ROCm and Distributed Deep Learning on Spark and TensorFlowDatabricks
ROCm, the Radeon Open Ecosystem, is an open-source software foundation for GPU computing on Linux. ROCm supports TensorFlow and PyTorch using MIOpen, a library of highly optimized GPU routines for deep learning. In this talk, we describe how Apache Spark is a key enabling platform for distributed deep learning on ROCm, as it enables different deep learning frameworks to be embedded in Spark workflows in a secure end-to-end machine learning pipeline. We will analyse the different frameworks for integrating Spark with Tensorflow on ROCm, from Horovod to HopsML to Databrick's Project Hydrogen. We will also examine the surprising places where bottlenecks can surface when training models (everything from object stores to the Data Scientists themselves), and we will investigate ways to get around these bottlenecks. The talk will include a live demonstration of training and inference for a Tensorflow application embedded in a Spark pipeline written in a Jupyter notebook on Hopsworks with ROCm.
Using Deep Learning on Apache Spark to Diagnose Thoracic Pathology from Chest...Databricks
Overview and extended description: AI is expected to be the engine of technological advancements in the healthcare industry, especially in the areas of radiology and image processing. The purpose of this session is to demonstrate how we can build a AI-based Radiologist system using Apache Spark and Analytics Zoo to detect pneumonia and other diseases from chest x-ray images. The dataset, released by the NIH, contains around 110,00 X-ray images of around 30,000 unique patients, annotated with up to 14 different thoracic pathology labels. Stanford University developed a state-of-the-art model using CNN and exceeds average radiologist performance on the F1 metric. This talk focuses on how we can build a multi-label image classification model in a distributed Apache Spark infrastructure, and demonstrate how to build complex image transformations and deep learning pipelines using BigDL and Analytics Zoo with scalability and ease of use. Some practical image pre-processing procedures and evaluation metrics are introduced. We will also discuss runtime configuration, near-linear scalability for training and model serving, and other general performance topics.
How to use Apache TVM to optimize your ML modelsDatabricks
Apache TVM is an open source machine learning compiler that distills the largest, most powerful deep learning models into lightweight software that can run on the edge. This allows the outputed model to run inference much faster on a variety of target hardware (CPUs, GPUs, FPGAs & accelerators) and save significant costs.
In this deep dive, we’ll discuss how Apache TVM works, share the latest and upcoming features and run a live demo of how to optimize a custom machine learning model.
Koalas is an open source project that provides pandas APIs on top of Apache Spark. Pandas is the standard tool for data science and it is typically the first step to explore and manipulate a data set, but pandas does not scale well to big data. Koalas fills the gap by providing pandas equivalent APIs that work on Apache Spark.
There are also many libraries trying to scale pandas APIs, such as Vaex, Modin, and so on. Dask is one of them and very popular among pandas users, and also works on its own cluster similar to Koalas which is on top of Spark cluster. In this talk, we will introduce Koalas and its current status, and the comparison between Koalas and Dask, including benchmarking.
Deep Dive of ADBMS Migration to Apache Spark—Use Cases SharingDatabricks
eBay has been using enterprise ADBMS for over a decade, and our team is working on batch workload migration from ADBMS to Spark in 2018. There has been so many experiences and lessons we got during the whole migration journey (85% auto + 15% manual migration) - during which we exposed many unexpected issues and gaps between ADBMS and Spark SQL, we made a lot of decisions to fulfill the gaps in practice and contributed many fixes in Spark core in order to unblock ourselves. It will be a really interesting and should be helpful sharing for many folks especially data/software engineers to plan and execute their migration work. And during this session we will share many very specific issues each individually we encountered and how we resolve & work-around with team in real migration processes.
MLeap: Deploy Spark ML Pipelines to Production API ServersDataWorks Summit
MLeap is an open-source technology that allows Data Scientists and Engineers to deploy Spark-trained ML Pipelines and Models to a scoring engine instantly. During our presentation, we will show you how to deploy any Spark ML Pipeline, as well as custom transformers, that are trained using Spark streaming to both a cloud-based API server as well as an IoT device.
Why MLeap? Data Scientists use a myriad tools to analyze datasets, clean them and build offline models and validate their performance. The resulting scripts are thrown across the wall to Data Engineers and Architects whose job is to bring these pipelines to production. The Engineers are left with the unenviable job of not only reproducing the Data Scientists’ conclusions, but to scale the resulting pipeline both of which require a deep understanding of Data Science itself. As a result, most if not all Data Science deployments in the wild end up either too simplistic or take too long to productionize.
MLeap solves this problem for Spark users by providing serialization of ML Pipelines’ transformers to an MLeap Bundle, which is a graph-based serialization framework built on top of Protobuf 3 and JSON. In addition, MLeap also provides a highly optimized execution engine that doesn’t rely on the Spark-context, making inference blazing fast and is capable of executing one model or thousands of models in parallel.
ROCm and Distributed Deep Learning on Spark and TensorFlowDatabricks
ROCm, the Radeon Open Ecosystem, is an open-source software foundation for GPU computing on Linux. ROCm supports TensorFlow and PyTorch using MIOpen, a library of highly optimized GPU routines for deep learning. In this talk, we describe how Apache Spark is a key enabling platform for distributed deep learning on ROCm, as it enables different deep learning frameworks to be embedded in Spark workflows in a secure end-to-end machine learning pipeline. We will analyse the different frameworks for integrating Spark with Tensorflow on ROCm, from Horovod to HopsML to Databrick's Project Hydrogen. We will also examine the surprising places where bottlenecks can surface when training models (everything from object stores to the Data Scientists themselves), and we will investigate ways to get around these bottlenecks. The talk will include a live demonstration of training and inference for a Tensorflow application embedded in a Spark pipeline written in a Jupyter notebook on Hopsworks with ROCm.
Using Deep Learning on Apache Spark to Diagnose Thoracic Pathology from Chest...Databricks
Overview and extended description: AI is expected to be the engine of technological advancements in the healthcare industry, especially in the areas of radiology and image processing. The purpose of this session is to demonstrate how we can build a AI-based Radiologist system using Apache Spark and Analytics Zoo to detect pneumonia and other diseases from chest x-ray images. The dataset, released by the NIH, contains around 110,00 X-ray images of around 30,000 unique patients, annotated with up to 14 different thoracic pathology labels. Stanford University developed a state-of-the-art model using CNN and exceeds average radiologist performance on the F1 metric. This talk focuses on how we can build a multi-label image classification model in a distributed Apache Spark infrastructure, and demonstrate how to build complex image transformations and deep learning pipelines using BigDL and Analytics Zoo with scalability and ease of use. Some practical image pre-processing procedures and evaluation metrics are introduced. We will also discuss runtime configuration, near-linear scalability for training and model serving, and other general performance topics.
How to use Apache TVM to optimize your ML modelsDatabricks
Apache TVM is an open source machine learning compiler that distills the largest, most powerful deep learning models into lightweight software that can run on the edge. This allows the outputed model to run inference much faster on a variety of target hardware (CPUs, GPUs, FPGAs & accelerators) and save significant costs.
In this deep dive, we’ll discuss how Apache TVM works, share the latest and upcoming features and run a live demo of how to optimize a custom machine learning model.
Koalas is an open source project that provides pandas APIs on top of Apache Spark. Pandas is the standard tool for data science and it is typically the first step to explore and manipulate a data set, but pandas does not scale well to big data. Koalas fills the gap by providing pandas equivalent APIs that work on Apache Spark.
There are also many libraries trying to scale pandas APIs, such as Vaex, Modin, and so on. Dask is one of them and very popular among pandas users, and also works on its own cluster similar to Koalas which is on top of Spark cluster. In this talk, we will introduce Koalas and its current status, and the comparison between Koalas and Dask, including benchmarking.
Deep Dive of ADBMS Migration to Apache Spark—Use Cases SharingDatabricks
eBay has been using enterprise ADBMS for over a decade, and our team is working on batch workload migration from ADBMS to Spark in 2018. There has been so many experiences and lessons we got during the whole migration journey (85% auto + 15% manual migration) - during which we exposed many unexpected issues and gaps between ADBMS and Spark SQL, we made a lot of decisions to fulfill the gaps in practice and contributed many fixes in Spark core in order to unblock ourselves. It will be a really interesting and should be helpful sharing for many folks especially data/software engineers to plan and execute their migration work. And during this session we will share many very specific issues each individually we encountered and how we resolve & work-around with team in real migration processes.
MLeap: Deploy Spark ML Pipelines to Production API ServersDataWorks Summit
MLeap is an open-source technology that allows Data Scientists and Engineers to deploy Spark-trained ML Pipelines and Models to a scoring engine instantly. During our presentation, we will show you how to deploy any Spark ML Pipeline, as well as custom transformers, that are trained using Spark streaming to both a cloud-based API server as well as an IoT device.
Why MLeap? Data Scientists use a myriad tools to analyze datasets, clean them and build offline models and validate their performance. The resulting scripts are thrown across the wall to Data Engineers and Architects whose job is to bring these pipelines to production. The Engineers are left with the unenviable job of not only reproducing the Data Scientists’ conclusions, but to scale the resulting pipeline both of which require a deep understanding of Data Science itself. As a result, most if not all Data Science deployments in the wild end up either too simplistic or take too long to productionize.
MLeap solves this problem for Spark users by providing serialization of ML Pipelines’ transformers to an MLeap Bundle, which is a graph-based serialization framework built on top of Protobuf 3 and JSON. In addition, MLeap also provides a highly optimized execution engine that doesn’t rely on the Spark-context, making inference blazing fast and is capable of executing one model or thousands of models in parallel.
Context-aware Fast Food Recommendation with Ray on Apache Spark at Burger KingDatabricks
For fast food recommendation use cases, user behavior sequences and context features (such as time, weather, and location) are both important factors to be taken into consideration. At Burger King, we have developed a new state-of-the-art recommendation model called Transformer Cross Transformer (TxT). It applies Transformer encoders to capture both user behavior sequences and complicated context features and combines both transformers through the latent cross for joint context-aware fast food recommendations. Online A/B testings show not only the superiority of TxT comparing to existing methods results but also TxT can be successfully applied to other fast food recommendation use cases outside of Burger King.
Scaling Apache Spark MLlib to Billions of Parameters: Spark Summit East talk ...Spark Summit
Apache Spark MLlib provides scalable implementation of popular machine learning algorithms, which lets users train models from big dataset and iterate fast. The existing implementations assume that the number of parameters is small enough to fit in the memory of a single machine. However, many applications require solving problems with billions of parameters on a huge amount of data such as Ads CTR prediction and deep neural network. This requirement far exceeds the capacity of exisiting MLlib algorithms many of who use L-BFGS as the underlying solver. In order to fill this gap, we developed Vector-free L-BFGS for MLlib. It can solve optimization problems with billions of parameters in the Spark SQL framework where the training data are often generated. The algorithm scales very well and enables a variety of MLlib algorithms to handle a massive number of parameters over large datasets. In this talk, we will illustrate the power of Vector-free L-BFGS via logistic regression with real-world dataset and requirement. We will also discuss how this approach could be applied to other ML algorithms.
Apache Spark MLlib 2.0 Preview: Data Science and ProductionDatabricks
This talk highlights major improvements in Machine Learning (ML) targeted for Apache Spark 2.0. The MLlib 2.0 release focuses on ease of use for data science—both for casual and power users. We will discuss 3 key improvements: persisting models for production, customizing Pipelines, and improvements to models and APIs critical to data science.
(1) MLlib simplifies moving ML models to production by adding full support for model and Pipeline persistence. Individual models—and entire Pipelines including feature transformations—can be built on one Spark deployment, saved, and loaded onto other Spark deployments for production and serving.
(2) Users will find it much easier to implement custom feature transformers and models. Abstractions automatically handle input schema validation, as well as persistence for saving and loading models.
(3) For statisticians and data scientists, MLlib has doubled down on Generalized Linear Models (GLMs), which are key algorithms for many use cases. MLlib now supports more GLM families and link functions, handles corner cases more gracefully, and provides more model statistics. Also, expanded language APIs allow data scientists using Python and R to call many more algorithms.
Finally, we will demonstrate these improvements live and show how they facilitate getting started with ML on Spark, customizing implementations, and moving to production.
Lessons Learned while Implementing a Sparse Logistic Regression Algorithm in ...Spark Summit
This talk tells the story of implementation and optimization of a sparse logistic regression algorithm in spark. I would like to share the lessons I learned and the steps I had to take to improve the speed of execution and convergence of my initial naive implementation. The message isn’t to convince the audience that logistic regression is great and my implementation is awesome, rather it will give details about how it works under the hood, and general tips for implementing an iterative parallel machine learning algorithm in spark. The talk is structured as a sequence of “lessons learned” that are shown in form of code examples building on the initial naive implementation. The performance impact of each “lesson” on execution time and speed of convergence is measured on benchmark datasets. You will see how to formulate logistic regression in a parallel setting, how to avoid data shuffles, when to use a custom partitioner, how to use the ‘aggregate’ and ‘treeAggregate’ functions, how momentum can accelerate the convergence of gradient descent, and much more. I will assume basic understanding of machine learning and some prior knowledge of spark. The code examples are written in scala, and the code will be made available for each step in the walkthrough.
Building a SIMD Supported Vectorized Native Engine for Spark SQLDatabricks
Spark SQL works very well with structured row-based data. Vectorized reader and writer for parquet/orc can make I/O much faster. It also used WholeStageCodeGen to improve the performance by Java JIT code. However Java JIT is usually not working very well on utilizing latest SIMD instructions under complicated queries. Apache Arrow provides columnar in-memory layout and SIMD optimized kernels as well as a LLVM based SQL engine Gandiva. These native based libraries can accelerate Spark SQL by reduce the CPU usage for both I/O and execution.
Building an ML Platform with Ray and MLflowDatabricks
A successful machine learning platform allows ML practitioners to focus solely on their experiments and models and minimizes the time it takes to develop ML applications and take them to production. However, building an ML Platform is typically not an easy task due to the many different components involved in the process. In this talk, we will show how two open source projects, Ray (https://ray.io/) and MLflow (https://mlflow.org/), work together to make it easy for ML platform developers to add scaling and experiment management to their platform.
We will first provide an overview of Ray and its native libraries: Ray Tune (https://tune.io) for distributed hyperparameter tuning and Ray Serve (https://docs.ray.io/en/master/serve/index.html) for scalable model serving. Then we will showcase how MLflow provides a perfect solution for managing experiments through integrations with Ray for tracking and model deployment. Finally, we will finish with a demo of an ML platform built on Ray, MLflow, and other open source tools.
Spark Autotuning: Spark Summit East talk by Lawrence SpracklenSpark Summit
While the performance delivered by Spark has enabled data scientists to undertake sophisticated analyses on big and complex data in actionable timeframes, too often, the process of manually configuring the underlying Spark jobs (including the number and size of the executors) can be a significant and time consuming undertaking. Not only it does this configuration process typically rely heavily on repeated trial-and-error, it necessitates that data scientists have a low-level understanding of Spark and detailed cluster sizing information. At Alpine Data we have been working to eliminate this requirement, and develop algorithms that can be used to automatically tune Spark jobs with minimal user involvement,
In this presentation, we discuss the algorithms we have developed and illustrate how they leverage information about the size of the data being analyzed, the analytical operations being used in the flow, the cluster size, configuration and real-time utilization, to automatically determine the optimal Spark job configuration for peak performance.
When OLAP Meets Real-Time, What Happens in eBay?DataWorks Summit
OLAP Cube is about pre-aggregations, it reduces the query latency by spending more time and resources on data preparation. But for real-time analytics, data preparation and visibility latency are critical. What happens when OLAP cube meets real-time use cases?
Can we pre-build the cubes in real-time with a quick and more cost effective way? This is hard but still doable.
In eBay,we built our own real-time OLAP solution based on Apache Kylin & Apache Kafka. We read unbounded events from Kafka cluster then divide the streaming data into 3 stages, In-Memory Stage (Continuously In-Memory Aggregations) , On Disk Stage (Flush to disk, columnar based storage and indexes) and Full Cubing Stage (with MR or Spark, save to HBase). Data are aggregated to different layers in different stage, but all query able. Data will be transformed from 1 stage to another stage automatically and transparent to user.
This solution is built to support quite a few realtime analytics use cases in eBay, we will share some use cases like site speed monitoring and eBay site deal performance in this session as well.
Speaker:
Qiaoneng Qian, Senior Product Manager, eBay
Extending Machine Learning Algorithms with PySparkDatabricks
Machine learning practitioners are most comfortable using high-level programming languages such as Python. This is a barrier to parallelizing algorithms with big data frameworks such as Apache Spark, which are written in lower-level languages. Databricks partnered with the Regeneron Genetics Center to create the Glow library for population-scale genomics data storage and analytics. Glow V1.0.0 includes PySpark-based implementations for both existing and novel machine learning algorithms. We will discuss how leveraging tooling for Python users, especially Pandas UDFs, accelerated our development velocity and impacted our algorithms’ computational performance.
Scaling Machine Learning with Apache SparkDatabricks
Spark has become synonymous with big data processing, however the majority of data scientists still build models using single machine libraries. This talk will explore the multitude of ways Spark can be used to scale machine learning applications. In particular, we will guide you through distributed solutions for training and inference, distributed hyperparameter search, deployment issues, and new features for Machine Learning in Apache Spark 3.0. Niall Turbitt and Holly Smith combine their years of experience working with Spark to summarize best practices for scaling ML solutions.
Benchmark Tests and How-Tos of Convolutional Neural Network on HorovodRunner ...Databricks
The freedom of fast iterations of distributed deep learning tasks is crucial for smaller companies to gain competitive advantages and market shares from big tech giants. Horovod Runner brings this process to relatively accessible spark clusters.
Build, Scale, and Deploy Deep Learning Pipelines Using Apache SparkDatabricks
Deep Learning has shown a tremendous success, yet it often requires a lot of effort to leverage its power. Existing Deep Learning frameworks require writing a lot of code to work with a model, let alone in a distributed manner. We’ll survey the state of Deep Learning at scale, and where we introduce the Deep Learning Pipelines, a new open-source package for Apache Spark. This package simplifies Deep Learning in three major ways:
1. It has a simple API that integrates well with enterprise Machine Learning pipelines.
2. It automatically scales out common Deep Learning patterns, thanks to Apache Spark.
3. It enables exposing Deep Learning models through the familiar Spark APIs, such as MLlib and Spark SQL.
In this talk, we will look at a complex problem of image classification, using Deep Learning and Spark. Using Deep Learning Pipelines, we will show:
how to build deep learning models in a few lines of code;
how to scale common tasks like transfer learning and prediction; and how to publish models in Spark SQL.
Data Agility—A Journey to Advanced Analytics and Machine Learning at ScaleDatabricks
This talk will walk you through the typical workflow of a data scientist or a data analyst at Uber, how they get access to Uber's Big data and fast data sources for ad hoc and experimental analysis, how the data platforms will make it easy to discover datasets, run interactive queries against our petabyte scale data lake to identify the features you're interested in, wrangle and prepare data for advanced analytics and machine learning. Our platforms also provide capabilities to do iterative machine learning and deep learning training seamless on single nodes and distributed on our Big data and GPU clusters, analyze, visualize and share the results of their experiments with colleagues and peers to get feedback, and even productionize data analytics jobs and ML models all without a degree in CS. Interested? Come, learn how Uber's Big data platforms and Data science workbench put the power of Spark in the hands of our Data scientists and data analysts for advanced analytics and ML/DL use cases.
Hyperspace: An Indexing Subsystem for Apache SparkDatabricks
At Microsoft, we store datasets (both from internal teams and external customers) ranging from a few GBs to 100s of PBs in our data lake. The scope of analytics on these datasets ranges from traditional batch-style queries (e.g., OLAP) to explorative, ‘finding needle in a haystack’ type of queries (e.g., point-lookups, summarization etc.).
Operationalizing Machine Learning at Scale with Sameer NoriDatabricks
Machine learning has quickly become the hot new tool in the big data ecosystem. Virtually every organization is looking to leverage machine learning and build deeper and richer predictive analytics into their applications.
How does this work though, in practice? What are the challenges organizations run into as they look to move hundreds of models into production? How can they
make the age of both data and models closer to real-time?
This session will focus on how leading practitioners have been able to scale their machine learning deployments in production with the MapR Converged Data Platform.
Use cases that will be featured include autonomous cars and analytics as a service for retail and financial services.
Context-aware Fast Food Recommendation with Ray on Apache Spark at Burger KingDatabricks
For fast food recommendation use cases, user behavior sequences and context features (such as time, weather, and location) are both important factors to be taken into consideration. At Burger King, we have developed a new state-of-the-art recommendation model called Transformer Cross Transformer (TxT). It applies Transformer encoders to capture both user behavior sequences and complicated context features and combines both transformers through the latent cross for joint context-aware fast food recommendations. Online A/B testings show not only the superiority of TxT comparing to existing methods results but also TxT can be successfully applied to other fast food recommendation use cases outside of Burger King.
Scaling Apache Spark MLlib to Billions of Parameters: Spark Summit East talk ...Spark Summit
Apache Spark MLlib provides scalable implementation of popular machine learning algorithms, which lets users train models from big dataset and iterate fast. The existing implementations assume that the number of parameters is small enough to fit in the memory of a single machine. However, many applications require solving problems with billions of parameters on a huge amount of data such as Ads CTR prediction and deep neural network. This requirement far exceeds the capacity of exisiting MLlib algorithms many of who use L-BFGS as the underlying solver. In order to fill this gap, we developed Vector-free L-BFGS for MLlib. It can solve optimization problems with billions of parameters in the Spark SQL framework where the training data are often generated. The algorithm scales very well and enables a variety of MLlib algorithms to handle a massive number of parameters over large datasets. In this talk, we will illustrate the power of Vector-free L-BFGS via logistic regression with real-world dataset and requirement. We will also discuss how this approach could be applied to other ML algorithms.
Apache Spark MLlib 2.0 Preview: Data Science and ProductionDatabricks
This talk highlights major improvements in Machine Learning (ML) targeted for Apache Spark 2.0. The MLlib 2.0 release focuses on ease of use for data science—both for casual and power users. We will discuss 3 key improvements: persisting models for production, customizing Pipelines, and improvements to models and APIs critical to data science.
(1) MLlib simplifies moving ML models to production by adding full support for model and Pipeline persistence. Individual models—and entire Pipelines including feature transformations—can be built on one Spark deployment, saved, and loaded onto other Spark deployments for production and serving.
(2) Users will find it much easier to implement custom feature transformers and models. Abstractions automatically handle input schema validation, as well as persistence for saving and loading models.
(3) For statisticians and data scientists, MLlib has doubled down on Generalized Linear Models (GLMs), which are key algorithms for many use cases. MLlib now supports more GLM families and link functions, handles corner cases more gracefully, and provides more model statistics. Also, expanded language APIs allow data scientists using Python and R to call many more algorithms.
Finally, we will demonstrate these improvements live and show how they facilitate getting started with ML on Spark, customizing implementations, and moving to production.
Lessons Learned while Implementing a Sparse Logistic Regression Algorithm in ...Spark Summit
This talk tells the story of implementation and optimization of a sparse logistic regression algorithm in spark. I would like to share the lessons I learned and the steps I had to take to improve the speed of execution and convergence of my initial naive implementation. The message isn’t to convince the audience that logistic regression is great and my implementation is awesome, rather it will give details about how it works under the hood, and general tips for implementing an iterative parallel machine learning algorithm in spark. The talk is structured as a sequence of “lessons learned” that are shown in form of code examples building on the initial naive implementation. The performance impact of each “lesson” on execution time and speed of convergence is measured on benchmark datasets. You will see how to formulate logistic regression in a parallel setting, how to avoid data shuffles, when to use a custom partitioner, how to use the ‘aggregate’ and ‘treeAggregate’ functions, how momentum can accelerate the convergence of gradient descent, and much more. I will assume basic understanding of machine learning and some prior knowledge of spark. The code examples are written in scala, and the code will be made available for each step in the walkthrough.
Building a SIMD Supported Vectorized Native Engine for Spark SQLDatabricks
Spark SQL works very well with structured row-based data. Vectorized reader and writer for parquet/orc can make I/O much faster. It also used WholeStageCodeGen to improve the performance by Java JIT code. However Java JIT is usually not working very well on utilizing latest SIMD instructions under complicated queries. Apache Arrow provides columnar in-memory layout and SIMD optimized kernels as well as a LLVM based SQL engine Gandiva. These native based libraries can accelerate Spark SQL by reduce the CPU usage for both I/O and execution.
Building an ML Platform with Ray and MLflowDatabricks
A successful machine learning platform allows ML practitioners to focus solely on their experiments and models and minimizes the time it takes to develop ML applications and take them to production. However, building an ML Platform is typically not an easy task due to the many different components involved in the process. In this talk, we will show how two open source projects, Ray (https://ray.io/) and MLflow (https://mlflow.org/), work together to make it easy for ML platform developers to add scaling and experiment management to their platform.
We will first provide an overview of Ray and its native libraries: Ray Tune (https://tune.io) for distributed hyperparameter tuning and Ray Serve (https://docs.ray.io/en/master/serve/index.html) for scalable model serving. Then we will showcase how MLflow provides a perfect solution for managing experiments through integrations with Ray for tracking and model deployment. Finally, we will finish with a demo of an ML platform built on Ray, MLflow, and other open source tools.
Spark Autotuning: Spark Summit East talk by Lawrence SpracklenSpark Summit
While the performance delivered by Spark has enabled data scientists to undertake sophisticated analyses on big and complex data in actionable timeframes, too often, the process of manually configuring the underlying Spark jobs (including the number and size of the executors) can be a significant and time consuming undertaking. Not only it does this configuration process typically rely heavily on repeated trial-and-error, it necessitates that data scientists have a low-level understanding of Spark and detailed cluster sizing information. At Alpine Data we have been working to eliminate this requirement, and develop algorithms that can be used to automatically tune Spark jobs with minimal user involvement,
In this presentation, we discuss the algorithms we have developed and illustrate how they leverage information about the size of the data being analyzed, the analytical operations being used in the flow, the cluster size, configuration and real-time utilization, to automatically determine the optimal Spark job configuration for peak performance.
When OLAP Meets Real-Time, What Happens in eBay?DataWorks Summit
OLAP Cube is about pre-aggregations, it reduces the query latency by spending more time and resources on data preparation. But for real-time analytics, data preparation and visibility latency are critical. What happens when OLAP cube meets real-time use cases?
Can we pre-build the cubes in real-time with a quick and more cost effective way? This is hard but still doable.
In eBay,we built our own real-time OLAP solution based on Apache Kylin & Apache Kafka. We read unbounded events from Kafka cluster then divide the streaming data into 3 stages, In-Memory Stage (Continuously In-Memory Aggregations) , On Disk Stage (Flush to disk, columnar based storage and indexes) and Full Cubing Stage (with MR or Spark, save to HBase). Data are aggregated to different layers in different stage, but all query able. Data will be transformed from 1 stage to another stage automatically and transparent to user.
This solution is built to support quite a few realtime analytics use cases in eBay, we will share some use cases like site speed monitoring and eBay site deal performance in this session as well.
Speaker:
Qiaoneng Qian, Senior Product Manager, eBay
Extending Machine Learning Algorithms with PySparkDatabricks
Machine learning practitioners are most comfortable using high-level programming languages such as Python. This is a barrier to parallelizing algorithms with big data frameworks such as Apache Spark, which are written in lower-level languages. Databricks partnered with the Regeneron Genetics Center to create the Glow library for population-scale genomics data storage and analytics. Glow V1.0.0 includes PySpark-based implementations for both existing and novel machine learning algorithms. We will discuss how leveraging tooling for Python users, especially Pandas UDFs, accelerated our development velocity and impacted our algorithms’ computational performance.
Scaling Machine Learning with Apache SparkDatabricks
Spark has become synonymous with big data processing, however the majority of data scientists still build models using single machine libraries. This talk will explore the multitude of ways Spark can be used to scale machine learning applications. In particular, we will guide you through distributed solutions for training and inference, distributed hyperparameter search, deployment issues, and new features for Machine Learning in Apache Spark 3.0. Niall Turbitt and Holly Smith combine their years of experience working with Spark to summarize best practices for scaling ML solutions.
Benchmark Tests and How-Tos of Convolutional Neural Network on HorovodRunner ...Databricks
The freedom of fast iterations of distributed deep learning tasks is crucial for smaller companies to gain competitive advantages and market shares from big tech giants. Horovod Runner brings this process to relatively accessible spark clusters.
Build, Scale, and Deploy Deep Learning Pipelines Using Apache SparkDatabricks
Deep Learning has shown a tremendous success, yet it often requires a lot of effort to leverage its power. Existing Deep Learning frameworks require writing a lot of code to work with a model, let alone in a distributed manner. We’ll survey the state of Deep Learning at scale, and where we introduce the Deep Learning Pipelines, a new open-source package for Apache Spark. This package simplifies Deep Learning in three major ways:
1. It has a simple API that integrates well with enterprise Machine Learning pipelines.
2. It automatically scales out common Deep Learning patterns, thanks to Apache Spark.
3. It enables exposing Deep Learning models through the familiar Spark APIs, such as MLlib and Spark SQL.
In this talk, we will look at a complex problem of image classification, using Deep Learning and Spark. Using Deep Learning Pipelines, we will show:
how to build deep learning models in a few lines of code;
how to scale common tasks like transfer learning and prediction; and how to publish models in Spark SQL.
Data Agility—A Journey to Advanced Analytics and Machine Learning at ScaleDatabricks
This talk will walk you through the typical workflow of a data scientist or a data analyst at Uber, how they get access to Uber's Big data and fast data sources for ad hoc and experimental analysis, how the data platforms will make it easy to discover datasets, run interactive queries against our petabyte scale data lake to identify the features you're interested in, wrangle and prepare data for advanced analytics and machine learning. Our platforms also provide capabilities to do iterative machine learning and deep learning training seamless on single nodes and distributed on our Big data and GPU clusters, analyze, visualize and share the results of their experiments with colleagues and peers to get feedback, and even productionize data analytics jobs and ML models all without a degree in CS. Interested? Come, learn how Uber's Big data platforms and Data science workbench put the power of Spark in the hands of our Data scientists and data analysts for advanced analytics and ML/DL use cases.
Hyperspace: An Indexing Subsystem for Apache SparkDatabricks
At Microsoft, we store datasets (both from internal teams and external customers) ranging from a few GBs to 100s of PBs in our data lake. The scope of analytics on these datasets ranges from traditional batch-style queries (e.g., OLAP) to explorative, ‘finding needle in a haystack’ type of queries (e.g., point-lookups, summarization etc.).
Operationalizing Machine Learning at Scale with Sameer NoriDatabricks
Machine learning has quickly become the hot new tool in the big data ecosystem. Virtually every organization is looking to leverage machine learning and build deeper and richer predictive analytics into their applications.
How does this work though, in practice? What are the challenges organizations run into as they look to move hundreds of models into production? How can they
make the age of both data and models closer to real-time?
This session will focus on how leading practitioners have been able to scale their machine learning deployments in production with the MapR Converged Data Platform.
Use cases that will be featured include autonomous cars and analytics as a service for retail and financial services.
Spark After Dark - LA Apache Spark Users Group - Feb 2015Chris Fregly
Spark After Dark is a mock dating site that uses the latest Spark libraries including Spark SQL, BlinkDB, Tachyon, Spark Streaming, MLlib, and GraphX to generate high-quality dating recommendations for its members and blazing fast analytics for its operators.
We begin with brief overview of Spark, Spark Libraries, and Spark Use Cases. In addition, we'll discuss the modern day Lambda Architecture that combines real-time and batch processing into a single system. Lastly, we present best practices for monitoring and tuning a highly-available Spark and Spark Streaming cluster.
There will be many live demos covering everything from basic topics such as ETL and data ingestion to advanced topics such as streaming, sampling, approximations, machine learning, textual analysis, and graph processing.
Foundations for Scaling ML in Apache SparkDatabricks
Apache Spark has become the most active open source Big Data project, and its Machine Learning library MLlib has seen rapid growth in usage. A critical aspect of MLlib and Spark is the ability to scale: the same code used on a laptop can scale to 100’s or 1000’s of machines. This talk will describe ongoing and future efforts to make MLlib even faster and more scalable by integrating with two key initiatives in Spark. The first is Catalyst, the query optimizer underlying DataFrames and Datasets. The second is Tungsten, the project for approaching bare-metal speeds in Spark via memory management, cache-awareness, and code generation. This talk will discuss the goals, the challenges, and the benefits for MLlib users and developers. More generally, we will reflect on the importance of integrating ML with the many other aspects of big data analysis.
Since its debut in 2010, Apache Spark has become one of the most popular Big Data technologies in the Apache open source ecosystem. In addition to enabling processing of large data sets through its distributed computing architecture, Spark provides out-of-the-box support for machine learning, streaming and graph processing in a single framework. Spark has been supported by companies like Microsoft, Google, Amazon and IBM and in financial services, companies like Blackrock (http://bit.ly/1Q1DVJH ) and Bloomberg (http://bit.ly/29LXbPv ) have started to integrate Apache Spark into their tool chain and the interest is growing. Unlike other big-data technologies which require intensive programming using Java etc., Spark enables data scientists to work with a big-data technology using higher level languages like Python and R making it accessible to conduct experiments and for rapid prototyping.
In this talk, we will introduce Apache Spark and discuss the key features that differentiate Apache Spark from other technologies. We will provide examples on how Apache Spark can help scale analytics and discuss how the machine learning API could be used to solve large-scale machine learning problems using Spark’s distributed computing framework. We will also illustrate enterprise use cases for scaling analytics with Apache Spark.
Predicting online user behaviour using deep learning algorithmsArmando Vieira
We propose a robust classifier to predict buying intentions based on user behaviour within a large e-commerce website. In this work we compare traditional machine learning techniques with the most advanced deep learning approaches. We show that both Deep Belief Networks and Stacked Denoising auto-Encoders achieved a substantial improvement by extracting features from high dimensional data during the pre-train phase. They prove also to be more convenient to deal with severe class imbalance.
The talk by Maksud Ibrahimov, Chief Data Scientist at InfoReady Analytics. He is going to share with us how to maximise the performance of Spark.
As a user of Apache Spark from very early releases, he generally sees that the framework is easy to start with but as the program grows its performance starts to suffer. In this talk Maksud will answer the following questions:
- How to reach higher level of parallelism of your jobs without scaling up your cluster?
- Understanding shuffles, and how to avoid disk spills
- How to identify task stragglers and data skews?
- How to identify Spark bottlenecks?
Building Robust, Adaptive Streaming Apps with Spark StreamingDatabricks
As the adoption of Spark Streaming increases rapidly, the community has been asking for greater robustness and scalability from Spark Streaming applications in a wider range of operating environments. To fulfill these demands, we have steadily added a number of features in Spark Streaming. We have added backpressure mechanisms which allows Spark Streaming to dynamically adapt to changes in incoming data rates, and maintain stability of the application. In addition, we are extending Spark’s Dynamic Allocation to Spark Streaming, so that streaming applications can elastically scale based on processing requirements. In my talk, I am going to explore these mechanisms and explain how developers can write robust, scalable and adaptive streaming applications using them. Presented by Tathagata "TD" Das from Databricks.
Building machine learning service in your business — Eric Chen (Uber) @PAPIs ...PAPIs.io
When making machine learning applications in Uber, we identified a sequence of common practices and painful procedures, and thus built a machine learning platform as a service. We here present the key components to build such a scalable and reliable machine learning service which serves both our online and offline data processing needs.
Low Latency Polyglot Model Scoring using Apache ApexApache Apex
Data science is fast becoming a complementary approach and process to solve business challenges today. The explosion of frameworks to help data scientists build models bears a testimony to this. However when a model needs to be turned into a production version in very low latency and enterprise grade environments, there are a very few choices with each one having their own strengths and weaknesses. Adding to this is the current disconnect between a data scientists world which is all about modelling and an engineers world which is about SLAs and service guarantees. A framework like Apache Apex can complement each of these roles and provide constructs for both these worlds. This would help enterprises to drastically cut down the cost of model deployment to production environments.
This talk was held at the 10th meeting on February 3rd 2014 by Sean Owen.
Having collected Big Data, organizations are now keen on data science and “Big Learning”. Much of the focus has been on data science as exploratory analytics: offline, in the lab. However, building from that a production-ready large-scale operational analytics system remains a difficult and ad-hoc endeavor, especially when real-time answers are required. Design patterns for effective implementations are emerging, which take advantage of relaxed assumptions, adopt a new tiered "lambda" architecture, and pick the right scale-friendly algorithms to succeed. Drawing on experience from customer problems and the open source Oryx project at Cloudera, this session will provide examples of operational analytics projects in the field, and present a reference architecture and algorithm design choices for a successful implementation.
Modern Data Warehousing with the Microsoft Analytics Platform SystemJames Serra
The traditional data warehouse has served us well for many years, but new trends are causing it to break in four different ways: data growth, fast query expectations from users, non-relational/unstructured data, and cloud-born data. How can you prevent this from happening? Enter the modern data warehouse, which is able to handle and excel with these new trends. It handles all types of data (Hadoop), provides a way to easily interface with all these types of data (PolyBase), and can handle “big data” and provide fast queries. Is there one appliance that can support this modern data warehouse? Yes! It is the Analytics Platform System (APS) from Microsoft (formally called Parallel Data Warehouse or PDW) , which is a Massively Parallel Processing (MPP) appliance that has been recently updated (v2 AU1). In this session I will dig into the details of the modern data warehouse and APS. I will give an overview of the APS hardware and software architecture, identify what makes APS different, and demonstrate the increased performance. In addition I will discuss how Hadoop, HDInsight, and PolyBase fit into this new modern data warehouse.
Auto-Train a Time-Series Forecast Model With AML + ADBDatabricks
Supply Chain, Healthcare, Insurance, and Finance often require highly accurate forecasting models in an enterprise large-scale fashion. With Azure Machine Learning on Azure Databricks, the scale and speed to large-scale many-models can be achieved and time-to-product decreases drastically. The better-together story poses an enterprise approach to AI/ML.
Azure AutoML offers an elegant solution efficiently to build forecasting models on Azure Databricks compute solving sophisticated business problems. The presentation covers the Azure Machine Learning + Azure Databricks approach (see slides attached) while the demo covers a hands-on business problem building a forecasting model in Azure Databricks using Azure Machine Learning. The AI/ML better-together story is elevated as MLFlow for Data Science Lifecycle Management and Hyperopt for distributed model execution completes AI/ML enterprise readiness for industry problems.
DataMass Summit - Machine Learning for Big Data in SQL ServerŁukasz Grala
Sesja pokazująca zarówno Machine Learning Server (czyli algorytmy uczenia maszynowego w językach R i Python), ale także możliwość korzystania z danych JSON w SQL Server, czy też łączenia się do danych znajdujących się na HDFS, HADOOP, czy Spark poprzez Polybase w SQL Server, by te dane wykorzystywać do analizy, predykcji poprzez modele w językach R lub Python.
Scaling AI in production using PyTorchgeetachauhan
Slides from my talk at MLOps World' 21
Deploying AI models in production and scaling the ML services is still a big challenge. In this talk we will cover details of how to deploy your AI models, best practices for the deployment scenarios, and techniques for performance optimization and scaling the ML services. Come join us to learn how you can jumpstart the journey of taking your PyTorch models from Research to production.
Emerging technologies /frameworks in Big DataRahul Jain
A short overview presentation on Emerging technologies /frameworks in Big Data covering Apache Parquet, Apache Flink, Apache Drill with basic concepts of Columnar Storage and Dremel.
Applied Machine learning using H2O, python and R WorkshopAvkash Chauhan
Note: Get all workshop content at - https://github.com/h2oai/h2o-meetups/tree/master/2017_02_22_Seattle_STC_Meetup
Basic knowledge of R/python and general ML concepts
Note: This is bring-your-own-laptop workshop. Make sure you bring your laptop in order to be able to participate in the workshop
Level: 200
Time: 2 Hours
Agenda:
- Introduction to ML, H2O and Sparkling Water
- Refresher of data manipulation in R & Python
- Supervised learning
---- Understanding liner regression model with an example
---- Understanding binomial classification with an example
---- Understanding multinomial classification with an example
- Unsupervised learning
---- Understanding k-means clustering with an example
- Using machine learning models in production
- Sparkling Water Introduction & Demo
How to Productionize Your Machine Learning Models Using Apache Spark MLlib 2....Databricks
Apache Spark has rapidly become a key tool for data scientists to explore, understand and transform massive datasets and to build and train advanced machine learning models. The question then becomes, how do you deploy these ML model to a production environment? How do you embed what you’ve learned into customer facing data applications?
In this talk I will discuss best practices on how data scientists productionize machine learning models, do a deep dive with actual case studies, and show live tutorials of a few example architectures and code in Python, Scala, Java and SQL.
Building Analytic Apps for SaaS: “Analytics as a Service”Amazon Web Services
TIBCO Jaspersoft® for AWS is a business intelligence suite that helps you deliver stunning interactive reports and dashboards inside your app that make it easy for your customers to get answers. Purpose-built for AWS, our reporting and analytics server quickly and easily connects to Amazon Relational Database Service (RDS), Amazon Redshift, and Amazon EMR. It includes ad-hoc reporting, dashboards, data analysis, data visualization, and data blending. In less than 10 minutes, you can be analyzing and reporting on your data. You get a full Cloud BI server starting at less than $1/hour, with no user or data limits and no additional fees.
This webinar deck shows how embeddable analytics with TIBCO Jaspersoft for AWS gives you the power to create the experience your end users demand and how to scale and manage that experience across your customer base with AWS.
Navigating SAP’s Integration Options (Mastering SAP Technologies 2013)Sascha Wenninger
Provides an overview of popular integration approaches, maps them to SAP's integration tools and concludes with some lessons learnt in their application.
Hyper-Parameter Tuning Across the Entire AI Pipeline GPU Tech Conference San ...Chris Fregly
Chris Fregly, Founder @ PipelineAI, will walk you through a real-world, complete end-to-end Pipeline-optimization example. We highlight hyper-parameters - and model pipeline phases - that have never been exposed until now.
While most Hyperparameter Optimizers stop at the training phase (ie. learning rate, tree depth, ec2 instance type, etc), we extend model validation and tuning into a new post-training optimization phase including 8-bit reduced precision weight quantization and neural network layer fusing - among many other framework and hardware-specific optimizations.
Next, we introduce hyperparameters at the prediction phase including request-batch sizing and chipset (CPU v. GPU v. TPU).
Lastly, we determine a PipelineAI Efficiency Score of our overall Pipeline including Cost, Accuracy, and Time. We show techniques to maximize this PipelineAI Efficiency Score using our massive PipelineDB along with the Pipeline-wide hyper-parameter tuning techniques mentioned in this talk.
Bio
Chris Fregly is Founder and Applied AI Engineer at PipelineAI, a Real-Time Machine Learning and Artificial Intelligence Startup based in San Francisco.
He is also an Apache Spark Contributor, a Netflix Open Source Committer, founder of the Global Advanced Spark and TensorFlow Meetup, author of the O’Reilly Training and Video Series titled, "High Performance TensorFlow in Production with Kubernetes and GPUs."
Previously, Chris was a Distributed Systems Engineer at Netflix, a Data Solutions Engineer at Databricks, and a Founding Member and Principal Engineer at the IBM Spark Technology Center in San Francisco.
Data Scientists and Machine Learning practitioners, nowadays, seem to be churning out models by the dozen and they continuously experiment to find ways to improve their accuracies. They also use a variety of ML and DL frameworks & languages , and a typical organization may find that this results in a heterogenous, complicated bunch of assets that require different types of runtimes, resources and sometimes even specialized compute to operate efficiently.
But what does it mean for an enterprise to actually take these models to "production" ? How does an organization scale inference engines out & make them available for real-time applications without significant latencies ? There needs to be different techniques for batch (offline) inferences and instant, online scoring. Data needs to be accessed from various sources and cleansing, transformations of data needs to be enabled prior to any predictions. In many cases, there maybe no substitute for customized data handling with scripting either.
Enterprises also require additional auditing and authorizations built in, approval processes and still support a "continuous delivery" paradigm whereby a data scientist can enable insights faster. Not all models are created equal, nor are consumers of a model - so enterprises require both metering and allocation of compute resources for SLAs.
In this session, we will take a look at how machine learning is operationalized in IBM Data Science Experience (DSX), a Kubernetes based offering for the Private Cloud and optimized for the HortonWorks Hadoop Data Platform. DSX essentially brings in typical software engineering development practices to Data Science, organizing the dev->test->production for machine learning assets in much the same way as typical software deployments. We will also see what it means to deploy, monitor accuracies and even rollback models & custom scorers as well as how API based techniques enable consuming business processes and applications to remain relatively stable amidst all the chaos.
Speaker
Piotr Mierzejewski, Program Director Development IBM DSX Local, IBM
Similar to Scaling Machine Learning to Billions of Parameters - Spark Summit 2016 (20)
THE IMPORTANCE OF MARTIAN ATMOSPHERE SAMPLE RETURN.Sérgio Sacani
The return of a sample of near-surface atmosphere from Mars would facilitate answers to several first-order science questions surrounding the formation and evolution of the planet. One of the important aspects of terrestrial planet formation in general is the role that primary atmospheres played in influencing the chemistry and structure of the planets and their antecedents. Studies of the martian atmosphere can be used to investigate the role of a primary atmosphere in its history. Atmosphere samples would also inform our understanding of the near-surface chemistry of the planet, and ultimately the prospects for life. High-precision isotopic analyses of constituent gases are needed to address these questions, requiring that the analyses are made on returned samples rather than in situ.
Comparing Evolved Extractive Text Summary Scores of Bidirectional Encoder Rep...University of Maribor
Slides from:
11th International Conference on Electrical, Electronics and Computer Engineering (IcETRAN), Niš, 3-6 June 2024
Track: Artificial Intelligence
https://www.etran.rs/2024/en/home-english/
Observation of Io’s Resurfacing via Plume Deposition Using Ground-based Adapt...Sérgio Sacani
Since volcanic activity was first discovered on Io from Voyager images in 1979, changes
on Io’s surface have been monitored from both spacecraft and ground-based telescopes.
Here, we present the highest spatial resolution images of Io ever obtained from a groundbased telescope. These images, acquired by the SHARK-VIS instrument on the Large
Binocular Telescope, show evidence of a major resurfacing event on Io’s trailing hemisphere. When compared to the most recent spacecraft images, the SHARK-VIS images
show that a plume deposit from a powerful eruption at Pillan Patera has covered part
of the long-lived Pele plume deposit. Although this type of resurfacing event may be common on Io, few have been detected due to the rarity of spacecraft visits and the previously low spatial resolution available from Earth-based telescopes. The SHARK-VIS instrument ushers in a new era of high resolution imaging of Io’s surface using adaptive
optics at visible wavelengths.
Slide 1: Title Slide
Extrachromosomal Inheritance
Slide 2: Introduction to Extrachromosomal Inheritance
Definition: Extrachromosomal inheritance refers to the transmission of genetic material that is not found within the nucleus.
Key Components: Involves genes located in mitochondria, chloroplasts, and plasmids.
Slide 3: Mitochondrial Inheritance
Mitochondria: Organelles responsible for energy production.
Mitochondrial DNA (mtDNA): Circular DNA molecule found in mitochondria.
Inheritance Pattern: Maternally inherited, meaning it is passed from mothers to all their offspring.
Diseases: Examples include Leber’s hereditary optic neuropathy (LHON) and mitochondrial myopathy.
Slide 4: Chloroplast Inheritance
Chloroplasts: Organelles responsible for photosynthesis in plants.
Chloroplast DNA (cpDNA): Circular DNA molecule found in chloroplasts.
Inheritance Pattern: Often maternally inherited in most plants, but can vary in some species.
Examples: Variegation in plants, where leaf color patterns are determined by chloroplast DNA.
Slide 5: Plasmid Inheritance
Plasmids: Small, circular DNA molecules found in bacteria and some eukaryotes.
Features: Can carry antibiotic resistance genes and can be transferred between cells through processes like conjugation.
Significance: Important in biotechnology for gene cloning and genetic engineering.
Slide 6: Mechanisms of Extrachromosomal Inheritance
Non-Mendelian Patterns: Do not follow Mendel’s laws of inheritance.
Cytoplasmic Segregation: During cell division, organelles like mitochondria and chloroplasts are randomly distributed to daughter cells.
Heteroplasmy: Presence of more than one type of organellar genome within a cell, leading to variation in expression.
Slide 7: Examples of Extrachromosomal Inheritance
Four O’clock Plant (Mirabilis jalapa): Shows variegated leaves due to different cpDNA in leaf cells.
Petite Mutants in Yeast: Result from mutations in mitochondrial DNA affecting respiration.
Slide 8: Importance of Extrachromosomal Inheritance
Evolution: Provides insight into the evolution of eukaryotic cells.
Medicine: Understanding mitochondrial inheritance helps in diagnosing and treating mitochondrial diseases.
Agriculture: Chloroplast inheritance can be used in plant breeding and genetic modification.
Slide 9: Recent Research and Advances
Gene Editing: Techniques like CRISPR-Cas9 are being used to edit mitochondrial and chloroplast DNA.
Therapies: Development of mitochondrial replacement therapy (MRT) for preventing mitochondrial diseases.
Slide 10: Conclusion
Summary: Extrachromosomal inheritance involves the transmission of genetic material outside the nucleus and plays a crucial role in genetics, medicine, and biotechnology.
Future Directions: Continued research and technological advancements hold promise for new treatments and applications.
Slide 11: Questions and Discussion
Invite Audience: Open the floor for any questions or further discussion on the topic.
Deep Behavioral Phenotyping in Systems Neuroscience for Functional Atlasing a...Ana Luísa Pinho
Functional Magnetic Resonance Imaging (fMRI) provides means to characterize brain activations in response to behavior. However, cognitive neuroscience has been limited to group-level effects referring to the performance of specific tasks. To obtain the functional profile of elementary cognitive mechanisms, the combination of brain responses to many tasks is required. Yet, to date, both structural atlases and parcellation-based activations do not fully account for cognitive function and still present several limitations. Further, they do not adapt overall to individual characteristics. In this talk, I will give an account of deep-behavioral phenotyping strategies, namely data-driven methods in large task-fMRI datasets, to optimize functional brain-data collection and improve inference of effects-of-interest related to mental processes. Key to this approach is the employment of fast multi-functional paradigms rich on features that can be well parametrized and, consequently, facilitate the creation of psycho-physiological constructs to be modelled with imaging data. Particular emphasis will be given to music stimuli when studying high-order cognitive mechanisms, due to their ecological nature and quality to enable complex behavior compounded by discrete entities. I will also discuss how deep-behavioral phenotyping and individualized models applied to neuroimaging data can better account for the subject-specific organization of domain-general cognitive systems in the human brain. Finally, the accumulation of functional brain signatures brings the possibility to clarify relationships among tasks and create a univocal link between brain systems and mental functions through: (1) the development of ontologies proposing an organization of cognitive processes; and (2) brain-network taxonomies describing functional specialization. To this end, tools to improve commensurability in cognitive science are necessary, such as public repositories, ontology-based platforms and automated meta-analysis tools. I will thus discuss some brain-atlasing resources currently under development, and their applicability in cognitive as well as clinical neuroscience.
Salas, V. (2024) "John of St. Thomas (Poinsot) on the Science of Sacred Theol...Studia Poinsotiana
I Introduction
II Subalternation and Theology
III Theology and Dogmatic Declarations
IV The Mixed Principles of Theology
V Virtual Revelation: The Unity of Theology
VI Theology as a Natural Science
VII Theology’s Certitude
VIII Conclusion
Notes
Bibliography
All the contents are fully attributable to the author, Doctor Victor Salas. Should you wish to get this text republished, get in touch with the author or the editorial committee of the Studia Poinsotiana. Insofar as possible, we will be happy to broker your contact.
Earliest Galaxies in the JADES Origins Field: Luminosity Function and Cosmic ...Sérgio Sacani
We characterize the earliest galaxy population in the JADES Origins Field (JOF), the deepest
imaging field observed with JWST. We make use of the ancillary Hubble optical images (5 filters
spanning 0.4−0.9µm) and novel JWST images with 14 filters spanning 0.8−5µm, including 7 mediumband filters, and reaching total exposure times of up to 46 hours per filter. We combine all our data
at > 2.3µm to construct an ultradeep image, reaching as deep as ≈ 31.4 AB mag in the stack and
30.3-31.0 AB mag (5σ, r = 0.1” circular aperture) in individual filters. We measure photometric
redshifts and use robust selection criteria to identify a sample of eight galaxy candidates at redshifts
z = 11.5 − 15. These objects show compact half-light radii of R1/2 ∼ 50 − 200pc, stellar masses of
M⋆ ∼ 107−108M⊙, and star-formation rates of SFR ∼ 0.1−1 M⊙ yr−1
. Our search finds no candidates
at 15 < z < 20, placing upper limits at these redshifts. We develop a forward modeling approach to
infer the properties of the evolving luminosity function without binning in redshift or luminosity that
marginalizes over the photometric redshift uncertainty of our candidate galaxies and incorporates the
impact of non-detections. We find a z = 12 luminosity function in good agreement with prior results,
and that the luminosity function normalization and UV luminosity density decline by a factor of ∼ 2.5
from z = 12 to z = 14. We discuss the possible implications of our results in the context of theoretical
models for evolution of the dark matter halo mass function.
4. 1.2 1.2 -78
6.3 -8.1
5.4 -8
4.2 2.3 -3.4
-1.1
2.3 4.9 7.4
4.5 2.1 -15
2.3 2.3
0.5 1.2 -0.9
-24
-1.3 -2.2 1.8
-4.9 -2.1 1.2
Web Scale ML
Billions of features
Hundredsofbillionsofexamples
Big Model
BigData
Ex: Yahoo word2vec - 120 billion parameters and 500 billion samples
5. 1.2 1.2 -78
6.3 -8.1
5.4 -8
4.2 2.3 -3.4
-1.1
2.3 4.9 7.4
4.5 2.1 -15
2.3 2.3
0.5 1.2 -0.9
-24
-1.3 -2.2 1.8
-4.9 -2.1 1.2
Web Scale ML
Billions of features
Hundredsofbillionsofexamples
Big Model
BigData
Ex: Yahoo word2vec - 120 billion parameters and 500 billion samples
Store Store Store
6. 1.2 1.2 -78
6.3 -8.1
5.4 -8
4.2 2.3 -3.4
-1.1
2.3 4.9 7.4
4.5 2.1 -15
2.3 2.3
0.5 1.2 -0.9
-24
-1.3 -2.2 1.8
-4.9 -2.1 1.2
Web Scale ML
Billions of features
Hundredsofbillionsofexamples
Big Model
BigData
Ex: Yahoo word2vec - 120 billion parameters and 500 billion samples
Worker
Worker
Worker
Store Store Store
7. 1.2 1.2 -78
6.3 -8.1
5.4 -8
4.2 2.3 -3.4
-1.1
2.3 4.9 7.4
4.5 2.1 -15
2.3 2.3
0.5 1.2 -0.9
-24
-1.3 -2.2 1.8
-4.9 -2.1 1.2
Web Scale ML
Billions of features
Hundredsofbillionsofexamples
Big Model
BigData
Ex: Yahoo word2vec - 120 billion parameters and 500 billion samples
Worker
Worker
Worker
Store Store Store
Each example depends only on a tiny fraction of the model
8. Two Optimization Strategies
Model
Multiple epochs…
BATCH
Example: Gradient Descent, L-BFGS
Model
Model
Model
SEQUENTIAL
Multiple random samples…
Example: (Minibatch) stochastic gradient method,
perceptron
Examples
9. Two Optimization Strategies
Model
Multiple epochs…
BATCH
Example: Gradient Descent, L-BFGS
Model
Model
Model
SEQUENTIAL
Multiple random samples…
Example: (Minibatch) stochastic gradient method,
perceptron
• Small number of model updates
• Accurate
• Each epoch may be expensive.
• Easy to parallelize.
Examples
10. Two Optimization Strategies
Model
Multiple epochs…
BATCH
Example: Gradient Descent, L-BFGS
Model
Model
Model
SEQUENTIAL
Multiple random samples…
Example: (Minibatch) stochastic gradient method,
perceptron
• Small number of model updates
• Accurate
• Each epoch may be expensive.
• Easy to parallelize.
• Requires lots of model updates.
• Not as accurate, but often good enough
• A lot of progress in one pass* for big data.
• Not trivial to parallelize.
*also optimal in terms of generalization error (often with a lot of tuning)
Examples
13. Requirements
✓ Support both batch and sequential optimization
✓ Sequential training: Handle frequent updates to the model
14. Requirements
✓ Support both batch and sequential optimization
✓ Sequential training: Handle frequent updates to the model
✓ Batch training: 100+ passes each pass must be fast.
15. Parameter Server (PS)
Client
Data
Client
Data
Client
Data
Client
Data
Training state stored in PS shards, asynchronous updates
PS Shard PS ShardPS Shard
ΔM
Model Update
M
Model
Early work: Yahoo LDA by Smola and Narayanamurthy based on memcached (2010),
Introduced in Google’s Distbelief (2012), refined in Petuum / Bösen (2013), Mu Li et al (2014)
17. ML in Spark alone
Executor Executor
CoreCore Core Core Core
Driver
Holds model
18. ML in Spark alone
Executor Executor
CoreCore Core Core Core
Driver
Holds model
MLlib optimization
19. ML in Spark alone
• Sequential:
– Driver-based communication limits frequency of model updates.
– Large minibatch size limits model update frequency, convergence suffers.
Executor Executor
CoreCore Core Core Core
Driver
Holds model
MLlib optimization
20. ML in Spark alone
• Sequential:
– Driver-based communication limits frequency of model updates.
– Large minibatch size limits model update frequency, convergence suffers.
• Batch:
– Driver bandwidth can be a bottleneck
– Synchronous stage wise processing limits throughput.
Executor Executor
CoreCore Core Core Core
Driver
Holds model
MLlib optimization
21. ML in Spark alone
• Sequential:
– Driver-based communication limits frequency of model updates.
– Large minibatch size limits model update frequency, convergence suffers.
• Batch:
– Driver bandwidth can be a bottleneck
– Synchronous stage wise processing limits throughput.
Executor Executor
CoreCore Core Core Core
Driver
Holds model
MLlib optimization
PS Architecture circumvents both limitations…
23. • Leverage Spark for HDFS I/O, distributed processing, fine-grained
load balancing, failure recovery, in-memory operations
Spark + Parameter Server
24. • Leverage Spark for HDFS I/O, distributed processing, fine-grained
load balancing, failure recovery, in-memory operations
• Use PS to sync models, incremental updates during training, or
sometimes even some vector math.
Spark + Parameter Server
25. • Leverage Spark for HDFS I/O, distributed processing, fine-grained
load balancing, failure recovery, in-memory operations
• Use PS to sync models, incremental updates during training, or
sometimes even some vector math.
Spark + Parameter Server
HDFS
Training state stored in PS shards
Driver
Executor ExecutorExecutor
CoreCore Core Core
PS Shard PS ShardPS Shard
control
control
33. Map PS API
• Distributed key-value store abstraction
• Supports batched operations in addition to usual get and put
• Many operations return a future – you can operate asynchronously or block
34. Matrix PS API
• Vector math (BLAS style operations), in addition to everything Map API provides
• Increment and fetch sparse vectors (e.g., for gradient aggregation)
• We use other custom operations on shard (API not shown)
42. L-BFGS Background
Exact, impractical
Step Size computation
- Needs to satisfy some technical (Wolfe) conditions
- Adaptively determined from data
Inverse Hessian Approximation
(based on history of L-previous gradients and model deltas)
Approximate, practical
Newton’s method
Gradient Descent
Using curvature information,
you can converge faster…
43. L-BFGS Background
Exact, impractical
Step Size computation
- Needs to satisfy some technical (Wolfe) conditions
- Adaptively determined from data
Inverse Hessian Approximation
(based on history of L-previous gradients and model deltas)
Approximate, practical
Newton’s method
Gradient Descent
Using curvature information,
you can converge faster…
44. L-BFGS Background
Exact, impractical
Step Size computation
- Needs to satisfy some technical (Wolfe) conditions
- Adaptively determined from data
Inverse Hessian Approximation
(based on history of L-previous gradients and model deltas)
Approximate, practical
Newton’s method
Gradient Descent
Using curvature information,
you can converge faster…
45. L-BFGS Background
Exact, impractical
Step Size computation
- Needs to satisfy some technical (Wolfe) conditions
- Adaptively determined from data
Inverse Hessian Approximation
(based on history of L-previous gradients and model deltas)
Approximate, practical
Newton’s method
Gradient Descent
Using curvature information,
you can converge faster…
dotprod
axpy (y ← ax + y)
copy
axpy
scal
scal
Vector Math
dotprod
46. L-BFGS Background
Exact, impractical
Step Size computation
- Needs to satisfy some technical (Wolfe) conditions
- Adaptively determined from data
Inverse Hessian Approximation
(based on history of L-previous gradients and model deltas)
Approximate, practical
Newton’s method
Gradient Descent
Using curvature information,
you can converge faster…
47. Executor ExecutorExecutor
HDFS HDFSHDFS
Driver
PS PS PS PS
Distributed LBFGS*
Compute gradient and loss
1. Incremental sparse gradient update
2. Fetch sparse portions of model
Coordinates executor
Step 1: Compute and update Gradient
*Our design is very similar to Sandblaster L-BFGS, Jeff Dean et al, Large Scale Distributed Deep Networks (2012)
state vectors
48. Executor ExecutorExecutor
HDFS HDFSHDFS
Driver
PS PS PS PS
Distributed LBFGS*
Compute gradient and loss
1. Incremental sparse gradient update
2. Fetch sparse portions of model
Coordinates executor
Step 1: Compute and update Gradient
*Our design is very similar to Sandblaster L-BFGS, Jeff Dean et al, Large Scale Distributed Deep Networks (2012)
state vectors
54. Speedup tricks
• Intersperse communication and computation
• Quicker convergence
– Parallel line search for step size
– Curvature for initial Hessian approximation*
*borrowed from vowpal wabbit
55. Speedup tricks
• Intersperse communication and computation
• Quicker convergence
– Parallel line search for step size
– Curvature for initial Hessian approximation*
• Network bandwidth reduction
– Compressed integer arrays
– Only store indices for binary data
*borrowed from vowpal wabbit
56. Speedup tricks
• Intersperse communication and computation
• Quicker convergence
– Parallel line search for step size
– Curvature for initial Hessian approximation*
• Network bandwidth reduction
– Compressed integer arrays
– Only store indices for binary data
• Matrix math on minibatch
*borrowed from vowpal wabbit
57. Speedup tricks
• Intersperse communication and computation
• Quicker convergence
– Parallel line search for step size
– Curvature for initial Hessian approximation*
• Network bandwidth reduction
– Compressed integer arrays
– Only store indices for binary data
• Matrix math on minibatch
0
750
1500
2250
3000
10
20
100
221612
2880
1260
96
MLlib
PS + Spark
1.6 x 108 examples, 100 executors, 10 cores
time(inseconds)perepoch
feature size (millions)
*borrowed from vowpal wabbit
67. Word2vec
• Skipgram with negative sampling:
– training set includes pairs of words and neighbors in corpus,
along with randomly selected words for each neighbor
68. Word2vec
• Skipgram with negative sampling:
– training set includes pairs of words and neighbors in corpus,
along with randomly selected words for each neighbor
– determine w → u(w),v(w) so that sigmoid(u(w)•v(w’)) is close
to (minimizes log loss) the probability that w’ is a neighbor of
w as opposed to a randomly selected word.
69. Word2vec
• Skipgram with negative sampling:
– training set includes pairs of words and neighbors in corpus,
along with randomly selected words for each neighbor
– determine w → u(w),v(w) so that sigmoid(u(w)•v(w’)) is close
to (minimizes log loss) the probability that w’ is a neighbor of
w as opposed to a randomly selected word.
– SGD involves computing many vector dot products e.g.,
u(w)•v(w’) and vector linear combinations
e.g., u(w) += α v(w’).
70. Word2vec Application at Yahoo
• Example training data:
gas_cap_replacement_for_car
slc_679f037df54f5d9c41cab05bfae0926
gas_door_replacement_for_car
slc_466145af16a40717c84683db3f899d0a fuel_door_covers
adid_c_28540527225_285898621262
slc_348709d73214fdeb9782f8b71aff7b6e autozone_auto_parts
adid_b_3318310706_280452370893 auoto_zone
slc_8dcdab5d20a2caa02b8b1d1c8ccbd36b
slc_58f979b6deb6f40c640f7ca8a177af2d
[ Grbovic, et. al. SIGIR 2015 and SIGIR 2016 (to appear) ]
73. Distributed Word2vec
• Needed system to train 200 million 300
dimensional word2vec model using minibatch
SGD
• Achieved in a high throughput and network
efficient way using our matrix based PS server:
74. Distributed Word2vec
• Needed system to train 200 million 300
dimensional word2vec model using minibatch
SGD
• Achieved in a high throughput and network
efficient way using our matrix based PS server:
– Vectors don’t go over network.
75. Distributed Word2vec
• Needed system to train 200 million 300
dimensional word2vec model using minibatch
SGD
• Achieved in a high throughput and network
efficient way using our matrix based PS server:
– Vectors don’t go over network.
– Most compute on PS servers, with clients aggregating
partial results from shards.
83. Distributed Word2vec
• Network lower by factor of #shards/dimension
compared to conventional PS based system
(1/20 to 1/100 for useful scenarios).
84. Distributed Word2vec
• Network lower by factor of #shards/dimension
compared to conventional PS based system
(1/20 to 1/100 for useful scenarios).
• Trains 200 million vocab, 55 billion word search
session in 2.5 days.
85. Distributed Word2vec
• Network lower by factor of #shards/dimension
compared to conventional PS based system
(1/20 to 1/100 for useful scenarios).
• Trains 200 million vocab, 55 billion word search
session in 2.5 days.
• In production for regular training in Yahoo search
ad serving system.
86. Other Projects using Spark + PS
• Online learning on PS
– Personalization as a Service
– Sponsored Search
• Factorization Machines
– Large scale user profiling
94. Summary
• Parameter server indispensable for big models
• Spark + Parameter Server has proved to be very
flexible platform for our large scale computing
needs
• Direct computation on the parameter servers
accelerate training for our use-cases