Recommender systems support the decision making processes of customers with personalized suggestions. These widely used systems influence the daily life of almost everyone across domains like ecommerce, social media, and entertainment. However, the efficient generation of relevant recommendations in large-scale systems is a very complex task. In order to provide personalization, engines and algorithms need to capture users’ varying tastes and find mostly nonlinear dependencies between them and a multitude of items. Enormous data sparsity and ambitious real-time requirements further complicate this challenge. At the same time, deep learning has been proven to solve complex tasks like object or speech recognition where traditional machine learning failed or showed mediocre performance.
Explore a use case for vehicle recommendations at mobile.de, Germany’s biggest online vehicle market. Marcel shares a novel regularization technique for the optimization criterion and evaluates it against various baselines. To achieve high scalability, he combines this method with strategies for efficient candidate generation based on user and item embeddings—providing a holistic solution for candidate generation and ranking.
The proposed approach outperforms collaborative filtering and hybrid collaborative-content-based filtering by 73% and 143% for MAP@5. It also scales well for millions of items and users returning recommendations in tens of milliseconds.
In this talk we will explain some of the main challenges that we faced at OLX Europe while trying to proof the value of a deep learning based recommender system, and to later productionize it with a high level of automation.
We'll talk about:
* Modern Recommender Systems
* Deep Learning
* Neural Item Embeddings
* Similarity Search
* Proving value through Experimentation
* From POC to PRD
* Lessons Learned
About the speakers:
Cristian Martinez works as Lead Data Scientist at OLX Group, mainly focused on Search and Recommenders, and has been working for more than a decade in different companies solving business problems with Machine Learning.
Ilia Ivanov is a Data Scientist in OLX Europe (online marketplace) with 4 years of experience in DS focusing on recommendations and NLP.
Word Embeddings, Application of Sequence modelling, Recurrent neural network , drawback of recurrent neural networks, gated recurrent unit, long short term memory unit, Attention Mechanism
Feature engineering--the underdog of machine learning. This deck provides an overview of feature generation methods for text, image, audio, feature cleaning and transformation methods, how well they work and why.
Dowhy: An end-to-end library for causal inferenceAmit Sharma
In addition to efficient statistical estimators of a treatment's effect, successful application of causal inference requires specifying assumptions about the mechanisms underlying observed data and testing whether they are valid, and to what extent. However, most libraries for causal inference focus only on the task of providing powerful statistical estimators. We describe DoWhy, an open-source Python library that is built with causal assumptions as its first-class citizens, based on the formal framework of causal graphs to specify and test causal assumptions. DoWhy presents an API for the four steps common to any causal analysis---1) modeling the data using a causal graph and structural assumptions, 2) identifying whether the desired effect is estimable under the causal model, 3) estimating the effect using statistical estimators, and finally 4) refuting the obtained estimate through robustness checks and sensitivity analyses. In particular, DoWhy implements a number of robustness checks including placebo tests, bootstrap tests, and tests for unoberved confounding. DoWhy is an extensible library that supports interoperability with other implementations, such as EconML and CausalML for the the estimation step.
Recommender systems support the decision making processes of customers with personalized suggestions. These widely used systems influence the daily life of almost everyone across domains like ecommerce, social media, and entertainment. However, the efficient generation of relevant recommendations in large-scale systems is a very complex task. In order to provide personalization, engines and algorithms need to capture users’ varying tastes and find mostly nonlinear dependencies between them and a multitude of items. Enormous data sparsity and ambitious real-time requirements further complicate this challenge. At the same time, deep learning has been proven to solve complex tasks like object or speech recognition where traditional machine learning failed or showed mediocre performance.
Explore a use case for vehicle recommendations at mobile.de, Germany’s biggest online vehicle market. Marcel shares a novel regularization technique for the optimization criterion and evaluates it against various baselines. To achieve high scalability, he combines this method with strategies for efficient candidate generation based on user and item embeddings—providing a holistic solution for candidate generation and ranking.
The proposed approach outperforms collaborative filtering and hybrid collaborative-content-based filtering by 73% and 143% for MAP@5. It also scales well for millions of items and users returning recommendations in tens of milliseconds.
In this talk we will explain some of the main challenges that we faced at OLX Europe while trying to proof the value of a deep learning based recommender system, and to later productionize it with a high level of automation.
We'll talk about:
* Modern Recommender Systems
* Deep Learning
* Neural Item Embeddings
* Similarity Search
* Proving value through Experimentation
* From POC to PRD
* Lessons Learned
About the speakers:
Cristian Martinez works as Lead Data Scientist at OLX Group, mainly focused on Search and Recommenders, and has been working for more than a decade in different companies solving business problems with Machine Learning.
Ilia Ivanov is a Data Scientist in OLX Europe (online marketplace) with 4 years of experience in DS focusing on recommendations and NLP.
Word Embeddings, Application of Sequence modelling, Recurrent neural network , drawback of recurrent neural networks, gated recurrent unit, long short term memory unit, Attention Mechanism
Feature engineering--the underdog of machine learning. This deck provides an overview of feature generation methods for text, image, audio, feature cleaning and transformation methods, how well they work and why.
Dowhy: An end-to-end library for causal inferenceAmit Sharma
In addition to efficient statistical estimators of a treatment's effect, successful application of causal inference requires specifying assumptions about the mechanisms underlying observed data and testing whether they are valid, and to what extent. However, most libraries for causal inference focus only on the task of providing powerful statistical estimators. We describe DoWhy, an open-source Python library that is built with causal assumptions as its first-class citizens, based on the formal framework of causal graphs to specify and test causal assumptions. DoWhy presents an API for the four steps common to any causal analysis---1) modeling the data using a causal graph and structural assumptions, 2) identifying whether the desired effect is estimable under the causal model, 3) estimating the effect using statistical estimators, and finally 4) refuting the obtained estimate through robustness checks and sensitivity analyses. In particular, DoWhy implements a number of robustness checks including placebo tests, bootstrap tests, and tests for unoberved confounding. DoWhy is an extensible library that supports interoperability with other implementations, such as EconML and CausalML for the the estimation step.
Fine-tuning BERT for Question AnsweringApache MXNet
This deck covers the problem of fine-tuning a pre-trained BERT model for the task of Question Answering. Check out the GluonNLP model zoo here for models and tutorials: http://gluon-nlp.mxnet.io/model_zoo/bert/index.html
Slides: Thomas Delteil
A presentation on Bidirectional Encoder Representations from Transformers (BERT) meant to introduce the model's use cases and training mechanism. Best viewed with powerpoint since it contain many slide animations.
Tutorial on Sequence Aware Recommender Systems - ACM RecSys 2018Massimo Quadrana
Slides of the Tutorial on Sequence Aware Recommenders held at ACM RecSys 2018 in Vancouver.
Link to the website: https://sites.google.com/view/seq-recsys-tutorial
Link to the hands-on: https://github.com/mquad/sars_tutorial
Counterfactual Learning for RecommendationOlivier Jeunen
Slides for our presentation at the REVEAL workshop for RecSys '19 in Copenhagen and a Data Science Leuven Meetup, titled "Counterfactual Learning for Recommendation".
Slidedeck from our seminar about Machine Learning (07/11/2014)
Topics covered:
- What is Machine Learning?
- Techiques (clustering, classification, ...)
- Tools (Mahout, R, Spark MlLib, Weka, ...)
- Practical example of Machine Learning applications
- How to embed Machine Learning in software development
- Demo's
PR-297: Training data-efficient image transformers & distillation through att...Jinwon Lee
안녕하세요 TensorFlow Korea 논문 읽기 모임 PR-12의 297번째 리뷰입니다
어느덧 PR-12 시즌 3의 끝까지 논문 3편밖에 남지 않았네요.
시즌 3가 끝나면 바로 시즌 4의 새 멤버 모집이 시작될 예정입니다. 많은 관심과 지원 부탁드립니다~~
(멤버 모집 공지는 Facebook TensorFlow Korea 그룹에 올라올 예정입니다)
오늘 제가 리뷰한 논문은 Facebook의 Training data-efficient image transformers & distillation through attention 입니다.
Google에서 나왔던 ViT논문 이후에 convolution을 전혀 사용하지 않고 오직 attention만을 이용한 computer vision algorithm에 어느때보다 관심이 높아지고 있는데요
이 논문에서 제안한 DeiT 모델은 ViT와 같은 architecture를 사용하면서 ViT가 ImageNet data만으로는 성능이 잘 안나왔던 것에 비해서
Training 방법 개선과 새로운 Knowledge Distillation 방법을 사용하여 mageNet data 만으로 EfficientNet보다 뛰어난 성능을 보여주는 결과를 얻었습니다.
정말 CNN은 이제 서서히 사라지게 되는 것일까요? Attention이 computer vision도 정복하게 될 것인지....
개인적으로는 당분간은 attention 기반의 CV 논문이 쏟아질 거라고 확신하고, 또 여기에서 놀라운 일들이 일어날 수 있을 거라고 생각하고 있습니다
CNN은 10년간 많은 연구를 통해서 발전해왔지만, transformer는 이제 CV에 적용된 지 얼마 안된 시점이라서 더 기대가 크구요,
attention이 inductive bias가 가장 적은 형태의 모델이기 때문에 더 놀라운 이들을 만들 수 있을거라고 생각합니다
얼마 전에 나온 open AI의 DALL-E도 그 대표적인 예라고 할 수 있을 것 같습니다. Transformer의 또하나의 transformation이 궁금하신 분들은 아래 영상을 참고해주세요
영상링크: https://youtu.be/DjEvzeiWBTo
논문링크: https://arxiv.org/abs/2012.12877
Feature Engineering for ML - Dmitry Larko, H2O.aiSri Ambati
This talk was given at H2O World 2018 NYC and can be viewed here: https://youtu.be/wcFdmQSX6hM
Description:
In this talk, Dmitry shares his approach to feature engineering which he used successfully in various Kaggle competitions. He covers common techniques used to convert your features into numeric representation used by ML algorithms.
Speaker's Bio:
Dmitry has more than 10 years of experience in IT. Starting with data warehousing and BI, now in big data and data science. He has a lot of experience in predictive analytics software development for different domains and tasks. He is also a Kaggle Grandmaster who loves to use his machine learning and data science skills on Kaggle competitions.
Charlie Greenbacker, founder and co-organizer of the DC NLP meetup group, provides a "crash course" in Natural Language Processing techniques and applications.
What’s New with Databricks Machine LearningDatabricks
In this session, the Databricks product team provides a deeper dive into the machine learning announcements. Join us for a detailed demo that gives you insights into the latest innovations that simplify the ML lifecycle — from preparing data, discovering features, and training and managing models in production.
2019.11.09에 있었던 제주 GDG 발표 슬라이드입니다.
Blog : http://coffeedjimmy.github.io
Cooperation github repo: https://github.com/coffeedjimmy/Pytorch-TensorFlow2-Comparison
Fine-tuning BERT for Question AnsweringApache MXNet
This deck covers the problem of fine-tuning a pre-trained BERT model for the task of Question Answering. Check out the GluonNLP model zoo here for models and tutorials: http://gluon-nlp.mxnet.io/model_zoo/bert/index.html
Slides: Thomas Delteil
A presentation on Bidirectional Encoder Representations from Transformers (BERT) meant to introduce the model's use cases and training mechanism. Best viewed with powerpoint since it contain many slide animations.
Tutorial on Sequence Aware Recommender Systems - ACM RecSys 2018Massimo Quadrana
Slides of the Tutorial on Sequence Aware Recommenders held at ACM RecSys 2018 in Vancouver.
Link to the website: https://sites.google.com/view/seq-recsys-tutorial
Link to the hands-on: https://github.com/mquad/sars_tutorial
Counterfactual Learning for RecommendationOlivier Jeunen
Slides for our presentation at the REVEAL workshop for RecSys '19 in Copenhagen and a Data Science Leuven Meetup, titled "Counterfactual Learning for Recommendation".
Slidedeck from our seminar about Machine Learning (07/11/2014)
Topics covered:
- What is Machine Learning?
- Techiques (clustering, classification, ...)
- Tools (Mahout, R, Spark MlLib, Weka, ...)
- Practical example of Machine Learning applications
- How to embed Machine Learning in software development
- Demo's
PR-297: Training data-efficient image transformers & distillation through att...Jinwon Lee
안녕하세요 TensorFlow Korea 논문 읽기 모임 PR-12의 297번째 리뷰입니다
어느덧 PR-12 시즌 3의 끝까지 논문 3편밖에 남지 않았네요.
시즌 3가 끝나면 바로 시즌 4의 새 멤버 모집이 시작될 예정입니다. 많은 관심과 지원 부탁드립니다~~
(멤버 모집 공지는 Facebook TensorFlow Korea 그룹에 올라올 예정입니다)
오늘 제가 리뷰한 논문은 Facebook의 Training data-efficient image transformers & distillation through attention 입니다.
Google에서 나왔던 ViT논문 이후에 convolution을 전혀 사용하지 않고 오직 attention만을 이용한 computer vision algorithm에 어느때보다 관심이 높아지고 있는데요
이 논문에서 제안한 DeiT 모델은 ViT와 같은 architecture를 사용하면서 ViT가 ImageNet data만으로는 성능이 잘 안나왔던 것에 비해서
Training 방법 개선과 새로운 Knowledge Distillation 방법을 사용하여 mageNet data 만으로 EfficientNet보다 뛰어난 성능을 보여주는 결과를 얻었습니다.
정말 CNN은 이제 서서히 사라지게 되는 것일까요? Attention이 computer vision도 정복하게 될 것인지....
개인적으로는 당분간은 attention 기반의 CV 논문이 쏟아질 거라고 확신하고, 또 여기에서 놀라운 일들이 일어날 수 있을 거라고 생각하고 있습니다
CNN은 10년간 많은 연구를 통해서 발전해왔지만, transformer는 이제 CV에 적용된 지 얼마 안된 시점이라서 더 기대가 크구요,
attention이 inductive bias가 가장 적은 형태의 모델이기 때문에 더 놀라운 이들을 만들 수 있을거라고 생각합니다
얼마 전에 나온 open AI의 DALL-E도 그 대표적인 예라고 할 수 있을 것 같습니다. Transformer의 또하나의 transformation이 궁금하신 분들은 아래 영상을 참고해주세요
영상링크: https://youtu.be/DjEvzeiWBTo
논문링크: https://arxiv.org/abs/2012.12877
Feature Engineering for ML - Dmitry Larko, H2O.aiSri Ambati
This talk was given at H2O World 2018 NYC and can be viewed here: https://youtu.be/wcFdmQSX6hM
Description:
In this talk, Dmitry shares his approach to feature engineering which he used successfully in various Kaggle competitions. He covers common techniques used to convert your features into numeric representation used by ML algorithms.
Speaker's Bio:
Dmitry has more than 10 years of experience in IT. Starting with data warehousing and BI, now in big data and data science. He has a lot of experience in predictive analytics software development for different domains and tasks. He is also a Kaggle Grandmaster who loves to use his machine learning and data science skills on Kaggle competitions.
Charlie Greenbacker, founder and co-organizer of the DC NLP meetup group, provides a "crash course" in Natural Language Processing techniques and applications.
What’s New with Databricks Machine LearningDatabricks
In this session, the Databricks product team provides a deeper dive into the machine learning announcements. Join us for a detailed demo that gives you insights into the latest innovations that simplify the ML lifecycle — from preparing data, discovering features, and training and managing models in production.
2019.11.09에 있었던 제주 GDG 발표 슬라이드입니다.
Blog : http://coffeedjimmy.github.io
Cooperation github repo: https://github.com/coffeedjimmy/Pytorch-TensorFlow2-Comparison
JMI Techtalk: 강재욱 - Toward tf.keras from tf.estimator - From TensorFlow 2.0 p...Lablup Inc.
이 Techtalk에서는 TensorFlow 2.0으로 이전시 tf.estimator 에서 tf.keras로 이전해야 하는 이유에 대하여 설명합니다.
This Techtalk explains why you need to migrate from tf.estimator to tf.keras when moving to TensorFlow 2.0.
온라인 커머스에서 판매되는 상품을 embedding 하는 방법에 대한 논문입니다.
아마존에서 2017년에 발표한 논문이고 범용적으로 활용할 수 있는 product embedding을 제안했습니다.
multi-task 기반의 bidirection RNN을 활용했다는 점이 가장 큰 차별점입니다.
[IGC 2017] 넥스트플로어 김영수 - Protocol:hyperspace Diver 개발 포스트모템강 민우
본 세션에서는 Protocol:hyperspace Diver의 개발 과정 전반에 대한 포스트 모템을 수행하며 기획적인 부분을 바탕으로 제기된 요구사항에 대응하기 위한 기술적인 이슈에 어떻게 대응하였는지를 살펴볼 예정입니다. 게임을 기획하며 게임에 어떤 기능들이 요구되었으며, 엔진 레벨에서부터 모바일 게임을 개발하는 과정에서 이런 요구사항들에 어떻게 대응하였는지를 살펴봅니다. 게임을 위한 전체적인 설계 및 문제 해결 전략과 각각의 문제 해결 과정에서 세부 내용에 대한 기술적 노하우를 공유합니다.
실전 애자일 게임 개발 (Agile Game Agile Game Development From The Trenches)Kay Kim
Noel Llopis가 Montreal International Game Summit 2006에서 발표한 내용을 번역.
Agile 중에서 XP에 대해서 주로 다룸.
자세한 것은 http://betterways.tistory.com/51 참조.
Source: http://www.gamesfromwithin.com/articles/0611/000112.html