AI has played a limited role in the COVID-19 pandemic so far, scoring a B- according to one expert. It has helped in some areas like early warning, image-based diagnosis, and optimizing clinical trials. However, it could not demonstrate great impact in regions with complex healthcare systems and high inertia. Going forward, AI may accelerate tasks like forecasting medical resource needs, optimizing logistics, and assisting vaccine and drug discovery for future pandemics if developed with proper objectives, less reliance on historical data, and alignment with human values.
Data pipeline and data lake for autonomous drivingYu Huang
This document outlines autonomous driving data pipelines and data lakes used by various companies. It discusses Tesla, Google Waymo, PlusAI, Alibaba Cloud, Nvidia, NetApp, Amazon AWS, Amazon TRI, Amazon Momenta, and data pipeline strategies from Eckerson DataOps and IBM. The document also provides a detailed overview of an autonomous driving data lake built on AWS that ingests vehicle telemetry data and processes drive data for labeling and search capabilities.
How to integrate salesforce data with azure data factoryservicesNitor
Integrating Salesforce data with Azure Data Factory is a great way to analyze your customers and get insights into their needs. Read our blog to learn more.
TAME: Trainable Attention Mechanism for ExplanationsVasileiosMezaris
Presentation of paper "TAME: Attention Mechanism Based Feature Fusion for Generating Explanation Maps of Convolutional Neural Networks", by M. Ntrougkas, N. Gkalelis, V. Mezaris, delivered at IEEE ISM 2022, Dec. 2022, Naples, Italy.
The apparent “black box” nature of neural networks is a barrier to adoption in applications where explainability is essential. This paper presents TAME (Trainable Attention Mechanism for Explanations), a method for generating explanation maps with a multi-branch hierarchical attention mechanism. TAME combines a target model’s feature maps from multiple layers using an attention mechanism, transforming them into an explanation map. TAME can easily be applied to any convolutional neural network (CNN) by streamlining the optimization of the attention mechanism’s training method and the selection
of target model’s feature maps. After training, explanation maps can be computed in a single forward pass. We apply TAME to two widely used models, i.e. VGG-16 and ResNet-50, trained on ImageNet and show improvements over previous top-performing methods. We also provide a comprehensive ablation study comparing the performance of different variations of TAME’s architecture.
When it comes to AI use for prediction, diagnosis and treatment of medical conditions, reality is often replaced with a hype. Limitations should be known. A review of AI failures and challenges in healthcare showing why it is not likely for algorithms to replace physicians in the nearest future.
Abbott overview medical device human factors standardsJones Wu
The document provides an overview of international medical device human factors standards, outlining their history and purposes. It summarizes that IEC 62366:2007 is now the key usability engineering standard, with IEC 60601-1-6:2010 pointing to it, and HE-74 being incorporated into IEC 62366 as an annex. Future revisions are planned to split and update IEC 62366 by 2014 to address additional topics and harmonize with FDA guidance.
Enabling Edge Processing & Surgical Suite Integration with AWS Snowball Edge ...Amazon Web Services
In this session, we walk through the architecture and key learnings of Alcon, the developer of SMART Suite, a digital platform designed to streamline and simplify cataract surgery for surgeons and patients. Leveraging Philips’ HSDP Edge (based on AWS Snowball Edge), Alcon deploys a local access point for IoT devices to locally connect to in order to access the discovery, proxy, data broker, and firmware services. By hosting these services locally at the site, Alcon achieves faster response times to local events and can continue to operate in the event of a loss of connectivity. Snowball Edge is also a consolidation point for local data sets that will be asynchronously replicated to the cloud.
The document introduces artificial immune systems (AIS), which are computational systems inspired by the human immune system. It provides an overview of the immune system and its properties such as diversity, learning, memory, pattern recognition, and self/non-self discrimination. These properties provide a biological paradigm for developing AIS algorithms. The document then discusses representation schemes, affinity measures, and generic algorithms that have been developed for AIS, including negative selection, clonal selection, and immune network models. Finally, it reviews applications of AIS and discusses current trends in the field.
AI has played a limited role in the COVID-19 pandemic so far, scoring a B- according to one expert. It has helped in some areas like early warning, image-based diagnosis, and optimizing clinical trials. However, it could not demonstrate great impact in regions with complex healthcare systems and high inertia. Going forward, AI may accelerate tasks like forecasting medical resource needs, optimizing logistics, and assisting vaccine and drug discovery for future pandemics if developed with proper objectives, less reliance on historical data, and alignment with human values.
Data pipeline and data lake for autonomous drivingYu Huang
This document outlines autonomous driving data pipelines and data lakes used by various companies. It discusses Tesla, Google Waymo, PlusAI, Alibaba Cloud, Nvidia, NetApp, Amazon AWS, Amazon TRI, Amazon Momenta, and data pipeline strategies from Eckerson DataOps and IBM. The document also provides a detailed overview of an autonomous driving data lake built on AWS that ingests vehicle telemetry data and processes drive data for labeling and search capabilities.
How to integrate salesforce data with azure data factoryservicesNitor
Integrating Salesforce data with Azure Data Factory is a great way to analyze your customers and get insights into their needs. Read our blog to learn more.
TAME: Trainable Attention Mechanism for ExplanationsVasileiosMezaris
Presentation of paper "TAME: Attention Mechanism Based Feature Fusion for Generating Explanation Maps of Convolutional Neural Networks", by M. Ntrougkas, N. Gkalelis, V. Mezaris, delivered at IEEE ISM 2022, Dec. 2022, Naples, Italy.
The apparent “black box” nature of neural networks is a barrier to adoption in applications where explainability is essential. This paper presents TAME (Trainable Attention Mechanism for Explanations), a method for generating explanation maps with a multi-branch hierarchical attention mechanism. TAME combines a target model’s feature maps from multiple layers using an attention mechanism, transforming them into an explanation map. TAME can easily be applied to any convolutional neural network (CNN) by streamlining the optimization of the attention mechanism’s training method and the selection
of target model’s feature maps. After training, explanation maps can be computed in a single forward pass. We apply TAME to two widely used models, i.e. VGG-16 and ResNet-50, trained on ImageNet and show improvements over previous top-performing methods. We also provide a comprehensive ablation study comparing the performance of different variations of TAME’s architecture.
When it comes to AI use for prediction, diagnosis and treatment of medical conditions, reality is often replaced with a hype. Limitations should be known. A review of AI failures and challenges in healthcare showing why it is not likely for algorithms to replace physicians in the nearest future.
Abbott overview medical device human factors standardsJones Wu
The document provides an overview of international medical device human factors standards, outlining their history and purposes. It summarizes that IEC 62366:2007 is now the key usability engineering standard, with IEC 60601-1-6:2010 pointing to it, and HE-74 being incorporated into IEC 62366 as an annex. Future revisions are planned to split and update IEC 62366 by 2014 to address additional topics and harmonize with FDA guidance.
Enabling Edge Processing & Surgical Suite Integration with AWS Snowball Edge ...Amazon Web Services
In this session, we walk through the architecture and key learnings of Alcon, the developer of SMART Suite, a digital platform designed to streamline and simplify cataract surgery for surgeons and patients. Leveraging Philips’ HSDP Edge (based on AWS Snowball Edge), Alcon deploys a local access point for IoT devices to locally connect to in order to access the discovery, proxy, data broker, and firmware services. By hosting these services locally at the site, Alcon achieves faster response times to local events and can continue to operate in the event of a loss of connectivity. Snowball Edge is also a consolidation point for local data sets that will be asynchronously replicated to the cloud.
The document introduces artificial immune systems (AIS), which are computational systems inspired by the human immune system. It provides an overview of the immune system and its properties such as diversity, learning, memory, pattern recognition, and self/non-self discrimination. These properties provide a biological paradigm for developing AIS algorithms. The document then discusses representation schemes, affinity measures, and generic algorithms that have been developed for AIS, including negative selection, clonal selection, and immune network models. Finally, it reviews applications of AIS and discusses current trends in the field.
Yole Intel RealSense 3D camera module and STM IR laser 2015 teardown reverse ...Yole Developpement
INNOVATIVE 3D CAMERA FOR FACIAL ANALYSIS AND HAND/FINGER TRACKING, BASED ON RESONANT MICRO-MIRROR, IR LASER, VISIBLE AND NEAR INFRARED IMAGE SENSORS.
Intel RealSense is an intelligent 3D camera equipped with a system of three components: a conventional camera, a near infrared image sensor and an infrared laser projector. Infrared parts are used to calculate the distance between objects, but also to separate objects on different planes. They serve for facial recognition as well as gestures tracking.
The Intel 3D camera can scan the environment from 0.2m to 1.2m. The fixed-focal length camera will support up to 1080p @30FPS capture in RGB with a 77° FOV. Its lens has a built in IR cut filter. The 640x480 pixel VGA camera has a frame rate up to 60fps with a 90° FOV, moreover its lens has an IR Band Pass filter.
More information on that report at http://www.i-micronews.com/reports.html
YZ insan istihbarat süreçlerinin makineler, özellikle de bilgisayar sistemleri ile simülasyonudur. Bu süreçler arasında öğrenme (bilgi edinme için bilgi ve kuralların edinilmesi), mantıksallaştırma (yaklaşık veya nihYZ sonuçlara ulaşmak için kuralları kullanarak) ve kendini düzeltme yer alır. YZ'nın özel uygulamaları, uzman sistem konuşma tanıma ve suni görme içerir.
How do we protect privacy of users when building large-scale AI based systems? How do we develop machine learned models and systems taking fairness, accountability, and transparency into account? With the ongoing explosive growth of AI/ML models and systems, these are some of the ethical, legal, and technical challenges encountered by researchers and practitioners alike. In this talk, we will first motivate the need for adopting a "fairness and privacy by design" approach when developing AI/ML models and systems for different consumer and enterprise applications. We will then focus on the application of fairness-aware machine learning and privacy-preserving data mining techniques in practice, by presenting case studies spanning different LinkedIn applications (such as fairness-aware talent search ranking, privacy-preserving analytics, and LinkedIn Salary privacy & security design), and conclude with the key takeaways and open challenges.
PARAS HMIS: A Patient Centric Comprehensive & Integrated Healthcare Delivery ...Ananta Sahai
The document provides an overview of PARAS HMIS, a patient-centric healthcare delivery platform developed by Srishti Software Applications Pvt. Ltd. It conducted a global survey of over 130 hospitals to understand key concerns and areas for improvement. PARAS aims to address these concerns by providing an integrated platform covering the entire spectrum of patient care, from outpatient to inpatient and more. It utilizes an EHR-centric approach and offers features like clinical decision support, specialty templates, and bidirectional integration between clinical and administrative processes. Case studies demonstrate successful implementations at large hospital groups in India and Africa delivering benefits like optimized resource utilization and reduced revenue leakages.
The document summarizes several bio-inspired algorithms including CLONALG, aiNet, ABNET, and Opt-aiNet. CLONALG is a clonal selection algorithm inspired by immune system principles of clonal selection, hypermutation, and affinity maturation. aiNet is an artificial immune network model that uses principles of clonal selection, affinity maturation, and network suppression to perform unsupervised learning and clustering. ABNET is an antibody network based on a feedforward neural network trained with immune system concepts. Opt-aiNet adapts the aiNet model for optimization problems by introducing dynamic population sizing, mutation proportional to fitness, and automatic stopping criteria.
This document proposes an active deep learning approach to CAPTCHA recognition using a small initial training set. A convolutional neural network is trained on CAPTCHAs containing 6 digits. During testing, samples that are classified correctly but with high uncertainty are added back to the training set to improve the model in subsequent rounds of learning, without needing human labels. The method is evaluated on CAPTCHAs generated using different configurations, and performance is shown to improve significantly with each round of active learning by selecting additional uncertain samples for retraining.
Yole Intel RealSense 3D camera module and STM IR laser 2015 teardown reverse ...Yole Developpement
INNOVATIVE 3D CAMERA FOR FACIAL ANALYSIS AND HAND/FINGER TRACKING, BASED ON RESONANT MICRO-MIRROR, IR LASER, VISIBLE AND NEAR INFRARED IMAGE SENSORS.
Intel RealSense is an intelligent 3D camera equipped with a system of three components: a conventional camera, a near infrared image sensor and an infrared laser projector. Infrared parts are used to calculate the distance between objects, but also to separate objects on different planes. They serve for facial recognition as well as gestures tracking.
The Intel 3D camera can scan the environment from 0.2m to 1.2m. The fixed-focal length camera will support up to 1080p @30FPS capture in RGB with a 77° FOV. Its lens has a built in IR cut filter. The 640x480 pixel VGA camera has a frame rate up to 60fps with a 90° FOV, moreover its lens has an IR Band Pass filter.
More information on that report at http://www.i-micronews.com/reports.html
YZ insan istihbarat süreçlerinin makineler, özellikle de bilgisayar sistemleri ile simülasyonudur. Bu süreçler arasında öğrenme (bilgi edinme için bilgi ve kuralların edinilmesi), mantıksallaştırma (yaklaşık veya nihYZ sonuçlara ulaşmak için kuralları kullanarak) ve kendini düzeltme yer alır. YZ'nın özel uygulamaları, uzman sistem konuşma tanıma ve suni görme içerir.
How do we protect privacy of users when building large-scale AI based systems? How do we develop machine learned models and systems taking fairness, accountability, and transparency into account? With the ongoing explosive growth of AI/ML models and systems, these are some of the ethical, legal, and technical challenges encountered by researchers and practitioners alike. In this talk, we will first motivate the need for adopting a "fairness and privacy by design" approach when developing AI/ML models and systems for different consumer and enterprise applications. We will then focus on the application of fairness-aware machine learning and privacy-preserving data mining techniques in practice, by presenting case studies spanning different LinkedIn applications (such as fairness-aware talent search ranking, privacy-preserving analytics, and LinkedIn Salary privacy & security design), and conclude with the key takeaways and open challenges.
PARAS HMIS: A Patient Centric Comprehensive & Integrated Healthcare Delivery ...Ananta Sahai
The document provides an overview of PARAS HMIS, a patient-centric healthcare delivery platform developed by Srishti Software Applications Pvt. Ltd. It conducted a global survey of over 130 hospitals to understand key concerns and areas for improvement. PARAS aims to address these concerns by providing an integrated platform covering the entire spectrum of patient care, from outpatient to inpatient and more. It utilizes an EHR-centric approach and offers features like clinical decision support, specialty templates, and bidirectional integration between clinical and administrative processes. Case studies demonstrate successful implementations at large hospital groups in India and Africa delivering benefits like optimized resource utilization and reduced revenue leakages.
The document summarizes several bio-inspired algorithms including CLONALG, aiNet, ABNET, and Opt-aiNet. CLONALG is a clonal selection algorithm inspired by immune system principles of clonal selection, hypermutation, and affinity maturation. aiNet is an artificial immune network model that uses principles of clonal selection, affinity maturation, and network suppression to perform unsupervised learning and clustering. ABNET is an antibody network based on a feedforward neural network trained with immune system concepts. Opt-aiNet adapts the aiNet model for optimization problems by introducing dynamic population sizing, mutation proportional to fitness, and automatic stopping criteria.
This document proposes an active deep learning approach to CAPTCHA recognition using a small initial training set. A convolutional neural network is trained on CAPTCHAs containing 6 digits. During testing, samples that are classified correctly but with high uncertainty are added back to the training set to improve the model in subsequent rounds of learning, without needing human labels. The method is evaluated on CAPTCHAs generated using different configurations, and performance is shown to improve significantly with each round of active learning by selecting additional uncertain samples for retraining.