Source  
Flashlight - Machine Learning library in C++ On April 17, Saturday, 2021

FAIR is open-sourcing a machine learning (ML) library written in C ++. It will make it easier for the AI community to tinker with the low-level code underpinning deep and ML frameworks. FAIR's Flashlight was built to power groundbreaking research. It's built on top of a shallow stack of basic abstractions that are modular and easy to use. It has incredibly low framework overhead, as modern C ++ enables first-class parallelism and out-of-the-box speed.

Read more at ai.facebook.com

Source  
Machine Learning method predicts battery health On April 12, Monday, 2021

Batteries are increasingly crucial in a variety of applications including electric vehicles, trains and ships. A key challenge is that batteries degrade quickly with operating conditions. It is difficult to estimate battery health without interrupting the operation of the battery or going through a lengthy procedure of charge-discharge. The developed framework scales up with new chemistries, including the new upcoming solid-state batteries, battery designs and operating conditions and has the potential to be used in new applications.

Read more at techxplore.com

Source  
Google Lyra to enable voice calls for another billion users On April 11, Sunday, 2021

Google released Lyra in February of 2021, an innovative audio codec that uses machine learning to enable high-quality voice calls. Google has made the product open source, inviting other developers to operate their own communication apps. Developers can use Lyra to encode and decode audio, as the tool runs on the 64-bit ARM Android platform, including development on Linux. The open source nature of Lyra seeks to advance accessibility for users, Google says. The tool can archive large amounts of speech, alleviating network congestion in emergency situations and saving battery power.

Read more at techxplore.com

Source  
Develop 3D structures from 2D material images using ML On April 06, Tuesday, 2021

A new machine learning algorithm can convert 2D images of composite materials into 3D structures. The tool learns what 2D cross-sections of composites look like and scales them up so their phases can be studied in a 3D space. Studying, designing, and manufacturing composite materials in three dimensions is currently challenging. The algorithm could be used to improve the design and production of materials like battery electrodes and aircraft parts in 3D. It is cheaper and faster than creating 3D computer representations from physical 3D objects.

Read more at techxplore.com

Source  
Build a CI/CD pipeline for Machine Learning Model in AWS On April 04, Sunday, 2021

SageMaker is a fully managed service that provides developers with the ability to build, train, and deploy machine learning (ML) models quickly. SageMaker removes the heavy lifting from each step of the ML process to make it easier to develop high-quality ML artifacts. AWS Serverless Application Model (AWS SAM) is an open-source framework for building serverless applications. For continuous integration and continuous delivery (CI/CD) pipelines, Amazon SageMaker Pipelines is the first purpose-built, easy-to-use CI/CD service for ML.

Read more at aws.amazon.com

Source  
Real-time inspection of final product using AI On March 30, Tuesday, 2021

Researchers at the Fraunhofer Institute for Computer Graphics Research IGD are set to showcase MARQUIS at the Hannover Messe. MARQUIS combines augmented reality with machine learning methods to allow comparisons to be made between a CAD specification and the real product. Workers simply pull out their smartphone or tablet and point it at the component concerned. The system recognizes the component and also identifies any discrepancies from the target dimensions, for example. The technology will be on display at Hannovers Messe from April 12-16, 2021.

Read more at techxplore.com

Source  
Replacing rewards with examples in Reinforcement Learning On March 25, Thursday, 2021

A general goal of robotics research is to design systems that can assist in a variety of tasks. Most reinforcement learning algorithms for teaching agents to perform new tasks require a reward function. We propose a machine learning algorithm that teaches agents how to solve new tasks by providing examples of success. This algorithm does not rely on hand-crafted reward functions, distance functions, or features. It learns to solve tasks directly from data, requiring the agent to solve the entire task by itself.

Read more at ai.googleblog.com

Source  
Binding Machine Learning with Game Development On March 20, Saturday, 2021

Online games have exploded in popularity, making it harder for game designers to balance games. We present an approach that leverages machine learning to adjust game balance by training models to serve as play-testers. By running millions of simulations using trained agents to collect data, this ML-based game testing approach can be used to more efficiently make a game more fun, balanced, and aligned with the designer's vision.

Read more at ai.googleblog.com

Source  
Music Circles - Interactive Visualisation tool to help create new music On March 19, Friday, 2021

Researchers at Seoul National University have created an interactive data visualisation tool called Music Circles. The tool represents songs as unique vectors and calculates similarities between different songs. It then generates visualisations of song clusters aligned with a user's musical preferences. The goal of the study was to help users search for music they might like and explore music streaming catalogues in a more intuitive and engaging way.

Read more at techxplore.com

Source  
Barlow Twins - Method for Self-Supervised Learning via Redundancy Reduction On March 18, Thursday, 2021

Barlow Twins is a new self-supervised learning approach for computer vision. Facebook AI and New York University research team introduce Barlow Twins. Method is based on redundancy-reduction, which can explain organization of visual systems. Researchers say method outperforms current state-of-the-art methods for self- supervised learning while being conceptually simpler and avoiding trivial constant representations.

Read more at syncedreview.com

Source  
Amazon Textract - PDF document pre-processing On March 15, Monday, 2021

Amazon Textract can detect text in a variety of documents, including financial reports, medical records, and tax forms. In many cases, there is a need to extract and analyze documents with various visuals, such as logos, photos, and charts. These visuals contain embedded text that isn't needed for your downstream process.

Read more at aws.amazon.com

Source  
Facebook announces Learning from Videos On March 15, Monday, 2021

Facebook is launching a project called Learning from Videos, designed to automatically learn audio, and visual representations from the data in publicly available videos uploaded to Facebook. This is part of broader efforts toward building machines that learn like humans do. The first application is now live in Instagram Reels' recommendation system. Building AI that learns from public available videos will help create machines that better analyze real-world sights and sounds, Mark Zuckerberg says.

Read more at ai.facebook.com

Source  
Long-Term Impact of COVID-19 pandemic using AI On March 11, Thursday, 2021

Fraunhofer IAO and seracom GmbH collaborate together to investigate long-term impact of COVID-19 using AI algorithms. This project was named as WIBCE. This study aimed to identify specific groups of people which are especially burdened and vulnerable regarding their mental health during the pandemic. The research investigation analyzes the responses of 275 respondents across Germany through an online survey and used machine learning algorithms to unravel hidden inferences in the data.

Read more at techxplore.com

Source  
LazyTensor - Facebook & Google's research enables expressive Domain-Specific Compilers On March 06, Saturday, 2021

Researchers from Facebook and Google developed a unique technique that combines eager execution and domain-specific compilers (DSCs) to enable full use of host programming languages throughout the tensor portions of user programs. Both the conglomerates have the research published in the paper, LazyTensor : combining eager execution with domain specific compilers.

Read more at syncedreview.com

Source  
Deno 1.8 offers GPU accelerated Machine Learning On March 04, Thursday, 2021

Deno 1.8, released on March 2, offers support for an API to bring optimized machine learning to the secure JavaScript/TypeScript runtime. WebGPU API renders computation on GPU by providing path towards accelerated machine learning in Deno. It gives developers a low-level, high-performance cross-architecture mechanism to program GPU hardware from JavaScript.

Read more at www.infoworld.com

Source  
Fast Video Object Segmentation On March 03, Wednesday, 2021

Yu Li, Zhuoran Shen and Ying Shan developed a real-time high-quality semi-supervised video object segmentation algorithm. The novel feature of the algorithm is a global context module that challenges the pre-existing model which only use one frame or a few frames to guide the segmentation of the current frame. Global context module uses all past frames and achieves the top performance at a real-time speed.

Read more at syncedreview.com

Source  
Towards Causal Representation Learning On March 03, Wednesday, 2021

Turing award honoree Yoshua Bengio and his research team makes a progressive effort to unite causality (cause-effect relationship) with ML research. Bengio in his paper argues it is time to look at causal learning for possible solutions and proposes critical areas for future research. The idea behind research is to reduce the significant gap between state-of-art machine learning-driven AI and the intelligence that is seen in humans and animals.

Read more at syncedreview.com

Source  
Machine Learning in our pockets On February 27, Saturday, 2021

EPFL- INRIA research has shown that it is possible for complex machine learning algorithms to run on mobile devices in real time as a part of distributed network. Big Tech companies like facebook and google collects behavioral data of user every time they visit internet in order to provide better web experience. This research exposes the prospect of real-time machine learning without sharing the data thereby protecting the privacy of user.

Read more at techxplore.com

Source  
A Tensorflow Framework for AutoML On February 26, Friday, 2021

Model Search is an open-source, TensorFlow based python framework for building AutoML algorithms at a large scale. The idea of Model Search was presented at Google Interspeech 2019 by Hanna Mazzawi, Javier Gonzalvo, Aleks Kracun, Prashant Sridhar, Niranjan Subrahmanya, Ignacio Lopez Moreno, Hyun Jin Park, Patrick Violette. The key idea is to develop a novel Neural Architecture search that aims at defining an incremental search.

Read more at analyticsindiamag.com

Source  
Reinforcement Learning Algorithm beats all in classic video games On February 26, Friday, 2021

Researchers at Uber AI Labs at San Francisco have developed a set of learning algorithms that manifested better results in playing classic video games than humans and many AI systems. Publishing the paper in the journal Nature, they explains how their algorithm is better than the pre-existing and the vast application of it in the areas of robotics, language processing and even creating new drugs.

Read more at techxplore.com

Source  
COVID mask detection on an Arm Cortex On February 24, Wednesday, 2021

In this post, it is shown how to approach and solved the challenges. Convert a PyTorch ResNet to TensorFlow and quantize it to use 8-bit integer values. Collect, select, and annotate data of faulty and non-faulty masks. Use self-supervised pre-training to boost model performance on fewer images.

Read more at dev.to

Source  
Sony Patents on Behavioral AI On February 24, Wednesday, 2021

Japanese tech giant gets patent on behavioral machine learning system that automatically adjusts the difficulty level when gamers are struggling. ML model is believe to understand the behavior of the opponent thereby making the game more or less difficult. Patent also suggests a reward distribution system where distinct prizes are awarded based on the difficulty level.

Read more at thenextweb.com

Source  
How AI And Machine Learning Are Improving Marketing In 2021? On February 22, Monday, 2021

CMOs must find the perfect balance between the emotional and rational, data-driven portions of their work. 36% of marketers expect that this year, AI will have a major effect on marketing success. CMOs who lead high-performance marketing teams put a high emphasis on learning constantly and embracing an attitude of growth.

Read more at www.forbes.com

Source  
Machine Learning in 4 steps On February 21, Sunday, 2021

Machine learning projects can be difficult to use in production. There is no clear way to test models on your own examples to make sure they work. This article shows how to turn a project from only-code to a production-ready ML project. It's a step-by-step guide to turning a project into a production readyML project.

Read more at dev.to

Source  
Apply Machine Learning to Business Problems On February 21, Sunday, 2021

Emerj market research focuses on cutting through the AI hype. Tech executives often have trouble identifying where their business might apply machine learning (ML) to business problems. Is the prediction you are trying to make? Do you have new data and clean data? Does your data have existing labels to help a machine make sense of it? Can your solution to this problem afford for some allowance of error? Emerj's research helps innovation and strategy leaders make a better business case for AI.

Read more at www.predictiveanalyticsworld.com

Source  
Machine Learning Identifies Existing Drugs to Treat COVID-19 On February 17, Wednesday, 2021

Machine learning could help identify existing drugs that could be repurposed to fight the coronavirus. MIT researchers used big data analytics to identify drugs that might respond to COVID-19 infection. Clinical testing is necessary to determine the efficacy of these drugs, they say. The study was published in the journal Nature Communications.

Read more at healthitanalytics.com

Source  
Comau Validates Machine Learning Platform On February 15, Monday, 2021

Comau is an Italian industrial automation company. It is a subsidiary of automaker Stellantis. This Machine Learning project names "MUSKETEER" enables data analytics on organized datasets. It has two test datasets from two manufacturing units. The increase in the dataset calls for new ways to preserve privacy while still allowing data sharing among multiple data owners. This project aims to create a validated, federated, privacy-preserving machine learning platform tested on industrial data that is interoperable, scalable, and efficient enough to be deployed in real use cases. Data can continue to be stored in different locations with different privacy constraints but shared securely. The MUSKETEER project has received funding from the European Union's Horizon 2020 research and innovation program under grant agreement No 824988.

Read more at metrology.news

TAGS