Sepp Hochreiter on Parallels Between Attention Mechanisms and Modern Hopfield Networks
Researchers argue that “attention mechanism is the update rule of a modern Hopfield network with continuous states.”
Category: Machine Learning & Data Science
New ML Algorithm Tunes Quantum Devices Faster Than Human Experts
The proposed algorithm is approximately 180 times faster than an automated random search of the parameter space.
Using Deep RL to Model Human Locomotion Control in Neuromechanical Simulations
Researchers from Stanford University, UC Berkeley and CMU review neuromechanical simulations and DRL, with a focus on modelling the control of human locomotion.
Fast and Furious! DRL-Fuelled Agents Grab Pole Position in Gran Turismo Sport
University of Zurich and ETH Zurich’s deep reinforcement learning agents have achieved superhuman performance in the time trials of popular racing simulator game Gran Turismo Sport.
Tesla Video Data Processing Supercomputer ‘Dojo’ – Building a 4D System Aiming at L5 Self-Driving
Elon Musk tweeted that Tesla is recruiting AI or chip talents for the company’s neural network training supercomputer project “Dojo.”
DeepMind & McGill Employ Operations Generalization to Divide & Conquer Complex RL Problems
DeepMind and McGill University researchers believe the problem could be solved through a strategic “divide-and-conquer” approach.
Yoshua Bengio’s Improved Rate-Based Curriculum Learning Outperforms Progress-Based Algorithms
Bengio and researchers from Université de Montréal and the École Normale Supérieure in Paris have introduced a new curriculum learning algorithm based on the notion of mastering rate.
Google Introduces NLP Model Understanding Tool
Google Research released a paper tackling this issue with a new open-source analytic platform: the Language Interpretability Tool (LIT).
Advancing Safety & Privacy for Trustworthy AI Inference Systems
Researchers from academia and industry offer input on the design, deployment, and operation of trustworthy AI inference systems.
Love-Love: Stanford Researchers Generate Realistic ‘Fake’ Wimbledon
A Stanford University research team has responded with an AI-powered model capable of realistically simulating a Wimbledon final and more.
How Smart is BERT? Evaluating the Language Model’s Commonsense Knowledge
Researchers dive deep into the large language model to discover how it encodes the structured commonsense knowledge it leverages on downstream commonsense tasks.
Tour the World From Your Couch: Google ‘NeRF-W’ Delivers Accurate 3D Scene Reconstruction of Complex Outdoor Environments
Google researchers have introduced a series of extensions to the SOTA view-synthesis method Neural Radiance Fields (NeRF) that enable it to produce high-quality 3D representations of complex scenes with only unstructured image collections as input.
Preferred Networks Releases Optuna v2.0
Open-source library for machine learning now gives the ability to evaluate which hyperparameters most affect performance.
Eyes on Me: Google AI ‘MediaPipe Iris’ Improves Iris Tracking and Distance Estimation
MediaPipe Iris, a novel machine learning model designed to deliver accurate iris estimation without using depth sensors.
Pixel2Style2Pixel: Novel Encoder Architecture Boosts Facial Image-To-Image Translation
Researchers from Penta-AI and Tel-Aviv University introduce a generic image-to-image translation framework dubbed Pixel2Style2Pixel (pSp).
SIAT Researchers Design Novel Deep Learning System to Assess Skeletal Maturity
Researchers introduced an ensemble-based deep learning pipeline to automatically assess the distal radius and ulna (DRU) maturity from left-hand radiographs.
DeepMind & Google Explore Hyperparameter Selection for Offline RL
Researchers conducted a thorough empirical study of hyperparameter selection for offline RL, aiming to identify and develop more reliable and effective approaches.
Unselfie Turns Arm-Extended Selfies into Neutral-Pose Portraits
“Unselfie,” a novel photographic transformation model that can automatically translate selfies into neutral-pose portraits.
AI Generator Learns to ‘Draw’ Like Cartoonist Lee Mal-Nyeon in Just 10 Hours
A Seoul National University Master’s student and developer has trained a face generating model to transfer normal face photographs into cartoon images in the distinctive style of Lee Mal-nyeon.
As Its GPT-3 Model Wows the World, OpenAI CEO Suggests ‘the Hype Is Way Too Much’
OpenAI’s 175 billion parameter language model GPT-3 has gone viral once again.
Google ‘BigBird’ Achieves SOTA Performance on Long-Context NLP Tasks
BigBird is shown to dramatically improve performance across long-context NLP tasks, producing SOTA results in question answering and summarization.
Applying Linearly Scalable Transformers to Model Longer Protein Sequences
Researchers proposed a new transformer architecture called “Performer” — based on what they call fast attention via orthogonal random features (FAVOR).
AI-Powered ‘Genderify’ Platform Shut Down After Bias-Based Backlash
Genderify is an AI-powered tool designed to identify a person’s gender by analyzing their name, username or email address.
Preferred Networks Releases PFRL Deep Reinforcement Learning Library for PyTorch Users
PFRL succeeds ChainerRL as comprehensive library with cutting-edge deep reinforcement learning algorithms and features.
MLPerf Training v0.7 Results Released: Google & NVIDIA Lead the Race
The industry-standard MLPerf benchmark today released the results of the third round of its ongoing ML Training Systems competition.
Google Proposes Lasso Algorithm Variant for Learning Convolutions: ‘Bridging the Gap Between Fully-Connected and Convolutional Nets’
Unlike previous work that suggested fully-connected networks can only deliver subpar performance, Neyshabur’s experiments showed a significant performance jump.
Asia China Computer Vision & Graphics Health Machine Learning & Data Science Research Shenzhen Institute of Advanced Technology
About Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences
SIAT aims to enhance the innovative capacity of the equipment manufacturing and service industries in the Guangdong-Hong Kong region, promote the development of emerging industries possessing their own proprietary intellectual property, and become a world-class industrial research institute.
SIAT Researchers Propose Novel Correction Method for Gas Sensor Drift Problem
Researchers developed a subspace learning-based domain adaptation method, which could potentially alleviate the gas sensor drift problem.
New White-Box Framework Will Cartoonize Your World
This framework can generate high-quality cartoonized images with much-improved controllability.
Adobe and Stanford Unveil SOTA Method for Human Pose Estimation
The researchers say the approach produces motions that are visually and physically much more plausible than state-of-the-art methods.
DeepMind’s AlignNet Learns Stable Object Representations Across Time
New research from UK based AI company and research lab DeepMind is enabling AI agents to perceive dynamic real-world environments more like humans do.
Microsoft & MIT Apply Adversarially Robust Models for Better Transfer Learning
Despite being less accurate on ImageNet, adversarially robust neural networks can match or improve their standard counterparts’ transfer performance.
Joint Spatial-Temporal Transformation Learning Boosts Video Inpainting Performance
Researchers introduce a joint Spatial-Temporal Transformer Network (STTN) to tackle such video inpainting challenges.
Facebook & Inria Propose High-Performance Self-Supervised Technique for CV Tasks
Researchers have developed a new technique for self-supervised training of convolutional networks used for image classification and other computer vision tasks.
Automating Machine Learning: Google AutoML-Zero Evolves ML Algorithms From Scratch
Google researchers propose an “AutoML-Zero” approach designed to automatically search for machine learning (ML) algorithms from scratch, requiring minimal human expertise or input.
Meet ByteDance AI’s Xiaomingbot: World’s First Multilingual and Multimodal AI News Agent
Researchers from ByteDance AILab and Shanghai Jiao Tong University have introduced Xiaomingbot, a multilingual and multimodal news reporter.
Fujitsu Develops World’s First AI technology to Accurately Capture Characteristics of High-Dimensional Data Without Labeled Training Data
Fujitsu has developed the world’s first AI technology that accurately captures the characteristics of high-dimensional data without labeled training data.
SIAT Researchers Propose Novel Deep Learning Method for Positron Emission Tomography
Researchers developed a novel deep learning method that reconstructs PET images using a cascading back-projection neural network (bpNet).
Google Bridges Domain Gaps in Semantic Segmentation of LiDAR Point Clouds
A team of Google researchers recently proposed a novel “complete and label” domain adaptation approach.
Grounded Language Learning: A Look at the Paper ‘Understanding Early Word Learning in Situated Artificial Agents’
This paper carefully engineers artificial-language learning experiments to replicate sources of information infants learn from and under what conditions.