Microsoft

Researchers from McGill University and Microsoft Introduces Convolutional vision Transformer (CvT) that improves Vision Transformer (ViT) in Performance and Efficiency by Introducing Convolutions into...

Transformers have been widely used in the natural language processing (NLP) domain for years, and their introduction was a turning point for many NLP...

Microsoft Research Introduces a General-Purpose Multimodal Foundation Model ‘BEIT-3,’ that Achieves State-of-the-Art Transfer Performance on Both Vision and Vision Language Tasks

The machine learning community has recently diverted its focus on the convergence of language, vision, and multimodal pretraining. The main intention behind this is...

Researchers at Microsoft Introduce Z-Code++, A Pre-Trained Language Model Optimized For Abstractive Summarization

In this research article, Microsoft researchers introduce Z-Code++, an advanced encoder-decoder PLM (Pre-trained Language Model) created for machine translation that considerably enhances Z-Code and...

To Enable Advanced Research on Artificial Humanoid Control, Microsoft’s Robotics Team is Releasing A Library of Pre-Trained Simulated Humanoid Control Models with Enriched Data...

Simulated humanoids present an intriguing platform for investigating motor intelligence with their ability to mimic the whole spectrum of human motion. An important area...

Researchers from Microsoft Asia and Peking University Proposed NUWA-Infinity, a Model to Generate High-Resolution, Arbitrarily-Sized Images and Videos

In recent years, the generation of images or videos from different types of inputs (text, visual, or multimodal) has gained increased popularity. In this...

Latest Computer Vision Research At Microsoft Explains How This Proposed Method Adapts The Pretrained Language Image Models To Video Recognition

Numerous vision applications heavily rely on video recognition, including autonomous driving, sports video analysis, and microvideo recommendation. A temporal video model is showcased in...

Researchers Develop ‘TiCoder’ Framework For Code Generation Using User Feedback With 90.4% Consistency To User Intent

One of the key drivers of the recent success of powerful pretrained large language models (LLMs) in natural language processing is the model's capacity...

Researchers at Microsoft Research and TUM Have Made Robots to Change Trajectory by Voice Command Using A Deep Machine Learning Model

While deploying a robot in a real environment, many obstacles can often come. Like a robot arm is deployed to pick up an object,...

University of Arizona and Microsoft AI Research Presents ‘TextWorldExpress’: A High-Performance Text-Game Simulator That Boosts Simulation Upto One Million Steps Per Second

Even though they only have a virtual, not a physical, embodiment, agents that are graphically portrayed with a body, such as a human or...

Meet Microsoft’s ‘Bonsai Brain’: A Low-Code AI Platform That Speeds AI-Powered Automation Development

Microsoft's recent ongoing project called Bonsai Brain is dedicated to modeling and creating a low-code-based AI component that can be applied to various autonomous...

Researchers at Microsoft Propose a Low-Precision Training Algorithm for GBDT, Based on Gradient Quantization

Gradient boosting decision trees is a sophisticated machine learning technique frequently utilized in real-world applications such as electronic advertising, search ranking, time-series prediction, and...

Researchers From China Propose A New Machine Learning Framework Called BootMAE (Bootstrapped Masked Autoencoders) For Vision BERT Pretraining

In the computer vision field, self-supervised representation learning has been a challenging problem for a long time. Self-supervised representation learning aims to learn transferrable...

Recent articles

Be the first to know the latest AI research breakthroughs.

X