Research
Discover an index of datasets, SDKs, APIs and open-source tools developed by Microsoft researchers and shared with the global academic community below. These experimental technologies—available through Azure AI Foundry Labs (opens in new tab)—offer a glimpse into the future of AI innovation.
This data set contains 244,497 images from 140 camera locations in the Southwestern United States, with species-level labels for 22 species, and approximately 66,000 bounding box annotations.
TensorWatch is a comprehensive library of tools to debug and monitor training phase for Deep Learning and Reinforcement Learning models as well as perform analysis on trained models. TensorWatch is a debugging and visualization tool…
This project can be used to reproduce the DQN implementation presented in the ICML2019 paper: Safe Policy Improvement with Baseline Bootstrapping, by Romain Laroche, Paul Trichelair, and Rémi Tachet des Combes. For the finite MDPs…
This project can be used to reproduce the finite MDPs experiments presented in the ICML2019 paper: Safe Policy Improvement with Baseline Bootstrapping, by Romain Laroche, Paul Trichelair, and Rémi Tachet des Combes. For the DQN…
The presentation starts with a brief introduction of Reinforcement Learning (RL) and an overview of its success. Even though these achievements are compelling, state-of-the-art algorithms require an unreasonable amount of data. Moreover, they sometimes converge…
Scripts to generate the CoDraw and i-CLEVR datasets used for the GeNeVA Neural Visual Artist (GeNeVA) task proposed in Tell, Draw, and Repeat: Generating and modifying images based on continual linguistic instruction.
Bing Artificial Search Sessions(BASS) is a collection of 18m Artificial Search session that were created by taking real conversational Search Sessions and mapping them to publicly available queries using vector space embeddings.
MS MARCO is a collection of datasets focused on deep learning in search. The first dataset was a question answering dataset featuring 100,000 real Bing questions and a human generated answer. Since then, we released…
Access Publication Publication Publication Publication Publication
A python package with a reinforcement learning algorithm that decodes latent states from rich observations.
The Bosque programming language is an experiment in regularized design for a machine-assisted rapid and reliable software development.