Released tools
-
Torch contributions
- 2017-RL PyTorch: A RL simple package for PyTorch and OpenAI Gym (approx 150 stars on GitHub).
- 2016-2017-Participation to the development of the PyTorch Platform (Deep Learning) developed by Facebook AI Research.
- 2016-RLTorch: A RL simple package for Torch and OpenAI Gym (approx 50 stars on GitHub).
-
Machine learning
-
Computer vision
- 2021
- MixMo: Mixing Multiple Inputs for Multiple Outputs via Deep SubNetworks (ICCV 2021)
- Continuum: Continual learning / data loaders (CVPR 2021 workshop)
- PLOP: Continual learning for segmentation (CVPR 2021)
- 2020
- ESL: Entropy-guided Self-supervised Learning for Domain Adaptation in Semantic Segmentation (CVPR 2020 Workshop on Scalability in Autonomous Driving)
- Stochastic Latent Residual Video Prediction (ICML 2020)
- Incremental learning for image classification (+ zero-shot) (ECCV 2020)
- 2019
- Unsupervised Object Segmentation by Redrawing (neurIPS 2019)
- Unsupervised Scalable Representation Learning for Multivariate Time Series (LLD@ICLR2019)
- 2017 – FaderNetworks (NIPS 2017) — work in collaboration with Facebook Research
- 2017 – MUTAN: Multimodal Tucker Fusion for Visual Question Answering (ICCV 2017)
- 2017 – Pretrained computer vision models for Pytorch
- 2021
-
Natural language processing – information retrieval
- 2021:
- OpenNIR-Lifelong – lifelong learning framework for information retrieval (ECIR 2021)
- QuestEval – NLG evaluation metric
- 2020:
- PARENTing via Model-Agnostic Reinforcement Learning to Correct Pathological Behaviors in Data-to-Text Generation (INLG 2020)
- Hierarchical model for data-to-text (ECIR 2020)
- Information extraction setup – “Let’s Stop Incorrect Comparisons in End-to-End Relation Extraction!” EMNLP 2020
- Contextualized Embeddings in Named-Entity Recognition (ECIR 2020)
- 2019 – Answers Unite! Unsupervised Metrics for Reinforced Summarization Models (EMNLP 2019)
- 2017 – Semantic embedding evaluation platform (word similarity, concretness, feature norm prediction)
- 2015-BIOASQ challenge platform. BIOASQ was a EU project (2012-2014) – MLIA was in charge of developing the evaluation platform for the project. The platform is still active and challenges are organized every year from 2013 to 2017.
- 2021:
-
Spatio-temporal prediction
- 2021: Spatiotemporal prediction and disentanglement (ICLR 2021)
-
Scattering for audio, image, 3D
- 2019 – On Lazy Training in Differentiable Programming (Neurips 2019)
- 2018 – KYMAT (for torch, tf, keras, np) – connected to https://jmlr.org/papers/volume21/19-047/19-047.pdf
-
Social network
- 2014-Learning Social Network Embeddings for Predicting Information Diffusion – (WSDM 2014) (approx 25 stars on Github).
-
Experiment manager and datasets
- 2020 – Experiments for IR (SIGIR 2020):
- 2017 – Experimaestro: Experiment manager based on a server that contains a job scheduler
- 2017 – Dataset manager