Google Scholar, Linkedin, Mastodon, Research Gate, Github, Semantic Scholar, Twitter
My name is Timothée Lesort. I am a senior Data Scientist at Aignostics Gmhb in Berlin. I train large-scale vision self-supervised models for histopathology to improve the diagnostics of cancer and rare diseases.
I am specialized in deep learning for vision and language. My main topic of interest is continual learning and representation learning for generalization and to make it work efficiently at scale.
I also worked on large-scale continual pretraining of LLMs. (large language models) during my postdoc at UdeM, Mila – Quebec Artificial Intelligence Institute under the supervision of Irina Rish.
I obtained my PhD in Computer Science from IP Paris - Institut Polytechnique de Paris (France) in the U2IS lab under the supervision of David Filliat. My Subject was “Continual Learning: Tackling Catastrophic Forgetting in Deep Neural Networks with Replay Processes”, I studied how replay can be used for continual learning and in particular how generative models can be used for replay in continual learning. I also studied replay for continual reinforcement learning and the theoretical limitation of regularization of dynamic architecture for continual learning. I have a master's degree in electronics and robotics from CPE Lyon.
Some Research Projects:
→ Pretraining of Vision transformer for Histopathology
→ Continual pre-training of large language models.
→ Characterization of data distribution drifts.
→ A better understanding of continual learning models.
→ The impact of large pre-trained models for continual learning.
→ Continuum a python library based on pytorch built to make easy continual learning experimentation by proposing various benchmarks. The idea is to build a large set of experiments settings to help/boost continual learning progress.
→ **continual_learning_papers:** A catalogue of continual learning papers is also available on Overleaf as bibfile for writing papers [here].
Simple and scalable strategies to continually pre-train large language models Adam Ibrahim, Benjamin Thérien, Kshitij Gupta, Mats L Richter, Quentin Anthony, Timothée Lesort, Eugene Belilovsky, Irina Rish