Post doc in self-supervised deep learning (3 years M/F)

The successful candidate will join our lab and engage a research work to design and publish self-supervised deep learning method suited to enhance capabilities of current microscopes. He/she will interact with the other lab members to use and possibly augment his/her expertise in deep and machine learning to the benefit of his/her project under the form of novel compelling idea, proposition of numerical experiment based on data, algorithms or network architectures. He/she will also be in charge to write and submit manuscripts to international machine learning conferences and journal with review committee, describing a novel method validated quantitatively.

For about ten years, deep learning has enabled the community to obtain impressive results in supervised learning tasks for pattern recognition from various types of large datasets and especially those made of images. Biology, medicine and healthcare has recently been recognized as one of research field that produces the largest amount of data. Thus, it is an ideal environment for computer and data scientists that aim to develop sophisticated, cutting edge algorithms related to current scientific questions.

Robotics coupled with high throughput microscopy can generate hundreds of thousands images from automated experiments. Using such datasets, we and others showed that deep network models, properly conceived and trained, have enabled to discriminate subtle, and even sometimes invisible, cellular phenotypes from image. A performance that we couldn't hope to achieve a few years ago with classical image analysis algorithms. Approaches were also developed to classify, annotate or generate biological image data and analyze complex interaction between biological objects from large set of experiments. Applications of this work range from basic research in biology to drug discovery and diagnostic with identified collaborators in these fields.

In this project we aim to leverage the large image datasets we produced with colleagues to develop cutting edge methods able to significantly increase the resolution of microscopy images using computational mean and especially self-supervised learning approaches.

The Institut de biologie de l'ENS (IBENS) is a fundamental research center that conducts original research aimed at deciphering the fundamental mechanisms at the heart of biological processes.
A joint ENS-CNRS-INSERM unit, IBENS hosts more than 300 people grouped into 30 autonomous teams conducting highly collaborative and multidisciplinary research that combines experimental and theoretical approaches.
The research activity covers various thematic fields: Neurosciences, Developmental Biology, Functional Genomics, Ecology and Evolutionary Biology.

The host lab (Computational Bioimaging and Bioinformatics: https://www.ibens.ens.fr/spip.php?rubrique47), led by Auguste Genovesio comprises about 10 computer scientists. The lab is part of the Center for Computational Biology, an interdisciplinary and international research center located at IBENS, Ecole Normale Supérieure in Paris, France. The lab performs and publish research in computer science to designs methods typically using large set of biological images data. It currently focusses on the development of deep learning methods for data analytics study of the cell morphology and dynamic at large scale. It also works in functional genomics and the interplay between gene expression and cell image phenotype. The École Normale Supérieure is a renown public higher education and research institution, located in the Latin Quarter in the center of Paris, close to numerous public transportation options (RER, subway, bus) and in a very nice and student area.