Software

The project provides a Python package and command-line tool for generating feature vectors from molecules. The feature sets to include in the feature vectors are configurable via a simple YAML configuration file. Molecules are specified as lists of Inchis.

A Python framework for machine- and deep-learning that makes it easy to use Hydra for hyperparameter configuration and MLflow for experiment tracking and result distribution. The user only needs to create a single YAML configuration file and a subclass of Hydronaut.Experiment to use the framework.

Hydra allows the user to systematically sweep all hyperparameter combinations or optimize them use different strategies with plugins for libraries such as Optuna.

MLflow provides a web interface, command-line interface and Python API for exploring and sharing the results.

The framework is fully compatible with PyTorch Lightning and provides a custom subclass to facilitate its use.

A Python library and command-line tool to automatically export presentations from Jupyter notebooks along with custom CSS and other user-configurable website content.

Metabograph is a Python library for generating metabolic networks from BioPAX data. The generated NetworkX graphs can be used in various application such as training graph neural networks (GNNs) or saved to files for use with viewers such as Gephi.

Utility scripts and matching python module for working with MLflow directories. It mainly provides functionality for moving around and regrouping "mlruns" directories which is useful in contexts such as migrating results from a cluster.

A Hydronaut-based framework that uses ChemFeat to generate feature vectors for training machine- and deep-learning models to predict properties of molecules.

Python library containing classes for representing sequences and chronicles, ie a representation of a temporal pattern. It implements efficient recognition algorithms to match chronicles in a long sequence.

Parse PicoQuant PTU files in Python. It includes a Python package and a command-line utility.

The SNDS, formerly SNIIRAM, is a huge database (consisting of several TBs of data and about 700 tables) that contains information about healthcare reimbursements of about 60 million French insured patients. This database is used to carry out epidemiological and medical-economic studies. Due to its sensitive medical content, identifying information (names, social security numbers) is removed or replaced by spurious information. This repository contains a solution to generate a synthetic version of the database.

The software generates relational data compliant with the original database schema with realistic distributions. This guarantees privacy preservation thanks to the use of open data only.

This work is currently under review to the AIME conference.

SWoTTeD is a tensor decomposition framework to extract temporal phenotypes from structured data. Most recent decomposition models allow extracting phenotypes that only describe snapshots of typical profiles, also called daily phenotypes. However, SWoTTeD extends the notion of daily phenotype into temporal phenotype describing an arrangement of features over a time window.

VCNet is a counterfactual generator for tabular datasets. Counterfactual generation is an explainable artificial intelligence technique that helps users understand the decisions of predictive models.
VCNet handles both categorical and numerical variables and can also manage immutable variables (i.e., variables that cannot be changed while generating counterfactuals). Currently, it is implemented for binary classification only. VCNet is based on conditional variational encoders and offers an easy and fast training architecture that generates realistic counterfactuals (i.e., counterfactuals that likely exist in the data distribution).
This library provides two versions of VCNet:
* Joint-learning architecture: This version includes both the classification model and the counterfactual generator.
* Post-hoc architecture: This version includes only the counterfactual generator, which can be fitted on any prediction model (Torch module).

Once the VCNet model is fitted, it can be used to generate efficiently valid counterfactuals of examples.