Multi-Learn is a Multiple Learning focused organization that gathers multiple projects developed as a result of close international collaboration. Originally focused on multi-view learning, the Multi-Learn projects now embraces a broader scope, encompassing diverse learning paradigms including large-scale single-image learning, segmentation, and multi-modal approaches.
Multi-Learn is a research and development initiative that explores a wide spectrum of learning paradigms in machine learning; with a historical focus on multi-view learning. Today, Multi-Learn encompasses a broader set of challenges including multimodal integration, synthetic data generation, single-image learning, and large-scale segmentation. It serves as a hub for interoperable tools and benchmarks designed to support reproducible research and scalable machine learning workflows. Key software and datasets developed under Multi-Learn it now addresses broader challenges such as:
- Multimodal integration
- Synthetic data generation
- Single-image learning
- Large-scale segmentation
It serves as a hub for interoperable tools and reproducible benchmarks, enabling scalable ML workflows.
- ULaval's GRAAL
- AMU's LIS machine learning team Qarma
- ARCHIMEDE's Development team ValoCell
- CHUQ's omics-oriented team Corbeil Lab
- LAM's Laboratoire d'Astrophysique de Marseille - Learning Projects BigSF
- TAGC's Theories and Approaches of Genomic Complexity gihub TAGC
Summit (Supervised MultiModal Integration Tool): A framework to benchmark mono- and multi-view classifiers on custom datasets. repo summit
scikit-multimodallearn: A Python package (scikit-learn compatible) for multimodal data classification. repo multimodallearn
MAGE (Multi-view Artificial Generation Engine): A toolkit to generate synthetic multi-view datasets for controlled experimentation. repo MAGE
Multimodal Protein Dataset: A dataset and toolkit for classifying highly multi-functional proteins using multimodal biological data. repo Protein Dataset
LIS² (Large Image Split Segmentation): A toolbox for large-scale, single-image semantic segmentation. repo LIS²
LIS²-SKELETON (Large Image Split Segmentation skeleton): external package for skeletization task of Galactic plane segmentation. repo LIS²-SKELETON
Multi-Learn brings together these tools to support the design, evaluation, and comparison of learning strategies across real and simulated datasets, and across a variety of modalities and learning tasks.
