Working with combinations of heterogeneous or unstructured data is a key challenge for research on big data processing. Multiple datasets can be acquired through different modalities, which calls for complex data fusion models.
The CRDM is particularly active in research projects on major unstructured and semi-structured data issues, such as :
- Developing a multimodal model, i.e., a data fusion model that could for example use an image dataset together with a text dataset made up of text descriptions of those same images.
- Approaches based on machine learning, including learning by deep networks—currently the state of the art in massive big data learning technology.