The MAIA Lab offers a series of courses covering theoretical aspects as well as best practices in artificial intelligence, machine learning, deep learning and software engineering. These are offered to both engineering students and computer science students.
The courses offered in artificial intelligence and machine learning cover:
- advances in artificial intelligence through deep learning applied to long-term dependency modeling, and in particular to natural language processing. The exploitation of DNNs, CNNs, RNNs, LSTMs, GRUs, Attention, Transformers, GPTs, generative models, GANs, and auto-encoders allow applications such as machine translation, information classification and extraction, chatbots, the extraction and search for information in "big data", etc.
- the field at the intersection of artificial intelligence and probability theory, including probabilistic graphical models, Bayesian networks, and their implementation via probabilistic programming. Decision-making in the face of uncertainty (missing information, noisy data, etc.) via statistical inference is one of the major contributions of probabilistic AI.
The scientific research skills within the lab cover these different aspects, as well as multimodal AI, human-agent interaction, and effective AI implementations: computational efficiency, but also more efficient learning, and more interpretable and trustable AI.
MAIA Lab took part in CVPR 2022 - NeuroVision workshop. Ahmad Hammoudeh and Prof. Stéphabe Dupont presented their paper titled "How does explicit orientation encoding affect image classification of ConvNets?", which was accepted for a lightning talk and a poster presentation. Continue reading
A paper and and an extended abstract were accepted in LREC 2022 - SMILA workshop. Congrats to MAIA Lab researchers: Ahmad Hammoudeh, and Stéphabe Dupont. Continue reading
MAIA Lab took part in NeurIPS 2021 debriefing organized by University of Amsterdam on 15 March 2022, wherein PhD students and/or senior researchers briefly presented the papers they found the most interesting at NeurIPS 2021. Ahmad Hammoudeh talked about a paper titled "On the Frequency Bias of Generative Models" authored by Katja Schwarz, Yiyi Liao, and Andreas Geiger. Continue reading