📈 How can we use medical data or physiological recordings in user-centered machine learning methods?
🤖 Can such data inform us to enhance or design novel machine learning algorithms?
🧠 What are the similarities between human cognition and machine learning models?
I use electroencephalography (EEG) recordings to implicitly control generative models' (generative adversarial neural networks or GANs) output visual features to match the user's intentions (e.g. I think of a blond person, the GAN generates a blond person).
Applications include implicit multimodal visual feature generation (e.g. can I generate a person that is blond and smiling only from EEG? UIST'20), crowdsourcing, and design (e.g. can a crowd implicitly generate a design that combines their aggregate preferences from their brain signals? CSCW'21, under review).
I am also interested in cognitive modeling, where I have contributed to intelligent tutoring systems, combining a cognitive model of memory (infer how the user learns and forgets) and model-based planning (when to review or show a new item) IUI'21. I have also contributed to open science by replicating a connectionist model of memory, leading to the correction of the original manuscript (ReScience C).