Algonauts challenge 2023: predicting human fMRI activity in response to visual stimulation

• Název práce v češtině: Soutež Algonauts 2023: predikce lidské fMRI aktivity při stimulaci vizuálními stimuli
• Název v anglickém jazyce: Algonauts challenge 2023: predicting human fMRI activity in response to visual stimulation
• Klíčová slova: hluboké učení|predikce fMRI|vizualní stimuly|vision transformers
• Klíčová slova anglicky: deep learning|fMRI prediction|visual stimuli|vision transformers
• Akademický rok vypsání: 2022/2023
• Typ práce: diplomová práce
• Jazyk práce: angličtina
• Ústav: Katedra softwaru a výuky informatiky (32-KSVI)
• Vedoucí / školitel: Mgr. Ján Antolík, Ph.D.
• Řešitel: skrytý - zadáno a potvrzeno stud. odd.
• Datum přihlášení: 21.02.2023
• Datum zadání: 21.02.2023
• Datum potvrzení stud. oddělením: 24.05.2023
• Datum a čas obhajoby: 13.02.2024 09:00
• Datum odevzdání elektronické podoby: 11.01.2024
• Datum odevzdání tištěné podoby: 11.01.2024
• Datum proběhlé obhajoby: 13.02.2024
• Oponenti: doc. RNDr. Ondřej Bojar, Ph.D.
• Konzultanti: Luca Baroni, M.Sc.

Zásady pro vypracování

Computer vision deep learning models trained on neural responses to visual stimuli have proven to be powerful models in predicting neural responses in visual areas [1,2]. On top of this, numerous studies have shown that there is a strong parallelism between the hierarchical layered structure of these models and the hierarchy of brain areas [3]. For example, stimuli induced neural responses representations in low level brain areas resembles image representation in early layers of object recognition pretrained artificial neural network and it has been shown that selecting matching layers and brain areas is possible to predict neural responses with simple linear readout layers. Improving neural prediction performances can help the understanding of how our brain works, and potentially, provide insights on how to build better machine learning models. In this project the student will take part in the Algonauts 2023 challenge [4]. The goal of the challenge is simple: predicting fMRI voxel responses to natural images visual stimuli. To do so the student will fine pretrained vision transformers models as well as explore alternative approaches. Lastly, the student will analyze the best performing model to gain insights about

Seznam odborné literatury

Cadena, Santiago A., et al. “Deep convolutional models improve predictions of macaque V1 responses to natural images.” PLoS computational biology 15.4 (2019): e1006897.
Cadieu, Charles F., et al. “Deep neural networks rival the representation of primate IT cortex for core visual object recognition.” PLoS computational biology 10.12 (2014): e1003963.
Yamins, Daniel LK, et al. “Performance-optimized hierarchical models predict neural responses in higher visual cortex.” Proceedings of the national academy of sciences 111.23 (2014): 8619-8624.
Gifford, Alessandro T., et al. “The Algonauts Project 2023 Challenge: How the Human Brain Makes Sense of Natural Scenes.” arXiv preprint arXiv:2301.03198 (2023).