Dear All, The CEU Department of Cognitive Science and the Center for Cognitive Computation (CCC) invites you to the following talk. *Please note that this event will be held at CEU Budapest site (N13. building), from 5 PM.* Speaker: Ádám Gosztolai (EPFL) Title: Interpretable representations of neural dynamics using geometric deep learning https://arxiv.org/abs/2304.03376 Abstract: It is increasingly recognised that computations in the brain and artificial neural networks can be understood as outputs of a high-dimensional dynamical system conformed by the activity of large neural populations. Yet revealing the structure of the underpinning latent dynamical processes from data and interpreting their relevance in computational tasks remains a fundamental challenge. A prominent line of research has observed that task-relevant neural activity often takes place on low-dimensional smooth subspaces of the state space called neural manifolds. However, there is a lack of theoretical frameworks for the unsupervised representation of neural dynamics that are interpretable based on behavioural variables, comparable across systems, and decodable to behaviour with high accuracy. To address these challenges, we introduce Manifold Representation Basis Learning (MARBLE), a fully unsupervised representation-learning framework for non-linear dynamical systems. Our approach combines empirical dynamical modelling and geometric deep learning to transform neural activations during a set of trials into statistical distributions of local flow fields (LFFs). Our central insight is that LFFs vary continuously over the neural manifold, allowing for unsupervised learning, and are preserved under different manifold embeddings, allowing the comparison of neural computations across networks and animals. We show that MARBLE offers a well-defined similarity metric between different neural systems that is expressive enough to compare computations and detect fine-grained changes in dynamics due to task variables, e.g., decision thresholds and gain modulation. Being unsupervised, MARBLE is uniquely suited to biological discovery. Indeed, we show that it discovers more interpretable neural representations in several motor, navigation and cognitive tasks than generative models such as LFADS or (semi-)supervised models such as CEBRA. Intriguingly, this interpretability implies significantly higher decoding performance than state-of-the-art. Our results suggest that using the manifold structure yields a new class of algorithms with higher performance and the ability to assimilate data across experiments. Time: 17:00, Wednesday, March 20., 2024. Location: CEU Budapest site (1051 Budapest, Nádor u. 15.) N13. building, room 302.* and Zoom (Meeting ID: 975 2152 9826<https://ceu-edu.zoom.us/j/97521529826?pwd=bXhOTDNzK054VFd2cUdMcVVCMkdUUT09> Passcode: 996748) *Anyone not affiliated with CEU wishing to attend in-person in Budapest must RSVP to vargai(a)ceu.edu to get access to the lecture hall. Please, be informed that video/photo recording might take place at the event and the edited version of the video material might be published to communicate or promote CEU's activities. Please, find our Privacy Notice here<https://www.ceu.edu/privacy>cy>. Best regards, Ildikó Varga Department Coordinator (Budapest) Department of Cognitive Science [cid:a9c19747-9116-484d-b203-0aeefb20295d] H-1051 Budapest Nádor u. 15. FT room 404. tel: +36-1 327-3000 2941 http://www.ceu.edu<http://www.ceu.edu/> http://cognitivescience.ceu.edu<http://cognitivescience.ceu.edu/> ______________________________________________ Subscribe by sending an empty mail to talks-subscribe(a)cogsci.ceu.edu Unsubscribe by sending an empty mail to talks-unsubscribe(a)cogsci.ceu.edu