The Mackelab (Prof. Jakob Macke, University Tübingen) is looking for PhD, Postdoc and Scientific Programmer applicants interested in working with us on using deep learning to build, optimize and study mechanistic models of neural computations! In a first project, funded by the ERC Grant DeepCoMechTome, we want to make use of connectomic reconstructions of the fruit fly to build large-scale simulations of the fly brain that can explain visually driven behavior—see, e.g., our prior work with Srinivas Turaga’s group, described in Lappalainen et al., Nature, 2024. In a second project, funded by the DFG through the CRC Robust Vision, we want to use differentiable simulators of biophysical models (Deistler et al., 2024) to build data-driven models of visual processing in the retina. We are open to candidates who are more interested in neurobiological questions, as well as to ones more interested in machine learning aspects (e.g. training large-scale mechanistic neural networks, learning efficient emulators, coding frameworks for collaborative modelling, automated model discovery for mechanistic models, …) of these projects.

This webinar addressed computational perspectives on how animals and humans make decisions, spanning normative, descriptive, and mechanistic models. Sam Gershman (Harvard) presented a capacity-limited reinforcement learning framework in which policies are compressed under an information bottleneck constraint. This approach predicts pervasive perseveration, stimulus‐independent “default” actions, and trade-offs between complexity and reward. Such policy compression reconciles observed action stochasticity and response time patterns with an optimal balance between learning capacity and performance. Jonathan Pillow (Princeton) discussed flexible descriptive models for tracking time-varying policies in animals. He introduced dynamic Generalized Linear Models (Sidetrack) and hidden Markov models (GLM-HMMs) that capture day-to-day and trial-to-trial fluctuations in choice behavior, including abrupt switches between “engaged” and “disengaged” states. These models provide new insights into how animals’ strategies evolve under learning. Finally, Kenji Doya (OIST) highlighted the importance of unifying reinforcement learning with Bayesian inference, exploring how cortical-basal ganglia networks might implement model-based and model-free strategies. He also described Japan’s Brain/MINDS 2.0 and Digital Brain initiatives, aiming to integrate multimodal data and computational principles into cohesive “digital brains.”

I’ll give an overview of internal - or motivational - states in larval zebrafish. Specifically we will focus on the role of the Oxytocin system in regulating the detection of, and behavioral responses to, conspecifics. The appeal here is that Oxytocin has likely conserved roles across all vertebrates, including humans, and that the larval zebrafish allows us to study some of the general principles across the brain but nonetheless at cellular resolution. This allows us to propose mechanistic models of emotional states.

The SARS-CoV-2 pandemic is awash in data, including daily, spatially-resolved COVID case data, virus sequence data, patients `omics data, and mobility data. Journals are now also awash in studies that make use of quantitative modeling approaches to gain insight into the geographic spread of SARS-CoV-2 and its temporal dynamics, as well as studies that predict the impact of control strategies on SARS-CoV-2 circulation. Some, but by no means all, of these studies are informed by the massive amounts of available data. Some, but by no means all, of these studies have been useful — in that their predictions revealed something beyond simple back of the envelope calculations. To summarize some of these findings, in this symposium, we will address questions such as: What do we want from models of disease spread? What can and should be predicted? Which data are the most useful for predictions? When do we need mechanistic models? What have we learned about how to model disease spread from unmet and/or conflicting predictions? The workshop speakers will explore these questions from different perspectives on what data need to be considered and how models can be evaluated. As at other TMLS workshops, each speaker will deliver a 10-minute talk with ample time set aside for moderated questions/discussion. We expect the talks to be provocative and bold, while respecting different perspectives.