SpeakerOrganized by Kristin Branson and Edda "Floh" Thiels, see link for full list of speakers
Date and LocationWednesday June 26, 2019 8:00am to 2:00pm
The Interdepartmental Graduate Program in Dynamical Neuroscience (DYNS) is hosting the live broadcast of the SfN Virtual Conference
Machine Learning in Neuroscience: Fundamentals and Possibilities
Machine learning methods enable researchers to discover statistical patterns in large datasets to solve a wide variety of tasks, including in neuroscience. Recent advances have led to an explosion in the scope and complexity of problems to which machine learning can be applied, with an accuracy rivaling or surpassing that of humans in some domains.
This virtual conference will illuminate the many ways machine learning and neuroscience intersect in the context of data analysis and modeling brain function, and how neuroscience can benefit from the machine learning revolution.
Join the virtual conference’s Neuronline community to connect with colleagues and share your insight and questions about machine learning ahead of June 26.
Event Date: Wednesday, June 26, 2019, 8 am to 2 pm PDT (11 a.m. to 5 p.m. EDT)
Organizers: Kristin Branson, HHMI Janelia Farm Research Campus; Edda (Floh) Thiels, University of Pittsburgh
Session 1: The Fundamentals of Machine Learning, Speakers: Floh Thiels, Sanjoy Dasgupta, Time: 8 am – 8:45 am PDT (11 am – 11:45 am EDT)
The field of machine learning encompasses a broad range of data modeling and predictive tasks, and for each of these, a variety of approaches has been developed to address different application characteristics (e.g., different types of data). This session will provide a taxonomy of machine learning tasks and algorithms, illustrated by scientific applications.
Session 2: Machine Learning for Automating Analysis of Big Neuroscience Data, Speakers: Kristin Branson, Sebastian Seung, Leila Wehbe, Time: 9 am – 10 am PDT (12 pm – 1 pm EDT)
Across neuroscience, from fMRI brain imaging of cognitive processing in humans to electron microscopy dissection of neuron connectivity in mice, scientists are producing datasets of unprecedented scale and complexity. Machine learning is a powerful tool for automating the processing and analysis of these datasets. In this session, Sebastian Seung and Leila Wehbe will discuss how machine learning is being used to analyze these types of data, and as a source of proposed models for the brain processes involved in the tasks behind the experiments. They will discuss different machine learning approaches and highlight pitfalls to avoid while pursuing them.
Session 3: Machine Learning for Modeling the Brain, Speakers: Andrew Saxe, Kim Stachenfel, Time: 10:30 am – 11:30 am PDT (1:30 pm – 2:30 pm EDT)
In addition to providing state-of-the-art tools for neural data analysis, machine learning methods can be useful to neuroscience as models of neural systems themselves. This session will focus on how machine learning principles provide an orienting normative perspective for making sense of brain data. Andrew Saxe and Kim Stachenfeld will show how understanding properties of learning in an artificial context can translate to insights about learning in the biological context, and discuss how machine learning problems and unexplained neuroscience data can mutually inform each other. Their talks will be augmented with examples from their work highlighting ML-neuro translational insights.
Session 4: Machine Learning for Discovering Structure in Data, Speakers: Floh Thiels, Scott Linderman, Srini Turuga, Time: 11:45 am to 12:45 pm PDT (2:45 pm – 3:45 pm EDT)
Unsupervised machine learning methods combine our prior hypotheses about the world with big data sets to discover new hypotheses. This session will describe the frontier of unsupervised machine learning algorithms and how they are being used to understand neuroscience data. Scott Linderman will present a short tutorial, “Finding Structure in Neural Data: From HMMs to Deep State Space Models,” which will cover both past and new ideas in state space modeling of neural data. Srini Turaga will present the use of variational autoencoders for Bayesian inference of spikes, synaptic inputs, and connectivity from calcium imaging and optogenetic perturbation experiments.
Session 5: Panel Discussion for Future Directions of Machine Learning in Neuroscience, Speakers: Kristin Branson, Andrew Saxe, Kim Stachenfeld, Leila Wehbe, Time: 1 pm – 2 pm PDT
Panelists will share their views on the future of machine learning in neuroscience and answer questions from attendees.
Contact Anna Spickard if you have any questions: firstname.lastname@example.org