· Organizers · Abstract · Schedule ·

Adam J Calhoun
Talmo Pereira
Scott Linderman
John Cunningham
Liam Paninski

One of the major goals of neuroscience is to understand the biological mechanisms underlying behavior. Modern recording technologies now enable us to simultaneously measure the activity of hundreds of neurons while making high-dimensional measurements of behavior, yet we struggle to make use of all of this information. Traditional tools such as PCA and GLMs seem insufficient to fully capture the richness and dynamics of these data sets suggesting the need for new computational methods. Combining these new technologies and techniques will offer an unprecedented opportunity to study the relationship between neural activity and natural behaviors. This workshop will bring together a mix of experimental and computational neuroscientists to address these major challenges, as well as identify where models developed for one community can be profitably used by the other community.

Day 1: Automated Behavioral Analysis
While great strides have been made in neural data analysis, behavioral analysis has lagged far behind. Traditionally, quantification and analysis of behavioral data has largely relied on laborious and subjective manual labeling. With recent advances in machine learning and related fields, more powerful and unsupervised computational methods for signal processing and for segmenting behavioral time series into biologically meaningful data have begun to be adopted more widely in the neuroscience community. However, this rapidly coalescing community does not yet have a single place to gather to discuss these techniques or their limitations. By unifying the diverse perspectives of systems neuroscientists and computer vision/machine learning researchers in the first day of this workshop, we hope this burgeoning community will be able to overcome the challenges of dealing with highly heterogeneous behavioral data. In particular, we believe that these methods can help in finally answering questions from classic ethological frameworks with an emphasis on its mechanistic links to systems neuroscience.

Day 2: Toward Joint Neuro-Behavioral Analysis
Ultimately, our goal is to relate these rich behavioral measurements to underlying neural activity, which we can now collect at a massive scale. The second day will emphasize the challenges of modeling joint neural and behavioral recordings when both domains are equally complex. Interestingly, we have seen a convergence of methodology, with both neural and behavioral analyses drawing upon similar sets of tools. Day 2 will highlight the many similarities between neural and behavioral modeling and leverage these similarities to suggest new directions for future work.

Monday, 27 February 2017 (Day 1 – Automated Behavior Analysis)
8:00 – 8:05     Introduction (Adam J Calhoun)
8:10 – 8:40      Ji Hyun Bak (Adaptive optimal training of animal behavior)
8:45 – 9:15     Gordon Berman (Predictability and hierarchy in animal behavior)
9:15 – 9:45      Coffee Break
9:45 – 10:15    Adam J Calhoun (Estimating behavioral state)
10:20 – 10:50 Benjamin de Bivort (Systematic exploration of unsupervised methods for mapping behavior)

4:30 – 5:00     Andre Brown (Natural variation in worm behaviour)
5:05 – 5:35      Megan Carey (Establishing a quantitative framework for locomotor coordination with LocoMouse)
5:35 – 6:05     Coffee Break
6:05 – 6:35     Elizabeth Buffalo (Large-scale recordings of memory and spatial navigation in the primate hippocampus)
6:40 – 7:10    Panel discussion (Chaired by Talmo Pereira)
  Panelists:   Ben de Bivort (Harvard), Gordon Berman (Emory), Andre Brown (Imperial College London)

Tuesday, 28 February 2017 (Day 2 – Toward Joint Neuro-Behavioral Analysis)
8:00 – 8:05     Introduction (Scott Linderman)
8:10 – 8:40     Kristin Branson (Mapping behavior to neural anatomy using machine vision and thermogenetics)
8:45 – 9:15      Nicholas Foti (Automatically Parsing Intracranial EEG using Bayesian Nonparametric Dynamic Models)
9:15 – 9:45      Coffee Break
9:45 – 10:15    Brian Duistermars (A neural module for threat display in Drosophila)
10:20 – 10:50   Rui Costa (Organizing self-paced actions)

4:30 – 5:00     Scott Linderman (Recurrent Switching Linear Dynamical Systems for Neural and Behavioral Data)
5:05 – 5:35     Andrea Giovannucci (Automated gesture tracking in head-fixed mice)
5:35 – 6:05     Coffee Break
6:05 – 6:35     Robert Sandeep Datta (Relating Brain to Behavior through Motion Sequencing)
6:40 – 7:10    Panel discussion (Chaired by Talmo Pereira)
  Panelists:   Robert Sandeep Datta (Harvard), Kristin Branson (Janelia Research Campus), Scott Linderman (Columbia)