From : Wednesday 29 September 2021 - 17:00 To : Friday 01 October 2021 - 21:15
From : Wednesday 29 September 2021 - 17:00
To : Friday 01 October 2021 - 21:15
Workshop on "Towards multipurpose neural network models II: Model testing and model fitting"
Organized by Anton Arkhipov (Allen Institute, Seattle), Gaute Einevoll (NMBU/University of Oslo)
Dates: Wednesday 29 September – Friday October 1st
Time: 5PM – 9.15PM (CET)
Click here to download flyer with abstract and list of speakers as PDF.
Online workshop. Free to attend, but registration is mandatory.
Abstract: Starting with the work of Hodgkin, Huxley, Cole, Rall, Katz, Eccles and others in the 1950s and 1960s, we have a reasonably good understanding of the biophysical principles by which single neurons operate. For neural circuits the understanding is much more limited. Most network studies have considered stylized models with a few populations of identical neurons and focused on explaining a particular experimental phenomenon. However, real neural networks consist of a variety of neuron types and have structured synaptic connections. Furthermore, real networks typically perform multiple functions and can be characterized by a variety of readouts from various measurement modalities, including spiking activity, local field potentials, and others. How can we move towards multipurpose models that incorporate the true biological complexity of neural circuits and faithfully reproduce multiple observables in many different situations?
The first workshop on the topic was arranged in August 2020 (see https://alleninstitute.org/what-we-do/brain-science/events-training/allen-institute-modeling-workshop-2020/ for program and videos of talks). In this second (also virtual) workshop in the series we will focus on two key aspects of the overall endeavor: model testing and model fitting.
Multipurpose network models mimicking real neural circuits will contain numerous model parameters that must be optimized. Efficient methods for fitting of model parameters to experimental data are thus needed. Further, the candidate models must be systematically tested against a variety of experimental data, requiring development of commonly accepted benchmarks and test suites. In the workshop these methodological challenges will be addressed from a variety of angles.
Confirmed speakers:
Anton Arkhipov (Allen Institute)
Marcus Covert (Stanford)
Sharon Crook (Arizona State U.)
James DiCarlo (MIT)
Gaute Einevoll (NMBU/U. Oslo)
Julijana Gjorgjeva (Max Planck Institute and TUM)
Peter Jedlicka (U. Giessen)
Szabolcs Kali (Institute of Experimental Medicine, Budapest, Hungary)
Arvind Kumar (KTH Stockholm)
Jakob Macke (U. Tübingen)
Stefan Mihalas (Allen Institute)
Aaron Milstein (Rutgers U.)
Kanaka Rajan (Mount Sinai)
Atle Rimehaug (U. Oslo)
Frances Skinner (Krembil Brain Institute, University Health Network, and U. Toronto)
Carsen Stringer (Janelia)
Kristin Tøndel (NMBU)
Detailled program:
Wednesday, September 29, 2021
Time (CEST)
Speaker
Affiliation
Title
8:00 am – 8:15 am
5:00 pm – 5:15 pm
Gaute Einevoll
NMBU/U. Oslo
Introduction
8:15 am – 8:55 am
5:15 pm – 5:55 pm
Jakob Macke
U. Tübingen
Keynote Simulation-based inference: Bridging the gap between mechanistic models and machine learning.
8:55 am – 9:25 am
5:55 pm – 6:25 pm
Aaron Milstein
Rutgers U.
Nested parallel simulation and multi-objective optimization of neuronal cell and circuit models
9:25 am – 9:40 am
6:25 pm – 6:40 pm
Break
9:40 am – 10:10 am
6:40 pm – 7:10 pm
Atle Rimehaug
U. Oslo
Enhancing model constraints by utilizing current source densities
10:10 am – 10:40 am
7:10 pm – 7:40 pm
Kristin Tøndel
NMBU
Facilitating optimization using metamodelling
10:40 am – 10:50 am
7:40 pm – 7:50 pm
10:50 am – 11:20 am
7:50 pm – 8:20 pm
Frances Skinner
Krembil Brain Institute, University Health Network, and University of Toronto
Clarity in model development and goals leads to model linkages and biological insights
11:20 am – 12:10 pm
8:20 pm – 9:10 pm
Panel debate – All participants of the day.
Thursday, September 30, 2021
Time (PDT)
Anton Arkhipov
Allen Institute
James DiCarlo
MIT
Keynote Reverse Engineering Visual Intelligence
Sharon Crook
Arizona State U.
Testing the Data-driven Model
Szabolcs Kali
Institute of Experimental Medicine, Budapest, Hungary
Systematic construction and evaluation of models of rodent hippocampal neurons
Peter Jedlicka
U. Giessen
Building consistent and robust models of hippocampal granule cells and CA1 pyramidal cells
Stefan Mihalas
Computing with a mess: How nonstationary, heterogeneous and noisy components help the brain’s computational power
Friday, October 1, 2021
8:00 am – 8:40 am
5:00 pm – 5:40 pm
Markus Covert
Stanford
Special lecture Simultaneous cross-evaluation of heterogeneous E. coli datasets via mechanistic simulation
8:40 am – 9:20 am
5:40 pm – 6:20 pm
Kanaka Rajan
Mount Sinai
Keynote Data-constrained neural network models of adaptive learning in the brain
9:20 am – 9:40 am
6:20 pm – 6:40 pm
Julijana Gjorgjeva
Max Planck Institute and TUM
Biologically plausible learning in developing networks
Carsen Stringer
Janelia
Rastermap: Extracting structure from high-dimensional neural data
Arvind Kumar
KTH Stockholm
Structure and activity dynamics relationship in biological neuronal networks: Measurements and models