EVENTS





Oxford Neurotheory Forum (ONTF)

Seminars usually take place at 1:30 pm in the Le Gros Clark lecture theatre (view map) unless stated otherwise. This forum is jointly organised by , and . If you would like to be added to the Neurotheory mailing list, please email one of us or use the OxfordTalks platform.

Upcoming Seminars

ics calendar link

 Mon 11 December 2017 13:30
Le Gros Clark Lecture Theatre (map)
Dr Andrew Saxe
Harvard
A theory of the dynamics of deep learning: Consequences for semantic development
Abstract: Anatomically, the brain is deep; and computationally, deep learning is known to be hard. How might depth impact learning in the brain? To understand the specific ramifications of depth, I develop the theory of learning in deep linear neural networks. I give exact solutions to the dynamics of learning which specify how every weight in the network evolves over the course of training. The theory answers fundamental questions such as how learning speed scales with depth, and why unsupervised pretraining accelerates learning. Turning to generalization error, I use random matrix theory to analyze the cognitively-relevant "high-dimensional" regime, where the number of training examples is on the order or even less than the number of adjustable synapses. Consistent with the striking performance of very large deep network models in practice, I show that good generalization is possible in overcomplete networks due to implicit regularization in the dynamics of gradient descent. These results reveal a speed-accuracy trade-off between training speed and generalization performance in deep networks. Drawing on these findings, I then describe an application to human semantic development. From a stream of individual episodes, we abstract our knowledge of the world into categories and overarching structures like trees and hierarchies. I present an exactly solvable model of this process by considering how a deep network will learn about richly structured environments specified as probabilistic graphical models. This scheme illuminates empirical phenomena documented by developmental psychologists, such as transient illusory correlations and changing patterns of inductive generalization. Neurally, the theory suggests that the representation of complex structures resides in the similarity structure of neural population responses, not the detailed activity patterns of individual neurons. Overall, these results suggest that depth may be an important computational principle influencing learning in the brain. Deep linear networks yield a tractable theory of layered learning that interlinks computation, neural representations, and behavior.

About the speaker: Dr. Andrew Saxe is a Swartz Postdoctoral Fellow in Theoretical Neuroscience at Harvard University. He completed his PhD in Electrical Engineering at Stanford University, advised by Jay McClelland, Surya Ganguli, Andrew Ng, and Christoph Schreiner. His dissertation received the Robert J. Glushko Dissertation Prize from the Cognitive Science Society. His research focuses on the theory of deep learning and its applications to phenomena in neuroscience and psychology.
 Mon 26 March 2018 13:30
Le Gros Clark Lecture Theatre (map)
Prof Sam Gershman
Harvard
TBA

Past Seminars

 Wed 29 November 2017
Matthew Botvinick, M.D., Ph.D.
DeepMind and Gatsby Computational Neuroscience Unit, University College London
Prefrontal cortex as a meta-reinforcement learning system
 Fri 24 November 2017
Dr Willem Wybo
Ecole polytechnique federale de Lausanne EPFL
Dynamic compartmentalization in dendrites enables branch-specific learning
 Wed 22 November 2017
Prof Jeff Moehlis
University of California, Santa Barbara
Controlling Populations of Neurons
 Wed 15 November 2017
Dr Dan Goodman
Imperial College London
What is simplicity in a complex auditory world?
 Wed 01 November 2017
Prof Jochen Braun
Otto-von-Guericke-Universitaet Magdeburg
Continuous visual inference and hierarchical attractor dynamics
 Wed 11 October 2017
Prof Thomas Nowotny
University of Sussex
Biophysical model of the olfactory system of honey bees predicts qualitatively different responses to mixtures
 Wed 27 September 2017
Prof Gaute Einevoll
Norwegian University of Life Sciences and University of Oslo
What can we learn from local field potentials (LFPs) recorded in the brain?
 Thu 21 September 2017
Prof Jonathan Rubin
University of Pittsburgh
Basal ganglia activity: The good, the bad, and the depressed
 Wed 13 September 2017
Danilo Jimenez Rezende, PhD
DeepMind
Approximate Inference and Deep Generative Models
 Wed 05 July 2017
Prof Josh Berke
UC San Francisco
Rethinking Reward: Dopamine Signals in Learning and Motivation
 Thu 29 June 2017
Dr Mark Humphries
University of Manchester
A Spiral Attractor Network Drives Rhythmic Locomotion in Aplysia
 Thu 22 June 2017
Dr David Kastner
UCSF
Mechanism underlying the retinal computation of predictive sensitization
 Wed 14 June 2017
Dr Simon Schultz
Imperial College London
Optical Tools for Dissecting Neural Information Processing
 Wed 17 May 2017
Prof Ingo Bojak
University of Reading
Towards state and parameter estimation in neural populations models (of anaesthesia)
 Tue 16 May 2017
Note: joint Cortex Club talk
Dr Christian Machens
Champalimaud Centre for the Unknown
Efficient codes and balanced networks
 Thu 11 May 2017
Note: talk at the room C1, Mathematical Institute, 12-1pm
Dr Kyle Wedgwood
University of Exeter
Riding the waves in spiking neural networks
 Wed 10 May 2017
Note: joint Neurotheory/Cortex club talk at the Le Gros Clark Lecturer Theatre
Prof Alexander Pouget
University of Geneva
The Agony of Choice: Optimal Policies for Value-based Decision Making
 Wed 26 April 2017
Dr Nicolangelo Iannella
University of Nottingham
STDP balance and morphology influences the appearance of the dendritic mosaic
 Wed 05 April 2017
Dr Yann Sweeney
Imperial College London
Diverse population coupling ensures robust yet flexible stimulus representation in a recurrent network model of perceptual learning
 Wed 29 March 2017
Note: joint Neurotheory/Cortex club talk
Prof Xiao-Jing Wang
New York University Shanghai
What does it mean to build a large-scale brain circuit model?
 Thu 23 March 2017
Michael Markie
F1000 (and Wellcome Open Research)
Wellcome Open Research - author led publishing
 Fri 17 March 2017
Dr Thomas E. Nichols
University of Warwick
Large Scale Evaluation of Random Field Theory Inference in fMRI
 Wed 15 March 2017
Dr Nikolaus Kriegeskorte
University of Cambridge
Testing complex brain-computational models to understand how the brain works
 Wed 15 February 2017
Ms Grace Lindsay
Columbia University
Modeling Neural and Performance Correlates of Attention
 Thu 26 January 2017
Dr Tiago Pereira
Imperial College London
Dynamics of hubs: emergent behaviour and phase transitions
 Wed 11 January 2017
Prof Stephen Coombes
University of Nottingham
Next generation neural field modelling
 Wed 14 December 2016
Dr Michael Spratling
Kings College London
Predictive coding as a model of cortical function
 Wed 23 November 2016
Dr Ben Torben-Nielsen
University of Hertfordshire
Neuronal computation: A tale of dendritic structure and function
 Wed 16 November 2016
Mr Dane Corneil
Ecole Polytechnique Federale de Lausanne
Preplay and path planning in maze-like environments with attractor networks
 Wed 19 October 2016
Dr Tara Keck
University College London
Spatial scales of homeostatic plasticity in mouse visual cortex
 Wed 21 September 2016
Dr Tim O'Leary
University of Cambridge
Reconciling flexible neuromodulation, variability and regulatory control in neurons
 Wed 17 August 2016
Mr Alex Seeholzer
Ecole Polytechnique Federale de Lausanne
A mean-field theory for drift-diffusion dynamics of spiking continuous attractor models
 Wed 22 June 2016
Dr Bob Wilson
University of Arizona
Information seeking and randomness drive human exploration
 Mon 20 June 2016
Dr Angela Yu
University of California San Diego
A Decision-Theoretic Framework for Active Cognition
 Mon 13 June 2016
Dr Matthias Munk
Technical University Darmstadt
Neural Dynamics in Prefrontal Cortex: Ensemble Codes, Oscillatory Structure and Avalanches during short-term memory
 Wed 18 May 2016
Dr Cian O'Donnell
University of Bristol
Multidimensional imbalances in excitation/inhibition in Fragile-X Syndrome
 Wed 04 May 2016
Dr Jill O'Reilly
University of Oxford
Control of entropy in internal models
 Mon 25 April 2016
Note: Joint CNCB / Neurotheory seminar
Dr Shaul Druckmann
Janelia Research Campus
Relating Circuit Dynamics to Computation: Robustness and Dimension-specific Computation in Cortical Dynamics
 Wed 06 April 2016
Dr Eleni Vasilaki
University of Sheffield
Emergence of connectivity motifs via the interaction of long-term and sort-term plasticity
 Wed 16 March 2016
Prof Andreas Herz
LMU Munich
Decoding the population activity of grid cells for spatial localization and goal-directed navigation
 Tue 15 March 2016
Note: this talk will be at the Department of Experimental Psychology, C113 Weiskrantz Room between 1pm to 2pm
Dr Thomas Akam
University of Oxford
Optogenetic silencing of anterior cingulate cortex disrupts model-based reinforcement learning in mice
 Thu 10 March 2016
Note: Joint Experimental Psychology / Neurotheory seminar
Dr Keren Haroush
Harvard University
Neuronal underpinnings of social interaction
 Wed 02 March 2016
Dr Pablo Jercog
IDIBAPS and Cellex Inst., Barcelona
Time dependent spatial information carried by the neural population in hippocampal area CA1 during familiarization
 Wed 17 February 2016
Note: Joint CNCB / Neurotheory seminar
Prof Leslie Griffith
Brandeis University
A single pair of neurons links sleep to memory consolidation in Drosophila melanogaster
 Wed 10 February 2016
Mr James Whittington
University of Oxford
Learning in cortical networks through error back-propagation
 Wed 27 January 2016
Prof Peter Latham
Gatsby Unit for Computational Neuroscience, UCL
Probabilistic Synapses
 Wed 13 January 2016
Dr Greg Wayne
Deep Mind
Differentiable Neural Computers for Memory-Based Control
 Wed 16 December 2015
Dr Sophie Deneve
Ecole Normale Superieure
Learning to represent signals spike by spike
 Wed 02 December 2015
Dr Claudia Clopath
Imperial College
Broad inhibitory tuning shaped by synaptic plasticity with homeostasis
 Tue 24 November 2015
Note: Joint CNCB / Neurotheory seminar
Dr Mate Lengyel
University of Cambridge
Adaptation to natural input statistics: a key to dendritic computation and plasticity
 Wed 18 November 2015
Dr Pedro Goncalves
Gatsby Unit for Computational Neuroscience, UCL
Dynamics of an oculomotor integrator revealed by instantaneous optogenetic perturbations
 Wed 11 November 2015
Dr James Bennett
University of Oxford
Patterning of receptive fields by spatiotemporally inseparable correlations and STDP
 Wed 04 November 2015
Arvind Kumar
KTH Royal Institute of Technology/University of Freiburg
Closed-loop feedback control for spiking neuronal network dynamics
 Wed 14 October 2015
Paul Brodersen
University of Oxford
How does the topology of neuronal network change during spatial learning?
 Mon 28 September 2015
Kishore Kuchibhotla
New York University
Cholinergic switching conveys behavioral context by controlling cortical inhibition
 Wed 16 September 2015
Piotr Suffczynski
University of Warsaw
Modelling seizure transitions with neuronal and ionic dynamics
 Fri 11 September 2015
Sen Song
Tsinghua University
Attentional Neural Network: Feature Selection Using Cognitive Feedback
 Wed 08 July 2015
N. Alex Cayco Gajic
University College London
Coordinated neural activity: mechanistic origins and impact on stimulus coding
 Mon 06 July 2015
Yashar Ahmadian
Columbia University
Contextual modulation of gamma rhythms in inhibition stabilized cortical networks
 Wed 24 June 2015
Michael Okun
University College London
Diverse coupling of neurons to populations in sensory cortex
 Fri 19 June 2015
Note: Joint Statistics / Neurotheory seminar
Geoffrey Hinton
University of Toronto/Google
Deep Learning
 Wed 27 May 2015
John Mikhael
University of Oxford
Learning Reward Uncertainty in the Basal Ganglia
 Wed 13 May 2015
Carlos Stein N. Brito
Ecole polytechnique federale de Lausanne EPFL
Theory of synaptic plasticity for receptive field development
 Thu 07 May 2015
Guillaume Hennequin
University of Cambridge
Inhibition-stabilized balanced dynamics account for stimulus-induced changes of noise variability in the cortex
 Wed 08 April 2015
Ali Neishabouri
Imperial College London
Noise and other constraints on the design of neural fibres
 Wed 18 February 2015
Rafal Bogacz
University of Oxford
A tutorial on the free-energy model of perception and learning
 Wed 04 February 2015
Friedemann Zenke
Ecole polytechnique federale de Lausanne EPFL
Memory formation and recall in spiking neural networks
 Wed 21 January 2015
Alejo Nevado-Holgado
University of Oxford
Model-guided analysis of beta band oscillations in the basal ganglia
 Fri 09 January 2015
Patrick Simen
Oberlin College
Interval timing and decision making with an opponent Poisson diffusion model
 Wed 07 January 2015
Adam Ponzi
Okinawa Institute of Science and Technology (OIST)
Edge of chaos in the striatum and for temporal expectation
 Wed 10 December 2014
Hayriye Cagnan
University of Oxford
Phase specific stimulation of tremor oscillators
 Wed 26 November 2014
Mark Humphries
University of Manchester
Tonic dopamine's control of the exploration-exploitation trade-off
 Wed 12 November 2014
Alvaro Tejero-Cantero
University of Oxford
Automatic brain state discovery from multivariate LFP time series
 Wed 29 October 2014
Tim Behrens
University of Oxford
Can we look at neuronal and synaptic computations in humans? And why should we want to.
 Thu 16 October 2014
Blake Richards
University of Toronto
Memories are not formed in a vacuum: how multiple memories are combined during consolidation
 Wed 15 October 2014
Bilal Haider
University College London
Synaptic inhibition regulates cortical activity in vivo.
 Wed 01 October 2014
Adam Packer
University College London
All-optical manipulation and recording of neural circuit activity in vivo.
 Wed 17 September 2014
Guillaume Hennequin
University of Cambridge
From dynamics to statistics: Trial-average trajectories and across-trial variability of cortical responses.
 Tue 10 June 2014
Brent Doiron
Dept. of Mathematics, University of Pittsburgh
Formation of neuronal assemblies and their maintenance during spontaneous cortical activity.
 Wed 28 May 2014
Tim Vogels
CNCB, Oxford
Connection-type specific biases make uniform random network models consistent with cortical recordings.
 Wed 16 April 2014
Bill Podlaski
CNCB, Oxford
Learning sensory representations with retroaxonal feedback.
 Wed 02 April 2014
Christian Pozzorini
Ecole polytechnique federale de Lausanne EPFL
Adaptive coding in single neurons.
 Wed 19 March 2014
Marius Usher
Dept. of Psychology, Tel Aviv University
Absolute vs relative decisions: Contrasting choice models via controlled evidence manipulation.
 Wed 05 March 2014
Timothy Lillicrap
Dept. of Pharmacology, Oxford
Random feedback weights support learning in deep neural networks.
 Wed 19 February 2014
Ben Willmore
Biomedical Science, Oxford
Gain control in the auditory system.
 Wed 05 February 2014
Akihiro Eguchi
Dept. of Experimental Psychology, Oxford
Modelling the development of object shape representations in V4 and TEO.
 Wed 22 January 2014
Shamik Dasgupta
CNCB, Oxford
Models of decision making in fruit flies.
 Wed 08 January 2014
Rui Ponte Costa
Inst. for Adaptive and Neural Computation, Univ. of Edinburgh
Pre and post-synaptic plasticity in local cortical circuits.
 Thu 02 January 2014
Christine Constantinople
Princeton University
Thalamic drive of deep cortical layers.
 Thu 02 January 2014
Rob Froemke
NYU School of Medicine
Cortical plasticity improves sensory perception.
 Wed 11 December 2013
Rafal Bogacz
Div. of Clinical Neuroscience, Oxford
Do we have Bayes' theorem hard-wired in the basal ganglia?
 Wed 27 November 2013
Daniel Walters
Dept. of Experimental Psychology, Oxford
Accurate path integration of head direction: mechanisms for updating a packet of neural activity at the correct speed.
 Wed 13 November 2013
Chris Summerfield
Dept. of Experimental Psychology, Oxford
Adaptation and prediction in perceptual categorisation.