mackelab

As we are no longer at research center caesar

Fri, 01 May 2020 00:00:00 +0000

The pages on research, resources and publications will stay up, so that one can find our old work, but no new content will be added.

We are excited about joining the vibrant research community in Tuebingen, with exceptionally strong groups in Machine Learning, AI, as well as in Computational, Experimental, and Clinical Neuroscience!

Check out the Machine Learning Cluster of Excellence, the International Max Planck Research School Intelligent Systems, and the new International Max Planck Research School Mechanisms of Mental Function and Dysfunction!

Check google scholar for new publications, and twitter for updates!

Are choice probabilities explained by serial dependencies?

Thu, 18 Jan 2018 00:00:00 +0000

Our paper “Can serial dependencies in choices and neural activity explain choice probabilities?” by Jan-Matthis, Jakob, and Hendrikje Nienborg has been accepted to Journal of Neuroscience and is available online.

Correlations, unexplained by the sensory input, between the activity of sensory neurons and an animal’s perceptual choice (“choice probabilities”) have received attention from both a systems and computational neuroscience perspective. Conversely, while temporal correlations for both spiking activity (“non-stationarities”) and for a subject’s choices in perceptual tasks (“serial dependencies”) have long been established, they have typically been ignored when measuring choice probabilities. Some accounts of choice probabilities incorporating feedback predict that these observations are linked. Here, we explore the extent to which this is the case. We find that, contrasting with these predictions, choice probabilities are largely independent of serial dependencies, which adds new constraints to accounts of choice probabilities that include feedback.

Poster accepted at COSYNE

Sat, 13 Jan 2018 00:00:00 +0000

Alexandre’s poster submission “Inferring mesoscopic population models from population spike trains” has been accepted at COSYNE this year. There were 704 submissions to COSYNE this year, of which only 56% were accepted.

How does the interplay of single-neuron dynamics and neural connectivity give rise to the rich dynamical properties of neural populations? To tackle this question, it is desirable to have models which exhibit a wide range of population dynamics but remain interpretable in terms of connectivity and single-neuron dynamics. However, many commonly-used statistical models of neural population dynamics are based on generic models of dynamics (e.g. in Macke et al. 2011). Conversely, it has been challenging to link mechanistic spiking network models to empirical population data. To close this gap, we propose to model such data using mechanistic, but low-dimensional and hence statistically tractable models. We approximate neural populations as being composed of multiple homogeneous `pools’ of neurons, and model the dynamics of the aggregate population activity within each pool. We derive the likelihood of parameters (both single-neuron parameters and inter-pool connectivity) given this activity, which can then be used to either optimize parameters by gradient ascent on the log-likelihood, or to perform Bayesian inference using Markov Chain Monte Carlo (MCMC) sampling.

We illustrate this approach on a model based on generalized integrate-and-fire neurons (Schwalger et al., 2017). Using micro- and mesoscopic simulations of multiple neuron pools, we demonstrate that both single-neuron properties (membrane and adaptation constants) and connectivity-parameters (excitatory vs inhibitory connections and connection strengths) can be recovered on simulated data. Moving beyond point estimates, we compute the Bayesian posterior for combinations of parameters using MCMC sampling. Finally, we investigate how the approximations inherent to a mesoscopic population model impact the accuracy of the inferred single-neuron parameters. Ultimately, our method ensures compatibility between experimental multi-population data and mesoscopic dynamical models, by providing methods for statistical inference of low-dimensional mesoscopic models.

Three papers accepted at NIPS

Wed, 06 Sep 2017 00:00:00 +0000

We had three papers accepted at NIPS this year– Jan-Matthis’ and Pedro’s paper Flexible statistical inference for mechanistic models of neural dynamics and Marcel’s paper Extracting low-dimensional dynamics from multiple large-scale neural population recordings by learning to predict correlations were accepted for poster presentations. Artur’s paper Fast amortized inference of neural activity from calcium imaging data with variational autoencoders was selected for a spotlight. There were 3240 submissions to NIPS this year, 678 of which were accepted at the conference, and 112 of which for spotlights.

CorBinian

Mon, 21 Aug 2017 00:00:00 +0000

We put online our new code repository CorBinian for modelling multivariate binary data in Matlab.
It extends and succeeds our previous toolbox, and features new fast MCMC code for regularized fitting of high-dimensional maxEnt models with Rao-Blackwellization, code for minimum probability flow fitting, as well as for computation of specific heat capacity from maxEnt models.

Talk at CNS

Sat, 20 May 2017 00:00:00 +0000

Our abstract ``Flexible Bayesian inference for complex models of single neurons’’ by Pedro J. Goncalves, Jan-Matthis Lueckmann, Giacomo Bassetto and Marcel Nonnenmacher, has been accepted for a talk at the Computational Neuroscience meeting this year.

Cajal Course in Computational Neuroscience

Mon, 16 Jan 2017 00:00:00 +0000

Registration for this years Cajal Course in Computational Neuroscience has opened. The course will be hosted at the Champalimaud Centre for the Unknown in Lisbon, Portugal and is directed by Gilles Laurent from the MPI for Brain Research in Frankfurt, Christians Machens from the Champalimaud Centre for the Unknown and Jakob Macke from the Research Center caesar.

The course is designed for graduate students and postdoctoral fellows from a variety of disciplines. The application deadline is the 20th of March 2017 at midnight, CET time. More information about the course and how to apply visit the official website. This course is part of the CAJAL Advanced Neuroscience Training Programme - an initiative by FENS, IBRO and The Gatsby Foundation - and is hosted by Champalimaud Foundation, Portugal. We are grateful to Google DeepMind for additional funding.

Four Abstracts at Cosyne!

Thu, 12 Jan 2017 00:00:00 +0000

The annual meeting on Computational and Systems Neuroscience is the most important conference for exchange of experimental and theoretical/computational approaches to problems in systems neuroscience. We are happy that all four of our submissions to the Cosyne Conference were successful this year, and will be presented at the conference. In addition, Artur and Jan-Matthis received travel grants.

On day 1 (February 23), Giacomo Bassetto will present his work on data-efficient receptive field estimation (I-27).

On day 2 (February 24), you are spoilt for choice between two posters from the group: In II-3, ``Flexible Bayesian inference for mechanistic models of neural dynamics’’ by Pedro Goncalves, Jan-Matthis Lueckmann, Giacomo Bassetto and Marcel Nonnenmacher, we are presenting an approach for making statistical inference possible and painless for any neuron model of your choice.

In II-77 ``Can serial dependencies in choices and neural activity explain choice probabilities?’’, Jan-Matthis Lueckmann’s work from our collaboration with Hendrikje Nienborg will investigate whether and how choice probabilities are affected by temporal dependencies.

On day 3 (February 25), check out Artur Speiser’s work with Srinivas Turaga and Evan Archer, on ``Amortized inference for fast spike prediction from calcium imaging data.’’– using deep learning to make generative-model based inference of action potentials much faster.

International Max Planck Research School Brain and Behavior

Thu, 17 Nov 2016 00:00:00 +0000

The International Max Planck Research School for Brain and Behavior is a fully funded graduate program in the neurosciences jointly hosted between caesar and the Max Planck Florida Institute for Neuroscience , USA, and in collaboration with partner universities in Bonn and Florida. The Ph.D. program is open to highly qualified and motivated candidates from all over the world who hold an outstanding diploma or master degree. We will be taking in our second cohort of students this year, the deadline is soon (December 1st)– apply at http://www.imprs-brain-behavior.org! Selection symposium will be in March next year, and positions will start in 2017.

Workshop at NIPS

Fri, 26 Aug 2016 00:00:00 +0000

Together with Allie Fletcher (UCLA), Eva Dyer and Konrad Koerding (Northwestern), Jascha Sohl-Dickstein (Google Research) and Joshua Vogelstein (John Hopkins), Jakob is organising a two-day workshop on Brains and Bits: Neuroscience Meets Machine Learning at NIPS 2016. It’s all happening on Friday-Saturday, December 9th-10th, 2016, in Barcelona. We have an exciting line-up of speakers from machine learning, computational neuroscience and computer-science:

Yoshua Bengio, Université de Montréal
Adrienne Fairhall, University of Washington
Demis Hassabis, Google DeepMind
Christos Papadimitriou,UC Berkeley
Terry Sejnowski, Salk Institute, UCSD
Anima Anandkumar, UC Irvine
Mitya Chklovskii, Simons Foundation
David Cox, Harvard
Sophie Denève, ENS
Emily Fox, University of Washington
Surya Ganguli, Stanford
Fred Hamprecht, Heidelberg
Jonathan Pillow, Princeton
Maneesh Sahani, Gatsby Unit, University College London

Deadline for submissions is 29/09/2016– we will have both posters and contributed talks!

More instructions and information on the workshop website.