Yoshua Bengio

Yoshua Bengio

University of Montréal


Bridging the gap between brains, cognition and deep learning
We start by reviewing connectionist ideas from three decades ago which have fuelled a revolution in artificial intelligence with the rise of deep learning methods. We also discuss the new ideas from deep learning, including a discussion of the newly acquired theoretical understanding of the advantages brought by jointly optimizing a deep architecture. (more…)

Moritz Helmstaedter

Max Planck Institute for Brain Research


Cerebral Cortex Connectomics
Brains are highly interconnected networks of millions to billions of neurons. For a century, we
have not been able to map these connectivity networks. Only recently, using novel electron
microscopy techniques and machine-learning based data analysis, the mapping of neuronal
networks has become possible at a larger scale. (more…)

Sharon Crook

Arizona State University


Reproducibility and rigor in computational neuroscience: Testing the data driven model
As computational models in neuroscience increase in complexity, there are additional barriers for their creation, exchange, and re-use. Successful projects have created standards and open source tools to address these issues, but specific, rigorous criteria for evaluating models against experimental data during model development remain rare. (more…)

Yiota Poirazi

Foundation of Research and Technology-Hellas
Session: developing data-driven models of synapses and neurons


Dendritic contributions to complex functions: insights from computational modeling
My lab (www.dendrites.gr) uses computational modelling approaches to investigate the role of dendrites in learning and memory processes. Our models range in complexity from detailed biophysical single cells, to reduced microcircuits and large scale simplified neuronal networks. Brain areas of interest include the hippocampus, the amygdala, the prefrontal cortex and the visual cortex.

Russ Poldrack

Stanford University
Session: brain imaging standards and best practices


Towards a robust data organization scheme for neuroimaging: the Brain Imaging Data Structure
The field of neuroimaging has been at the vanguard of data sharing, but the utility of shared data has been limited by the lack of standards for data and metadata organization.  I will describe our data sharing efforts via the OpenfMRI and OpenNeuro projects, and how those efforts have been supported by the development of a community standard for data organization: The Brain Imaging Data Structure (BIDS) standard.  (more…)

Camille Maumet

Session: brain imaging standards and best practices


Tools and standards to make neuroimaging derived data reusable
Neuroimaging is becoming increasingly collaborative. More and more brain imaging datasets are made available to the community, effectively creating a massive distributed resource for neuroscientists. But to make the best use of this asset, we need tools & standards to model and understand the diverse sources of variability present in these data. This talk will discuss recent initiatives to represent and share neuroimaging metadata and how these can help us leverage heterogeneous datasets.

Andrew Davison

Centre Nationnal de la Recherche Scientifique (CNRS)
Session: reproducible neuroscience + open science


Improving reproducibility and reuse in computational and systems neuroscience
I will survey recent initiatives to improve reproducibility in computational and systems neuroscience, focusing on automatic capture of workflow provenance, on model sharing, and on community-driven open source tool development.

Michel Dumontier

Maastrictt University
Session: can we harmonize metadata


Accelerating biomedical discovery with FAIR

With its focus on investigating the basis for the sustained existence of living systems, modern biology has always been a fertile, if not challenging, domain to represent knowledge amenable to computational-based discovery. Indeed, the existence of millions of scientific articles, thousands of databases, and hundreds of ontologies, offers an exciting opportunity to reuse our collective knowledge, were we not stymied by incompatible formats, partial and overlapping standards, and heterogeneous data access.


Kenneth Harris

University College London
Session: computational infrastructure for neuroscience: automation / pipelines


High-Dimensional Geometry of the Cortical Population Code as Revealed by 10,000-Cell Recordings
We used 2-photon calcium imaging and improved analysis methods to record the responses of >10,000 neurons in the visual cortex of awake mice, to thousands of natural images. The recorded population code was high-dimensional, with the variance of its dimensions following a powerlaw. (more…)

Upinder Bhalla

National Centre for Biological Sciences
Session: standardization in multiscale modeling – connecting the levels


From neural recordings to subcellular neuronal computation.
Neural circuit activity is as much an outcome of subcellular and molecular computation as it is of spike integration. We are interested in three closely coupled problems: How to simulate multiscale neuronal models that span molecular, electrical, and structural events, how to define such models, and how to systematically parameterize them. (more…)

Fernando Perez

University of California, Berkeley
Session: computational infrastructure for neuroscience: automation / pipelines


Open science and reproducible research on Jupyter
Project Jupyter, evolved from the IPython environment, provides a platform for interactive computing that is widely used today in research, education, journalism and industry.  The core premise of the Jupyter architecture is to design tools around the experience of interactive computing. It provides an environment, protocol, file format and libraries optimized for the computational process when there is a human in the loop, in a live iteration with ideas and data assisted by the computer.


Doina Precup

Doina Precup

McGill University
Session: machine learning in neuroscience

A pioneer in the area known as reinforcement learning, which involves applying computer programs to solving problems by encouraging desired behaviour through rewards

Ivan Soltesz

Stanford University


Towards a complete description of the circuitry underlying sharp wave-mediated memory replay
Our team is making the first attempt to fully understand a cognitively important event, called memory replay during sharp-wave ripples, in terms of the detailed properties of the brain cells involved. We employ large-scale recording technologies to study and manipulate identified cell types in the behaving animal and construct the first full-scale computational model of the brain area


Pierre Bellec

Université de Montréal
Session: reproducible neuroscience + open science


Dealing with clinical heterogeneity in the discovery of new biomarkers of Alzheimer’s disease

Magnetic resonance imaging is routinely used as a biomarker for the prognosis of Alzheimer’s dementia. Prognosis based on machine learning models however has limited positive predictive value, likely due to the heterogeneity of clinical populations.


Nolan Nichols

Session: Can we Harmonize Metadata?


Meaningful (meta)data at scale: removing barriers to precision medicine research

Randomized controlled trials (RCTs) are the gold standard for evaluating therapeutics in patient populations. The data collected during RCTs include a wealth of clinical measures, biomarkers, and tissue samples – the analysis of which can lead to the approval of new medicines that improve the lives of patients. (more…)

Gael Varoquaux

Session: Machine learning in neuroscience


Towards psychoinformatics with machine learning and brain imaging

Informatics in the psychological sciences brings fascinating challenges as concepts or pathologies have fuzzy boundaries and are hard to quantify. Brain imaging brings rich data on the neural substrate of these concepts, yet it is a non trivial link.


Tatyana Sharpee

Computational Neurobiology Laboratory, Salk Institute for Biological Studies
Session: Building a framework for understanding circuit function session


Cortical representation of natural stimuli

In this talk I will describe our recent findings of several organizing principles for how natural visual and auditory stimuli are presented across stages of cortical processing. For visual processing, I will describe how signals in the secondary cortical visual area build on the outputs provided by the first cortical visual area, and how they relate to representation found in subsequent visual areas, such as area V4.


Shaul Druckmann

Stanford University
Session: Building a framework for understanding circuit function


Relating circuit dynamics to computation: robustness and dimension-specific computation in cortical dynamics

Neural dynamics represent the hard-to-interpret substrate of circuit computations. Advances in large-scale recordings have highlighted the sheer spatiotemporal complexity of circuit dynamics, portraying in detail the difficulty of interpreting such dynamics and relating it to computation.