OCNC: OIST Computational Neuroscience Course

OCNC2010

Okinawa Computational Neuroscience Course 2010

June 14 - July 1, 2010 in Okinawa, Japan

Okinawa Computational Neuroscience Course 2010

The aim of the Okinawa Computational Neuroscience Course is to provide opportunities for young researchers with theoretical backgrounds to learn the latest advances in neuroscience, and for those with experimental backgrounds to have hands-on experience in computational modeling.

We invite graduate students and postgraduate researchers to participate in the course, held from June 15th through July 2nd at an oceanfront seminar house of the Okinawa Institute of Science and Technology.

In 2004-2006, OCNC focused on the brain's computation at different levels: Bayesian computation by neural populations (2004), learning and prediction for behavior (2005), single neurons as computational devices (2006). In 2007, the course included a wide array of topics: single neurons, networks, and behaviors.

In 2008-2010, the course will run for three weeks with ample time in the afternoon for student projects. The first week will be devoted to methods only, with introductory talks in the morning and software tutorials in the afternoon.

The sponsor will provide lodging and meals during the course and support travel for those without funding. We hope that this course will be a good opportunity for theoretical and experimental neuroscientists to meet each other and to explore the attractive nature and culture of Okinawa, the southernmost island prefecture of Japan.

Schedule

A list of lecture topics can be found at the program page.

January 4:: Application process opens

February 15:: Application process closes

Early March:: Notification of acceptance

Mid March:: Confirmation of acceptance

June 13:: Arrival

June 14-19:: Methods

June 21-26:: Neurons, Networks and Behaviors I

June 27-July 1:: Neurons, Networks and Behaviors II

July 2:: Departure

Organizers

Lecturers

Tutors

Neural Microcircuitry Conference

Interested students will be able to attend evening and Sunday sessions of this conference, separate registration required.

OCNC 2010 Program

All lectures take place in the Seminar Room, OIST Seaside House unless otherwise indicated.

Week 1: Methods (Jun 14-19)

Monday, June 14

9:15-9:30: Introduction by the course organizers
9:30-10:30: Opening Lecture: Brain, Information and Mathematical Neuroscience (Shun-Ichi Amari)

11:00-12:30

Parallel track:

Biologists: Introduction to numerical methods for ordinary and partial differential equations (Kenji Doya)

Theoreticians: Neurobiology for Mathematicians (Meeting Room 1) (Gordon Arbuthnott)

14:00-16:00: Student poster presentations I (Abiva-Kanitscheider)

16:00-18:00: Student poster presentations II (Kubánek-You)

19:00-21:00: Reception

Tuesday, June 15

09:30-12:30: Introduction to modeling voltage-gated channels and interpretation of imaging data (Thomas Knöpfel)

14:00-15:00: Introduction of the Tutors

15:00-16:00: A broad overview of the different simulation environments and modeling tools available

16:30-18:30: Python tutorial

Wednesday, June 16

09:30-12:30: Introduction to modeling morphologically detailed neurons (Klaus Stiefel)

14:00-16:00: NEURON tutorial

17:00-18:00: Matlab tutorial

Thursday, June 17

09:30-12:30: Introduction to integrate-and-fire models (Romain Brette)

14:00-16:00: NEST tutorial

16:30-18:30: Advanced NEURON tutorial

Friday, June 18

09:30-12:30: Introduction to modeling biochemical reactions, diffusion and reaction-diffusion systems (Erik De Schutter)

14:00-16:00: STEPS tutorial or meeting with Dr. Brette or Dr. Knöpfel

16:30-17:30: High performance computing tutorial

17:30-18:30: Good modeling practice tutorial

Saturday, June 19

09:30-12:30: Introduction to reinforcement learning and Bayesian inference (Kenji Doya)

14:00-18:00: Project work

Week 2: Neurons, Networks and Behavior I (Jun 21-26)

Monday, June 21

09:30-12:30: Microcircuits involved in motion – insights gained through modeling (Jeanette Hellgren Kotaleski)

14:00-16:00: Project work or meeting with Dr. De Schutter

16:00-18:00: Project work

Tuesday, June 22

09:30-12:30: Mechanisms of dendritic integration and plasticity (Nelson Spruston)

14:00-16:00: Project work, or visit to Lab Building 1, or meeting with Dr. Doya or Dr. Stiefel

16:00-18:00: Project work

Wednesday, June 23

09:30-12:30: Psychological and Neural Reinforcement Learning (Peter Dayan)

14:00-16:00: Project work or meeting with Dr. Kotaleski or Dr. Spruston

16:00-18:00: Visit to Lab Building 1

Thursday, June 24

09:30-12:30: Cerebellar internal model, Purkinje cell LTD as supervised learning rule supported by bioinformatics, and control of learning degree-of-freedom by electrical coupling in Inferior Olive nucleus (Mitsuo Kawato)

14:00-16:00: Project work or meeting with Dr. Dayan

16:00-18:00: Project work

Friday, June 25

09:30-12:30: Extending the notion of receptive field for studying high-order visual neurons (Izumi Ohzawa)

14:00-16:00: Project work or meeting with Dr. Kawato

16:00-18:00: Project work

Saturday, June 26

09:30-12:30: Network Models of Primary Visual Cortex (Klaus Obermayer)

14:00-18:00: Project work

Week 3: Neurons, Networks and Behavior II (Jun 28-Jul 1)

Monday, June 28

09:30-12:30: Technologies for in vivo circuit research (Alla Karpova)

14:00-16:00: Project work or meeting with Dr. Obermayer

16:00-18:00: Project work

Tuesday, June 29

09:30-12:30: Large-scale modeling of the brain (Eugene Izhikevich)

14:00-16:00: Project work or meeting with Dr. Karpova or Dr. Izhikevich

16:00-18:00: Project work

Wednesday, June 30

09:30-12:30: Maximally informative irregularities in neural circuits (Tatyana Sharpee)

14:00-16:00: Project work or meeting with Dr. Sharpee or Dr. Gutkin

16:00-18:00: Project work

Thursday, July 1

09:30-12:30: Theoretical approaches to dynamics of neuronal excitability: spike generation, neuromodulation, synchrony and oscillatory dendrites (Boris Gutkin)

14:00-18:00: Student project presentations

19:00-21:00: Banquet

Lecture Abstracts & Readings

Shun-ichi Amari

Opening Lecture: Brain, Information and Mathematical Neuroscience

The brain consists of a huge number of mutually connected neurons. We need to study molecular and cellular basis of functions of neurons as well as networks or systems level functions. The brain is made of living materials, which are very special materials. Their functions are mostly information processing and memory formation. Further our mind has emerged and realized by using those materials and their functions. We will overview a history of evolution of the universe, showing how life has emerged from materials, what is the principal role of life and information in it, as well as the role of mind.

Theoretical neuroscience, in particular mathematical neuroscience, is one of the very important constituents of the brain research, although it has not yet been completely established. I will discuss its methods and scopes, introducing principles of mathematical neuroscience.

Gordon Arbuthnott

NEUROBIOLOGY FOR MATHEMATICIANS…

The usual accusation of 'unreasonable naivety', which is often presented to computational scientists by experimental biologists, is sometimes only a cover for their own ignorance of computational methods.

On the other hand, just because biology is complicated is no reason not to try to understand it! Perhaps the best format for this session would be a question and answer one. Not that I will have all the answers but at least some of the systems biology of the nervous system is within my ken and we can go through several of the brain areas that interest students in this year's course.

In the past I've used 'Brain Architecture' by Larry Swanson as a backbone to talk about order in the chaos of neuroanatomy but there is no reason to stick to the text if you would like to talk about other things. In particular there are very few details on synaptic relationships in Swanson's book. In fact there are very few details….

One recurring theme will inevitably be that many of the quantitative answers to questions that interest you – are extraordinarily hard to come by. We can cover why that is too!

Lets see what we can learn from each other in the time!

Gordon.

Romain Brette

Introduction to integrate-and-fire models

This lecture will focus on phenomenological descriptions of neurons known as "integrate-and-fire models", which are convenient analytical and numerical tools to understand neural function. In the first part of the lecture, I will show why these are reasonable neuron models, how to relate them with biophysical descriptions of neurons and how to model various electrophysiological properties (e.g. adaptation). I will then present the general properties of these models: coincidence detection vs. integration, balanced vs. oscillator regime, reliability of spike timing, irregularity of interspike intervals. In the second part of the lecture, I will introduce a few concepts of network dynamics and basic elements of synaptic plasticity models. The concepts will be illustrated with concrete computational examples using the Brian simulator.

Peter Dayan

Psychological and Neural Reinforcement Learning

We will expand on Kenji Doya's description of computational and neural aspects of reinforcement learning. We will review four systems involved in control: three instrumental decision-makers (model-based, model-free and episodic); and one Pavlovian controller. We will consider computational, psychological and neural aspects of each, and the ways that they can cooperate and compete. We will consider additional factors in the models such as prior expectations, uncertainty and exploration, and also model-free and model-based routes to psychiatric conditions such as depression.

Erik De Schutter

Modeling biochemical reactions, diffusion and reaction-diffusion systems

In my first talk I will use calcium dynamics modeling as a way to introduce deterministic solution methods for reaction-diffusion systems. The talk covers exponentially decaying calcium pools, diffusion, calcium buffers and buffered diffusion, and calcium pumps and exchangers. I will describe properties of buffered diffusion systems, ways to characterize them and new approximations that we developed for use in large neuron models.

In my second talk I will turn towards stochastic reaction-diffusion modeling.Two methods will be described: Gillespie's Stochastic Simulation algorithm extended to simulate diffusion and particle-based methods. I will discuss some of the problems in generating correct descriptions of microscopic 3D geometries and briefly describe the STEPS software. I will then describe two applications of stochastic reaction modeling of LTD induction in Purkinje cells and stochastic diffusion modeling of anomalous diffusion in spiny dendrites.

Kenji Doya

Introduction to reinforcement learning and Bayesian inference

The aim of this tutorial is to present the theoretical cores for modeling animal/human action and perception. In the first half of the tutorial, we will focus on "reinforcement learning", which is a theoretical framework for an adaptive agent to learn behaviors from exploratory actions and resulting reward or punishment. Reinforcement learning has played an essential role of understanding the neural circuit and neurochemical systems behind adaptive action learning, most notably the basal ganglia and the dopamine system. In the second half, we will familiarize ourselves with the framework of Bayesian inference, which is critical in understanding the process of perception from noisy, incomplete observations.

Boris Gutkin

Theoretical approaches to dynamics of neuronal excitability: spike generation, neuromodulation, synchrony and oscillatory dendrites

Neurons are far from passive integrators of their inputs. Ample data shows that in contrast to linear integrate and fire models, neurons perform active non-linear computations hence the structure if their activity is only partially related to the structure of the input. A classical example of such phenomenon is the voltage dependent mechanism for spike-generation explained Hodgkin and Huxley. More modern example include generation of various kinds of bursts, intermittency in spiking and intrinsic voltage-dependent non-linearities in the dendrties (dendritic spikes).

Over the path of my career, I have been interested in how the properties of neuronal excitability influence the behavior of neurons and furthermore how neuromodulators (e.g dopamine, acetylcholine) change this dynamics. By dynamics of neural excitability I mean the active spike generation, regenerative neural firing, and active dendritic properties. In my lecture I will discuss a geometric approaches to spike generating dynamics, show how to characterize these with Phase Response Functions and go over the weakly-coupled oscillator theory that allows us to link the spike-generating dynamics to collective network behavior. I will also show evidence of how neuromodulation shapes the spike-generation and what that implies to synchrony and asynchrony of networks. In the second part of the lecture I will show how the same mathematics can be used to understand the dynamics of active dendrites that generate oscillations. In conclusion I will show how we can use this theory to understand the possible mechanisms for the formation of the entorhinal grid-fields and general implications for dendritic integration.

Jeanette Hellgren Kotaleski

Microcircuits involved in motion – insights gained through modeling

One important goal of computational modeling is to synthezise data received from experiments, and predict the computational role of the different elements in a more physiological context, e.g. by scaling up the system or by extrapolating to ongoing network activity. To understand the computations going on in various neural systems, the microcircuitry is an important starting point. Understanding the microcircuit requires, however, a multi-scale approach. I will in my tutorial discuss insights gained about the microcircuitry in motor systems through recent modelling work. I will highlight the longstanding work related to the motor system in the lamprey. The lamprey is one of the few vertebrates in which the cellular and synaptic level mechanisms for goal-directed motor behaviour, including locomotion, steering and control of body orientation, are well described. Also recently the basal ganglia have been delineated in the lamprey. The insights gained from combining large-scale computational modeling with detailed experiments are reviewed. We are today for example able to model the motor system with biophysically detailed neurons and with the approximate number of neurons that are responsible in the behaving animal. I will show how activity in the whole motor network, including direction and speed of motion can be controlled from only a subpart of the network, highlighting the self-organizing nature of biological neuronal networks.

Eugene Izhikevich

Large-scale modeling of the brain

I will describe all the major steps in building a detailed large-scale thalamocortical model based on experimental measures in several mammalian species. The model spans three anatomical scales. (i) It is based on global (white-matter) thalamocortical anatomy obtained by means of diffusion tensor imaging (DTI) of a human brain. (ii) It includes multiple thalamic nuclei and six-layered cortical microcircuitry based on in vitro labeling and three-dimensional reconstruction of single neurons of cat visual cortex. (iii) It has 22 basic types of neurons with appropriate laminar distribution of their branching dendritic trees. The model simulates one million multicompartmental spiking neurons calibrated to reproduce known types of responses recorded in vitro in rats. It has almost half a billion synapses with appropriate receptor kinetics, short-term plasticity, and long-term dendritic spike-timing-dependent synaptic plasticity (dendritic STDP). The model exhibits behavioral regimes of normal brain activity that were not explicitly built-in but emerged spontaneously as the result of interactions among anatomical and dynamic processes. I will describe spontaneous activity, sensitivity to changes in individual neurons, emergence of waves and rhythms, and functional connectivity on different scales.

Alla Karpova

Technologies for in vivo circuit research

Understanding information processing by populations of neurons requires monitoring their activity during behavior as well as the ability to perturb this activity. Recent technical advances have given experimental neuroscientists a remarkable set of methods: extracellular electrophysiological measurement of hundrends and soon thousands of neurons simulaneously; recording of the subthreshold activity of multiple neurons in awake moving animals; visualization of somatic calcium transients in a cell-type specific manner using two-photon microscopy as well as perturbation of circuit activity using genetically-encoded tools, to name a few. Although I will only have time to speak about a subset of these methods, I will contrast the currently available tools in relation to how they are able to meet different experimental challenges as well as how they will help advance our understanding of neural circuit function from the computational point of view. Specifically, I will outline the advantages and disadvantages of current methods to measure neuronal activity in fixed and freely moving animals, and highlight how the field is attempting to combine the best of electrophysiology and imaging with freely moving behavior. In addition, I will describe current strategies for spatially and temporally precise circuit perturbations. Finally, I will discuss how these methods can be combined with identification of cell types and sub-circuits.

Mitsuo Kawato

Cerebellar internal model, Purkinje cell LTD as supervised learning rule supported by bioinformatics, and control of learning degree-of-freedom by electrical coupling in Inferior Olive nucleus

In my first talk I will introduce one of the most influential computational models of the cerebellum: cerebellar internal theory [1]. Each microzone of the cerebellar cortex acquires an internal model of some dynamical processes outside the cerebellum while guided by climbing fiber inputs as an error signal in supervised manner. Several experimental supports to this theory are also provided. Then I will move on to a systems biology model of Purkinje cell long-term depression LTD. Spike timing dependent plasticity of LTD is understood by IP3 dynamics as the eligibility trace. Memory itself is modeled by MAP kinase positive feedback loop model and this part is well supported by a series of experiments [2,3]. In my second talk, I will discuss about functions of electrical junctions between inferior olive neurons, whose density is highest in the mammalian brain. Synchronization, rhythmicity and chaotic firing [4] of inferior olive neurons and Purkinje cells can control degrees of freedom in cerebellar supervised learning.

Thomas Knöpfel

Introduction to modeling voltage-gated channels and interpretation of optical imaging data

Voltage-gated ion channels provide the ionic mechanisms that along with passive membrane properties determine the basic electrical behavior of nerve cells. For numerical simulations (compartmental models) the time and voltage-dependent properties of voltage-gated ion channel species are generally well described by the general formalism formulated by Hodgkin and Huxley in the 50ths where transitions between opened (conducting) and closed (non-conducting) states are determined by voltage-dependent rate constants.

After introducing these basics, I will highlight potential problems that frequently arise from the discrepancy between real electrophysiological data and idealistic modeling.

In the second portion of my presentation I will introduce optical imaging methods. The focus will be on interpreting imaging data with the aim to use them as a data source for computational models of single cells and large neuronal circuits. Finally, I will talk about the emerging optogenetic approaches that appear particularly well suited as a link between cellular neurophysiology and circuit modeling.

Klaus Obermayer

Network Models of Primary Visual Cortex

Primary visual cortex in higher animals like primates and cats is one of

the best characterized cortical areas and serves as a paradigmatic area

for understanding visual processing and cortical computation in general.

In my presentations I will describe computational approaches based on network

models of different complexity (from rate models to spiking models) and

discuss, how those models can be applied in oder to evaluate hypotheses

about the response properties of visual cortical neurons. Topics will cover

the computation of feature selective responses, the modulation of neural

response properties by stimulus context, the role of background activity

and noise, neural adaptation and activity dependent plasticity, and

implications for neural coding and the representation of visual information.

Izumi Ohzawa

Extending the notion of receptive field for studying high-order visual neurons

A standard definition of the receptive field of a visual neuron is the area of visual space in which stimuli can influence responses of the neuron. Applying this definition to neurons in high-order areas along the visual pathway is uninteresting, because receptive fields just become increasingly larger in high-order visual areas, providing minimally useful information about the stimulus specificity. I will describe our recent efforts in extending the notion of receptive fields to more useful space that is not necessarily limited to spatial dimensions. Examples of actual experimental measurements of such receptive fields are presented.

Tatyana Sharpee

Maximally informative irregularities in neural circuits

The question of how the nervous system achieves reliable computation using unreliable components has long interested scientists going back to John von Neumann. This is now a topic of renewed interest not only in neuroscience but also in engineering. In neuroscience, the unreliability in neural circuits is thought to be tied to the remarkable energy-efficiency of neural information processing, whereas in emerging nanotechnologies a significant number of faulty components may be impossible to isolate necessitating the need for extensive error-correction at the systems level.

In the first part of this talk, I will describe a theoretical framework for finding optimal sigmoidal input/output functions for biological networks based on the concept of decision boundaries. We will compare solutions for Gaussian, and more naturalistic exponential (Laplace) stimulus distributions, for independent and interacting neurons, and examine when and how optimal solutions change as a function of noise in individual neurons.

The second part of the talk will be devoted to applications of this approach to retinal circuits. We will discuss how nearly optimal solutions can be obtained in the retina by balancing two types of irregularities: those in receptive field (RF) shapes and those of RF center positions against each other. Either the scatter in receptive field center positions or irregularities in RF shapes would have been significantly detrimental for the retinal performance alone. A comparison with data on RF mosaics in the primate retina reveals that information maximization can predict irregularities in individual RFs shapes from the knowledge of RF center positions. Thus, in the retina we find one mechanistic demonstration of how irregular neural circuits could achieve a near-optimal performance.

Nelson Spruston

Mechanisms of dendritic integration and plasticity

This lecture will summarize work, done over the course of the last decade and more, on the subjects of dendritic integration and plasticity in pyramidal cells. The focus will be on work from my own lab, but I will also describe quite a bit of work from other labs that reinforce (or in some cases challenge) ideas that have emerged from our work. The main goal will be to convey our current thinking about how pyramidal neurons function. Both in the hippocampus (where we do most of our work) and in the cortex, pyramidal neurons have two dendritic trees: a basal tree and an apical tree. Collectively, these dendrites receive input from tens of thousands of synapses. How are these inputs integrated to determine when the cell should fire an action potential? Where in the cell does the action potential begin and how does it propagate in the dendrites? What kinds of electrical events in dendrites can lead to synaptic plasticity? An important aspect of answering these difficult questions it to understand how myriad types of ion channels function in the context of a complex dendritic tree. To address this problem, my laboratory has combined experimental and computational approaches. A major focus of the lecture will be on how these two approaches can be effectively combined in a synergistic way to advance a better understanding of dendritic integration and plasticity.

Klaus Stiefel

Introduction to modeling morphologically detailed neurons

I will first motivate the study of single neurons by showing how many fascinating computation they can carry out. Then I will introduce cable-theory, the mathematical description of current flow in dendrites. From cable theory, I will then move to multi-compartmental modeling, the simulation of complex morphologically detailed models of neurons. I will show the usefulness of multi-compartmental modeling by talking about some classic and insightful studies using this technique. I will briefly discuss non-electrical ways of information processing in neurons and how they can be modeled.

Finally, I will talk about some of the research dealing with modeling complex morphologies in our lab, particularly the generation of optimized neuronal morphologies for chosen computational functions.

Students

Abiva, Jeannine
Ahmadi, Mandana
Arizumi, Nana
Bush, Jeffrey
Bölinger, Daniel
Clemens, Jan
Cury, Kevin
Friedrich, Johannes
Ganguli, Deep
Giacomantonio, Clare
Herreros, Ivan
Hesse, Janina
Jimenez Rezende, Danilo
Kanitscheider, Ingmar
Kubánek, Jan
Kurikawa, Tomoki
Lindahl, Mikael
Luo, Ray
Pérez Schuster, Verónica
Romani, Armando
Schiess, Mathieu
Singh, Anand
Taouali, Wahiba
Tsai, Ming-Chi
Williford, Jonathan
Yamashita, Okito
You, Hongzhi
van Versendaal, Danielle

Jeannine Abiva

Affiliation:	The University of Iowa
URL:	http://myweb.uiowa.edu
About:	I have just started to work on a project that studies learning and memory. For this project, I will need to know how to interpret the multi-unit recording data to create a model for episodic memory. Since the University of Iowa does not offer a course in computational neuroscience, the Okinawa Computational Neuroscience Course will allow me to learn the skills of neural data analysis and modeling of neural activity, both essential to my research.

Mandana Ahmadi

Affiliation:	Gatsby Computational Neuroscience Unit, UCL
URL:	http://www.gatsby.ucl.ac.uk/~mandana
About:	I am interested in understanding adaptive goal-directed behaviour and its implementation at a neuronal level. In this course I hope to interact with and learn from leading experts in several sub-fields of neuroscience. This opportunity provides me with a complementary training to my current graduate studies, offering me a broader view for addressing the problems I am interested in solving. I am also looking forward to rich interactions with my fellow students.

Walther Akemann

Affiliation:

RIKEN Brain Science Institute

2-1 Hirosawa, Wako-City

Saitama 351-0198, JAPAN

URL:

http://www.brain.riken.jp/common/cv/w_akemann.pdf

About:

It is my pleasure to join OCNC 2010 as a tutor for a second time. This course provides a rare opportunity to advanced students and beginning researchers to explore new ideas away from the beaten track that you follow in our daily work, but yet may inspire new paths to follow in the future. The course brings together a diverse set of people with a common interest for computational biology and neuroscience. My own background is in neuroscience, physics and electrochemistry. Presently, I am a staff scientist at the RIKEN Brain Science Institute in Tokyo (Japan) and my research there is about functional dissection of cortical circuits of the mouse sensorimotor system using optogenetic probes. The brain, on many levels of its organization, performs model simulations to predict future events using internal representations of the physical world and multiple sensory feedbacks. Successful models suggest a gain of competence and understanding in dealing with the world. I am

looking forward to meet you at OCNC 2010 and wish all participants a worthwhile learning experience while creating new models, trying new computational strategies and, on the way, enjoying the many emotional leaps that go along with deeper understanding.