WS 2005/06

Computational Neuroscience: Oberseminar (31170)

Dr. Laurenz Wiskott, Prof. Andreas V. M. Herz, and Dr. Richard Kempter

Oberseminar: Monday 14-15 Uhr in the ITB seminar room (Invalidenstraße 43)

In this seminar various topics of current research in computational neuroscience are presented. Attendees should have basic knowledge in neuroscience and computational neuroscience, e.g. from the courses "Computational Neuroscience I-IV".

Talks

04.10.2005
Tuesday
11:00! Inés Samengo (Centro Atómico Bariloche, Argentina)
"Information transmission in burst spiking"

17.10.2005 Paula Kuokkanen
"Modeling Hearing Loss with the Auditory Modeling System"

24.10.2005 Norbert Krüger (Media Lab, Aalborg University Copenhagen)
"Multi-modal Representations for Visual Scene Analysis"

31.10.2005 Tiziano Zito
"The Impact of Intellectual Property on Science"

07.11.2005 Robert Schmidt
"Testing predictions of a model for phase precession in the hippocampus"

14.11.2005 Markus Diesmann (Computational Neurophysics, Albert-Ludwigs-University Freiburg)
"Spike synchronization by fast input transients"

Abstract: Synchronization in feed-forward neuronal subnetworks in the brain has been proposed to explain precisely timed spike patterns observed in experiments. While the attractor dynamics of the network is now well understood, the single neuron mechanisms remained unexplained. Previous attempts have captured the effects of the highly fluctuating membrane potential by relating spike intensity f(U) to the instantaneous voltage U of the input. We demonstrate that spike probability is high during the rise and low during the decay of U(t) exhibiting the d/dt U-dependence of f, not refractoriness, as the essential mechanism of synchronization. Moreover, The bifurcation scenario is quantitatively described by a simple f(U, d/dt U) relationship. The findings suggest f(U, d/dt U) as the relevant model class for the investigation of neural synchronization phenomena in a noisy environment.

21.11.2005 Robert Gütig
"Supervised learning of spike-timing based decision rules."

28.11.2005 Paul Szyszka (Institut für Biologie, Freie Universität Berlin)
"Sparsening and temporal sharpening of olfactory representations in the honeybee mushroom bodies."

Abstract: We characterized odor-evoked network activity in the honeybee brain at three consecutive neural compartments. Using Ca2+ imaging, we recorded activity in the dendrites of the projection neurons that connect the antennal lobe with the mushroom body, a higher-order integration center. Next, we recorded the presynaptic terminals of these projection neurons. Finally, we characterized their postsynaptic partners, the intrinsic neurons of the mushroom body, the Kenyon cells. We found fundamental differences in odor coding between the antennal lobe and the mushroom body. Odors evoked combinatorial activity patterns at all three processing stages, but the spatial patterns became progressively sparser along this path. Projection neuron dendrites and boutons showed similar response profiles, but the boutons were more narrowly tuned to odors. The transmission from projection neuron boutons to Kenyon cells was accompanied by a further sparsening of the population code. Activated Kenyon cells were highly odor specific. Furthermore, Kenyon cells responded to projection neuron activity only within the first 200 ms and transformed complex temporal patterns into brief phasic responses. Thus, two types of transformations occurred within the MB: Sparsening of a combinatorial code, mediated by pre- and postsynaptic processing within the mushroom body microcircuits, and temporal sharpening of postsynaptic Kenyon cell responses, probably involving a broader loop of inhibitory recurrent neurons.

05.12.2005 Andreas Herz
"About the Web-Pages of the ITB/Neuro Group"

12.12.2005 Peter Appleby
"A synaptic and temporal ensemble interpretation of spike-timing-dependent plasticity"

09.01.2006 Marton Danoczy
"Efficient estimation of hidden state dynamics from spike trains"

16.01.2006 Martin Stemmler

23.01.2006 Roland Schaette

30.01.2006 Henning Sprekeler
"SFA and Hamilton's principle of least action"

06.02.2006 Michael Bendels

13.02.2006 Tim Oppermann

27.02.2006 Felix Creutzig
"The past-future information bottleneck of dynamical systems"

Abstract: Predictive information was recently proposed as a general measure of complexity (Bialek et al, 2001). Additionally, the information bottleneck (IB) method (Tishby et al., 1999) extracts the relevant aspects of data soft clustering one variable while preserving information about another variable. Here, we combine those ideas and ask in the framework of linear dynamic systems, how the relevant information of the input past about the output future can be extracted. We formalize the intuition that the state space of dynamical systems acts as a bottleneck and transmits the relevant information of the input past about the output future. We state the Lagrangian for dynamical systems and formulate the relevant mutual informations in terms of the covariances which are explicitly dependent on the state space. That allows us to unravel the input history in the IB sense in terms of structural phase transitions corresponding to additional dimensions of the state space. The critical values are function of the Hankel singular values, recovering balanced-model reduction. Hence, complexity of observed data can be reduced to the underlying system's complexity (structural complexity).

06.03.2006 Samuel Glauser

13.03.2006 Christian Leibold
"Impressions from the cosyne meeting"

from now on at 14:00 o'clock

20.03.2006 Benjamin Lindner (Max-Planck-Institut für Physik komplexer Systeme, Dresden)
"Is the superposition of many random spike trains a Poisson process?"

Abstract: We study the sum of many independent spike trains and ask whether the resulting spike train has Poisson statistics or not. It is shown that for a non-Poissonian statistics of the single spike train, the resulting sum of spikes has exponential interspike interval (ISI) distribution, vanishing ISI correlation at finite lag but exhibits exactly the same power spectrum as the original spike train does. This paradox is resolved by considering what happens to ISI correlations in the limit of infinite number of superposed trains. Implications of our findings for stochastic models in the neurosciences are briefly discussed.
Ref. Phys. Rev. E 73, 022901 (2006)

27.03.2006 Raphael Ritz
"On software publishing"

Created by Laurenz Wiskott, http://itb.biologie.hu-berlin.de/~wiskott/

04.10.2005 Tuesday 11:00!		Inés Samengo (Centro Atómico Bariloche, Argentina) "Information transmission in burst spiking"
17.10.2005		Paula Kuokkanen "Modeling Hearing Loss with the Auditory Modeling System"
24.10.2005		Norbert Krüger (Media Lab, Aalborg University Copenhagen) "Multi-modal Representations for Visual Scene Analysis"
31.10.2005		Tiziano Zito "The Impact of Intellectual Property on Science"
07.11.2005		Robert Schmidt "Testing predictions of a model for phase precession in the hippocampus"
14.11.2005		Markus Diesmann (Computational Neurophysics, Albert-Ludwigs-University Freiburg) "Spike synchronization by fast input transients" Abstract: Synchronization in feed-forward neuronal subnetworks in the brain has been proposed to explain precisely timed spike patterns observed in experiments. While the attractor dynamics of the network is now well understood, the single neuron mechanisms remained unexplained. Previous attempts have captured the effects of the highly fluctuating membrane potential by relating spike intensity f(U) to the instantaneous voltage U of the input. We demonstrate that spike probability is high during the rise and low during the decay of U(t) exhibiting the d/dt U-dependence of f, not refractoriness, as the essential mechanism of synchronization. Moreover, The bifurcation scenario is quantitatively described by a simple f(U, d/dt U) relationship. The findings suggest f(U, d/dt U) as the relevant model class for the investigation of neural synchronization phenomena in a noisy environment.
21.11.2005		Robert Gütig "Supervised learning of spike-timing based decision rules."
28.11.2005		Paul Szyszka (Institut für Biologie, Freie Universität Berlin) "Sparsening and temporal sharpening of olfactory representations in the honeybee mushroom bodies." Abstract: We characterized odor-evoked network activity in the honeybee brain at three consecutive neural compartments. Using Ca2+ imaging, we recorded activity in the dendrites of the projection neurons that connect the antennal lobe with the mushroom body, a higher-order integration center. Next, we recorded the presynaptic terminals of these projection neurons. Finally, we characterized their postsynaptic partners, the intrinsic neurons of the mushroom body, the Kenyon cells. We found fundamental differences in odor coding between the antennal lobe and the mushroom body. Odors evoked combinatorial activity patterns at all three processing stages, but the spatial patterns became progressively sparser along this path. Projection neuron dendrites and boutons showed similar response profiles, but the boutons were more narrowly tuned to odors. The transmission from projection neuron boutons to Kenyon cells was accompanied by a further sparsening of the population code. Activated Kenyon cells were highly odor specific. Furthermore, Kenyon cells responded to projection neuron activity only within the first 200 ms and transformed complex temporal patterns into brief phasic responses. Thus, two types of transformations occurred within the MB: Sparsening of a combinatorial code, mediated by pre- and postsynaptic processing within the mushroom body microcircuits, and temporal sharpening of postsynaptic Kenyon cell responses, probably involving a broader loop of inhibitory recurrent neurons.
05.12.2005		Andreas Herz "About the Web-Pages of the ITB/Neuro Group"
12.12.2005		Peter Appleby "A synaptic and temporal ensemble interpretation of spike-timing-dependent plasticity"
09.01.2006		Marton Danoczy "Efficient estimation of hidden state dynamics from spike trains"
16.01.2006		Martin Stemmler
23.01.2006		Roland Schaette
30.01.2006		Henning Sprekeler "SFA and Hamilton's principle of least action"
06.02.2006		Michael Bendels
13.02.2006		Tim Oppermann
27.02.2006		Felix Creutzig "The past-future information bottleneck of dynamical systems" Abstract: Predictive information was recently proposed as a general measure of complexity (Bialek et al, 2001). Additionally, the information bottleneck (IB) method (Tishby et al., 1999) extracts the relevant aspects of data soft clustering one variable while preserving information about another variable. Here, we combine those ideas and ask in the framework of linear dynamic systems, how the relevant information of the input past about the output future can be extracted. We formalize the intuition that the state space of dynamical systems acts as a bottleneck and transmits the relevant information of the input past about the output future. We state the Lagrangian for dynamical systems and formulate the relevant mutual informations in terms of the covariances which are explicitly dependent on the state space. That allows us to unravel the input history in the IB sense in terms of structural phase transitions corresponding to additional dimensions of the state space. The critical values are function of the Hankel singular values, recovering balanced-model reduction. Hence, complexity of observed data can be reduced to the underlying system's complexity (structural complexity).
06.03.2006		Samuel Glauser
13.03.2006		Christian Leibold "Impressions from the cosyne meeting"
		from now on at 14:00 o'clock
20.03.2006		Benjamin Lindner (Max-Planck-Institut für Physik komplexer Systeme, Dresden) "Is the superposition of many random spike trains a Poisson process?" Abstract: We study the sum of many independent spike trains and ask whether the resulting spike train has Poisson statistics or not. It is shown that for a non-Poissonian statistics of the single spike train, the resulting sum of spikes has exponential interspike interval (ISI) distribution, vanishing ISI correlation at finite lag but exhibits exactly the same power spectrum as the original spike train does. This paradox is resolved by considering what happens to ISI correlations in the limit of infinite number of superposed trains. Implications of our findings for stochastic models in the neurosciences are briefly discussed. Ref. Phys. Rev. E 73, 022901 (2006)
27.03.2006		Raphael Ritz "On software publishing"