|2021||Feb Mar Apr May Jun Jul Aug|
|2020||Jan May Jun Jul Aug Sep Oct Nov Dec|
|2019||Jan Feb Mar Apr May Jun Jul Aug Oct Nov|
|2018||Jan Feb Mar Apr May Jun Jul Aug Oct Nov Dec|
|2017||Jan Feb Mar Apr May Jun Jul Aug Oct Nov Dec|
Event(s) on November 2020
- Friday, 20th November, 2020
Title: From Sparse Irregular Spikes to Critical Avalanches: Cost-Efficient Neural Dynamics Speaker: Prof Changsong Zhou, Department of Physics, Hong Kong Baptist University, Hong Kong Time/Place: 16:00 - 17:00
Zoom, (ID: 939 0417 7353 and Passcode: 594986 )
Abstract: The brain is highly energy consuming, therefore is under strong selective pressure to achieve cost-efficiency in both cortical connectivity and activity. Cortical neural circuits display highly irregular spiking in individual neurons but variably sized collective firing, oscillations and critical avalanches at the population level, all of which have functional importance for information processing. However, cost-efficiency as a design principle for cortical activities has been rarely studied. Especially it is not clear how cost-efficiency is related to ubiquitously observed multi-scale properties of irregular firing, oscillations and neuronal avalanches. In this talk, I review key features of the brain as complex dynamical network systems. Then I will give a brief introduction of our work demonstrating that prominent multilevel neural dynamics properties can be simultaneously observed in a generic, biologically plausible neural circuit model that captures excitation-inhibition balance and realistic dynamics of synaptic conductance. Their co-emergence achieves minimal energy cost as well as maximal energy efficiency on information capacity, when neuronal firing is coordinated and shaped by moderate synchrony to reduce otherwise redundant spikes, and the dynamical clustering is maintained in the form of neuronal avalanches. We propose a semi-analytical mean-field theory to derive the field equations governing the network macroscopic dynamics. It reveals that the E-I balanced state of the network manifesting irregular individual spiking is characterized by a macroscopic stable state, which can be either a fixed point or a periodic motion and the transition is predicted by a Hopf bifurcation in the macroscopic field. An analysis of the impact of network topology from random to modular networks shows that local dense connectivity under E-I balanced dynamics appears to be the key "less-is-more" solutions to achieve cost-efficiency organization in neural systems.
Dongping Yang, Haijun Zhou and Changsong Zhou, Co-emergence of Multi-scale Cortical Activities of Irregular firing, Oscillations and Avalanches Achieves Cost-efficient Information Capacity, PLoS Computational Biology 13, e1005384 (2017).
Junhao Liang, Tianshou Zhou and Changsong Zhou, Hopf Bifurcation in Mean Field Explains Critical Avalanches in E-I Balanced Neuronal Networks: A Mechanism for Multiscale Variability, Frontiers in Systems Neuroscience (in press).
Shenjun Wang, Junhao Liang and Changsong Zhou: Less is More: Wiring-Economical Modular Networks Support Self-Sustained Firing-Economical Neural Avalanches for Efficient Processing, submitted. arXiv: 2007.02511 (2020).
- Wednesday, 25th November, 2020
Title: Turnpike control and deep learning Speaker: Prof Enrique Zuazua, FAU Erlangen-Nürnberg, Germany Time/Place: 16:00 - 17:00
Zoom, (Meeting ID: 947 0888 1860)
Abstract: The turnpike principle asserts that in long time horizons optimal control strategies are nearly of a steady state nature. In this lecture we shall survey on some recent results on this topic and present some its consequences on deep supervised learning. This lecture will be based in particular on recent joint work with C. Esteve, B. Geshkovski and D. Pighin.
- Thursday, 26th November, 2020
Title: Euclidean Distance Matrix Optimization: Models, Algorithms and Applications Speaker: Prof Hou-Duo Qi, CORMSIS (Center of Operational Research, Management Science and Information Systems), University of Southampton, UK Time/Place: 16:00 - 17:00
Zoom, Meeting ID: 940 9975 0728
Abstract: Euclidean Distance Matrix (EDM) optimization has become a robust approach for analyzing dissimilarity data due to its capacity of handling hard constraints, availability of fast algorithms and guaranteed error bounds. It has found applications in machine learning (dimensionality reduction), engineering (sensor network localization), social sciences (multidimensional scaling), and computational chemistry (molecular conformation). This talk will give a brief review of the approach, showcasing its mathematical theory, algorithmic development and its performance on some challenging examples. We will also discuss an emerging and exciting application in computational finance on portfolio selections. We end the talk with an open question on software implementation.