Shifting attention to dynamics: Self-reconfiguration of neural networks

Task-Adaptive Neuromorphic Computing Using Reconfigurable Organic Neuristors with Tunable Plasticity and Logic-in-Memory Operations

The brain's function can be dynamically reconfigured through a unified neuron-synapse architecture, enabling task-adaptive network-level topology for energy-efficient learning and inferencing. Here, we demonstrate an organic neuristor utilizing a ferroelectric-electrolyte dielectric interface. This neuristor enables tunable short- to long-term plasticity and reconfigurable logic-in-memory functions by controlling the interfacial interaction between electrolyte ions and ferroelectric dipoles. Notably, the short-term plasticity of the organic neuristor allows for power-efficient reservoir computing in edge-computing scenarios, exhibiting impressive recognition accuracy, including images (90.6%) and acoustic signals (97.7%). For high-performance computing tasks, the neuristor based on long-term plasticity and logic-in-memory operations can construct all of the hardware circuits of a binarized neural network (BNN) within a unified framework. The BNN demonstrates excellent noise tolerance, achieving high recognition accuracies of 99.2% and 86.4% on the MNIST and CIFAR-10 data sets, respectively. Consequently, our research sheds light on the development of power-efficient artificial intelligence systems.

Robustness and Flexibility of Neural Function through Dynamical Criticality

In theoretical biology, robustness refers to the ability of a biological system to function properly even under perturbation of basic parameters (e.g., temperature or pH), which in mathematical models is reflected in not needing to fine-tune basic parameter constants; flexibility refers to the ability of a system to switch functions or behaviors easily and effortlessly. While there are extensive explorations of the concept of robustness and what it requires mathematically, understanding flexibility has proven more elusive, as well as also elucidating the apparent opposition between what is required mathematically for models to implement either. In this paper we address a number of arguments in theoretical neuroscience showing that both robustness and flexibility can be attained by systems that poise themselves at the onset of a large number of dynamical bifurcations, or dynamical criticality, and how such poising can have a profound influence on integration of information processing and function. Finally, we examine critical map lattices, which are coupled map lattices where the coupling is dynamically critical in the sense of having purely imaginary eigenvalues. We show that these map lattices provide an explicit connection between dynamical criticality in the sense we have used and “edge of chaos” criticality.

Information diversity in individual auditory cortical neurons is associated with functionally distinct coordinated neuronal ensembles

Neuronal activity in auditory cortex is often highly synchronous between neighboring neurons. Such coordinated activity is thought to be crucial for information processing. We determined the functional properties of coordinated neuronal ensembles (cNEs) within primary auditory cortical (AI) columns relative to the contributing neurons. Nearly half of AI cNEs showed robust spectro-temporal receptive fields whereas the remaining cNEs showed little or no acoustic feature selectivity. cNEs can therefore capture either specific, time-locked information of spectro-temporal stimulus features or reflect stimulus-unspecific, less-time specific processing aspects. By contrast, we show that individual neurons can represent both of those aspects through membership in multiple cNEs with either high or absent feature selectivity. These associations produce functionally heterogeneous spikes identifiable by instantaneous association with different cNEs. This demonstrates that single neuron spike trains can sequentially convey multiple aspects that contribute to cortical processing, including stimulus-specific and unspecific information.

Fish as a model in social neuroscience: conservation and diversity in the social brain network

Mechanisms for fish social behaviours involve a social brain network (SBN) which is evolutionarily conserved among vertebrates. However, considerable diversity is observed in the actual behaviour patterns amongst nearly 30000 fish species. The huge variation found in socio-sexual behaviours and strategies is likely generated by a morphologically and genetically well-conserved small forebrain system. Hence, teleost fish provide a useful model to study the fundamental mechanisms underlying social brain functions. Herein we review the foundations underlying fish social behaviours including sensory, hormonal, molecular and neuroanatomical features. Gonadotropin-releasing hormone neurons clearly play important roles, but the participation of vasotocin and isotocin is also highlighted. Genetic investigations of developing fish brain have revealed the molecular complexity of neural development of the SBN. In addition to straightforward social behaviours such as sex and aggression, new experiments have revealed higher order and unique phenomena such as social eavesdropping and social buffering in fish. Finally, observations interpreted as 'collective cognition' in fish can likely be explained by careful observation of sensory determinants and analyses using the dynamics of quantitative scaling. Understanding of the functions of the SBN in fish provide clues for understanding the origin and evolution of higher social functions in vertebrates.

Complex spatiotemporal behavior and coherent excitations in critically-coupled chains of neural circuits

We investigate a critically-coupled chain of nonlinear oscillators, whose dynamics displays complex spatiotemporal patterns of activity, including regimes in which glider-like coherent excitations move about and interact. The units in the network are identical simple neural circuits whose dynamics is given by the Wilson-Cowan model and are arranged in space along a one-dimensional lattice with nearest neighbor interactions. The interactions follow an alternating sign rule, and hence the "synaptic matrix" M embodying them is tridiagonal antisymmetric and has purely imaginary (critical) eigenvalues. The model illustrates the interplay of two properties: circuits with a complex internal dynamics, such as multiple stable periodic solutions and period doubling bifurcations, and coupling with a "critical" synaptic matrix, i.e., having purely imaginary eigenvalues. In order to identify the dynamical underpinnings of these behaviors, we explored a discrete-time coupled-map lattice inspired by our system: the dynamics of the units is dictated by a chaotic map of the interval, and the interactions are given by allowing the critical coupling to act for a finite period τ , thus given by a unitary matrix U = exp ⁡ ( τ 2 M ) . It is now explicit that such critical couplings are volume-preserving in the sense of Liouville's theorem. We show that this map is also capable of producing a variety of complex spatiotemporal patterns including gliders, like our original chain of neural circuits. Our results suggest that if the units in isolation are capable of featuring multiple dynamical states, then local critical couplings lead to a wide variety of emergent spatiotemporal phenomena.