Belle Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Belle Liu has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Chung-Chuan Lo
National Tsing Hua University
2 shared publications - Kuo-An Wu
Department of Physics, National Tsing Hua University
2 shared publications Alexander James White
Institute of Systems Neuroscience, National Tsing Hua University, Hsinchu, Taiwan
1 shared publication- Alexander White
Tsinghua University - School of Economics & Management
1 shared publication - Arthur Hong
Department of Neurobiology and Biophysics, University of Washington, Seattle, USA
1 shared publication Chung-Chuan Lo
Institute of Systems Neuroscience, National Tsing Hua University
1 shared publication
Latest Publications
- Open Access
Hyper Flexible Neural Networks Rapidly Switch between Logic Operations in a Compact Four Neuron Circuit
Biological neural circuits at various levels exhibit rapid adaptability to diverse environmental stimuli. Such fast response times imply that adaptation cannot rely solely on synaptic plasticity, which operates on a much slower timescale. Instead, circuits must be inherently hyper-flexible and receptive to switches in functionalities without changes in network structure. This biological flexibility is a fruitful mechanism for constructing artificial reconfigurable circuits, whether they are spiking or non-spiking. In this study, we demonstrate that a 4-neuron circuit can rapidly and controllably switch between 24 unique logical functions while maintaining the same set of synaptic weights. Moreover, we show that this reconfigurability works for several different underlying neuronal architectures and strikingly can be applied to a network composed of any sigmoid-shaped activation function. We conclude with proof-of-concept applications showing that we can perform standard tasks such as a full-adder, as well as event-based conditional computing, such as detecting unexpected motion.
Open AccessHyper Flexible Neural Networks Rapidly Switch between Logic Operations in a Compact Four Neuron Circuit
Biological neural circuits at various levels exhibit rapid adaptability to diverse environmental stimuli. Such fast response times imply that adaptation cannot rely solely on synaptic plasticity, which operates on a much slower timescale. Instead, circuits must be inherently hyper-flexible and receptive to switches in functionalities without changes in network structure. This biological flexibility is a fruitful mechanism for constructing artificial reconfigurable circuits, whether they are spiking or non-spiking. In this study, we demonstrate that a 4-neuron circuit can rapidly and controllably switch between 24 unique logical functions while maintaining the same set of synaptic weights. Moreover, we show that this reconfigurability works for several different underlying neuronal architectures and strikingly can be applied to a network composed of any sigmoid-shaped activation function. We conclude with proof-of-concept applications showing that we can perform standard tasks such as a full-adder, as well as event-based conditional computing, such as detecting unexpected motion.
- Open Access
Choose carefully, act quickly: Efficient decision making with selective inhibition in attractor neural networks
The ability to decide swiftly and accurately in an urgent scenario is crucial for an organism’s survival. The neural mechanisms underlying the perceptual decision and trade-off between speed and accuracy have been extensively studied in the past few decades. Among several theoretical models, the attractor neural network model has successfully captured both behavioral and neuronal data observed in many decision experiments. However, a recent experimental study revealed additional details that were not considered in the original attractor model. In particular, the study shows that the inhibitory neurons in the posterior parietal cortex of mice are as selective to decision making results as the excitatory neurons, whereas the previous attractor model assumes the inhibitory neurons to be unselective. In this study, we investigate the attractor model with selective inhibition (selective model), which can be considered as a general case of the previous attractor model (unselective model). Our model reproduces several behavioral and neurophysiological observations. To analyze the dynamics of the selective model, we reduce it and show that selectivity adds a time-varying component to the energy landscape. This time dependence of the energy landscape allows the selective model to integrate information carefully in initial stages, then quickly converge to an attractor once the choice is clear. This results in the selective model having a more efficient speed-accuracy trade-off that is unreachable by unselective models.
Open AccessChoose carefully, act quickly: Efficient decision making with selective inhibition in attractor neural networks
The ability to decide swiftly and accurately in an urgent scenario is crucial for an organism’s survival. The neural mechanisms underlying the perceptual decision and trade-off between speed and accuracy have been extensively studied in the past few decades. Among several theoretical models, the attractor neural network model has successfully captured both behavioral and neuronal data observed in many decision experiments. However, a recent experimental study revealed additional details that were not considered in the original attractor model. In particular, the study shows that the inhibitory neurons in the posterior parietal cortex of mice are as selective to decision making results as the excitatory neurons, whereas the previous attractor model assumes the inhibitory neurons to be unselective. In this study, we investigate the attractor model with selective inhibition (selective model), which can be considered as a general case of the previous attractor model (unselective model). Our model reproduces several behavioral and neurophysiological observations. To analyze the dynamics of the selective model, we reduce it and show that selectivity adds a time-varying component to the energy landscape. This time dependence of the energy landscape allows the selective model to integrate information carefully in initial stages, then quickly converge to an attractor once the choice is clear. This results in the selective model having a more efficient speed-accuracy trade-off that is unreachable by unselective models.
- 2 August 2021
- Published by Springer Nature in Nature Neuroscience
- Vol. 24 (9) , 1280-1291
Predictive encoding of motion begins in the primate retina
Predictive motion encoding is an important aspect of visually guided behavior that allows animals to estimate the trajectory of moving objects. Motion prediction is understood primarily in the context of translational motion, but the environment contains other types of behaviorally salient motion correlation such as those produced by approaching or receding objects. However, the neural mechanisms that detect and predictively encode these correlations remain unclear. We report here that four of the parallel output pathways in the primate retina encode predictive motion information, and this encoding occurs for several classes of spatiotemporal correlation that are found in natural vision. Such predictive coding can be explained by known nonlinear circuit mechanisms that produce a nearly optimal encoding, with transmitted information approaching the theoretical limit imposed by the stimulus itself. Thus, these neural circuit mechanisms efficiently separate predictive information from nonpredictive information during the encoding process.
- 2 August 2021
- Published by Springer Nature in Nature Neuroscience
- Vol. 24 (9) , 1280-1291
Predictive encoding of motion begins in the primate retina
Predictive motion encoding is an important aspect of visually guided behavior that allows animals to estimate the trajectory of moving objects. Motion prediction is understood primarily in the context of translational motion, but the environment contains other types of behaviorally salient motion correlation such as those produced by approaching or receding objects. However, the neural mechanisms that detect and predictively encode these correlations remain unclear. We report here that four of the parallel output pathways in the primate retina encode predictive motion information, and this encoding occurs for several classes of spatiotemporal correlation that are found in natural vision. Such predictive coding can be explained by known nonlinear circuit mechanisms that produce a nearly optimal encoding, with transmitted information approaching the theoretical limit imposed by the stimulus itself. Thus, these neural circuit mechanisms efficiently separate predictive information from nonpredictive information during the encoding process.
- Open Access
Augmenting Flexibility: Mutual Inhibition Between Inhibitory Neurons Expands Functional Diversity
Rapid, flexible response to an ever-changing environment is critical for an organism’s survival. Recently, multicellular recordings have shown that this rapid, flexible switching between activity patterns is present in neural microcircuits. However, the underlying neural mechanism is not clear. Strikingly, we show in a neural circuit model that mutually inhibitory connections are crucial for rapid and flexible switching between distinct functions without synaptic plasticity. Here, we develop a theoretical framework to explain how inhibitory recurrent circuits give rise to this flexibility and show that mutual inhibition doubles the number of cusp bifurcations in small neural circuits. As a concrete example, we study a commonly observed class of functional motifs we call Coupled Recurrent Inhibitory and Recurrent Excitatory Loops (CRIRELs). These CRIRELs have the advantage of being both multifunctional and controllable, performing a plethora of unique functions. Finally, we demonstrate how mutual inhibition maximizes storage capacity for larger networks.
Open AccessAugmenting Flexibility: Mutual Inhibition Between Inhibitory Neurons Expands Functional Diversity
Rapid, flexible response to an ever-changing environment is critical for an organism’s survival. Recently, multicellular recordings have shown that this rapid, flexible switching between activity patterns is present in neural microcircuits. However, the underlying neural mechanism is not clear. Strikingly, we show in a neural circuit model that mutually inhibitory connections are crucial for rapid and flexible switching between distinct functions without synaptic plasticity. Here, we develop a theoretical framework to explain how inhibitory recurrent circuits give rise to this flexibility and show that mutual inhibition doubles the number of cusp bifurcations in small neural circuits. As a concrete example, we study a commonly observed class of functional motifs we call Coupled Recurrent Inhibitory and Recurrent Excitatory Loops (CRIRELs). These CRIRELs have the advantage of being both multifunctional and controllable, performing a plethora of unique functions. Finally, we demonstrate how mutual inhibition maximizes storage capacity for larger networks.