Brian S. Robinson
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
Brian S. Robinson has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
Danilo Symonette
The Johns Hopkins University Applied Physics Laboratory, Laurel, USA
3 shared publications- Dong Song
University of Southern California
3 shared publications Erik C. Johnson
The Johns Hopkins University Applied Physics Laboratory, Laurel, USA
3 shared publicationsJustin Joyce
The Johns Hopkins University Applied Physics Laboratory, Laurel, USA
3 shared publicationsMartha Cervantes
The Johns Hopkins University Applied Physics Laboratory, Laurel, USA
3 shared publications- Raphael Norman-Tenazas
Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland, UNITED STATES
3 shared publications
Latest Publications
- Open Access
- 28 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Enabling local learning for generative-replay-based continual learning with a recurrent model of the insect memory center
Continual learning without catastrophic forgetting of previous experiences is an open general challenge for artificial neural networks, but is especially under-explored for artificial neural networks suitable to implement on neuromorphic platforms. An algorithmic understanding of how continual learning occurs in biological neural networks can inform solutions for artificial neural networks, especially in neuromorphic platforms whose biomimetic computing architectures lend themselves to more biofidelic algorithms. In this work, we derive an approach for generative-replay-based continual learning with a three-factor local learning rule based on recurrent connectivity in the insect's memory center, and characterize the model with a CIFAR-100 class incremental continual learning task. First, we investigate the properties of the model's internal representations and find that the high dimensional sparse representations enable this form of generative replay, and that these representations can be binary as required on spiking neuromorphic platforms with little detriment to model performance. Next, we derive a three-factor local learning rule by introducing simplifying assumptions to network updates from error backpropagation optimization which makes the learning rule biologically plausible (i.e., without weight transport) and amenable to neuromorphic implementation. Finally, we find that these simplifications enhance performance during gradient-based optimization for continual learning and when implemented locally achieve this increased performance. Overall, the outcomes of this work can inform a more detailed understanding for continual learning in this biological circuit, as well as introduce general approaches for neuromorphic continual learning.
- 28 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessEnabling local learning for generative-replay-based continual learning with a recurrent model of the insect memory center
Continual learning without catastrophic forgetting of previous experiences is an open general challenge for artificial neural networks, but is especially under-explored for artificial neural networks suitable to implement on neuromorphic platforms. An algorithmic understanding of how continual learning occurs in biological neural networks can inform solutions for artificial neural networks, especially in neuromorphic platforms whose biomimetic computing architectures lend themselves to more biofidelic algorithms. In this work, we derive an approach for generative-replay-based continual learning with a three-factor local learning rule based on recurrent connectivity in the insect's memory center, and characterize the model with a CIFAR-100 class incremental continual learning task. First, we investigate the properties of the model's internal representations and find that the high dimensional sparse representations enable this form of generative replay, and that these representations can be binary as required on spiking neuromorphic platforms with little detriment to model performance. Next, we derive a three-factor local learning rule by introducing simplifying assumptions to network updates from error backpropagation optimization which makes the learning rule biologically plausible (i.e., without weight transport) and amenable to neuromorphic implementation. Finally, we find that these simplifications enhance performance during gradient-based optimization for continual learning and when implemented locally achieve this increased performance. Overall, the outcomes of this work can inform a more detailed understanding for continual learning in this biological circuit, as well as introduce general approaches for neuromorphic continual learning.
- Open Access
Informing generative replay for continual learning with long-term memory formation in the fruit fly
Continual learning without catastrophic forgetting is a challenge for artificial systems but it is done naturally across a range of biological systems, including in insects. A recurrent circuit has been identified in the fruit fly mushroom body to consolidate long term memories (LTM), but there is not currently an algorithmic understanding of this LTM formation. We hypothesize that generative replay is occurring to consolidate memories in this recurrent circuit, and find anatomical evidence in synapse-level connectivity that supports this hypothesis. Next, we introduce a computational model which combines an initial experience phase and LTM phase to perform generative replay based continual learning. When evaluated on a CIFAR-100 class-incremental continual learning task, the modeled LTM phase increases classification performance by 20% and approaches within 2% of the performance for a non-incremental upper bound baseline. Unique elements of the proposed generative replay model include: 1) coupling high dimensional sparse activation patterns with generative replay and 2) sampling and reconstructing higher level representations for training generative replay (as opposed to reconstructing sensory-level or processed sensory-level representations). Additionally, we make the experimentally testable prediction that a specific set of synapses would need to undergo experience-dependent plasticity during LTM formation to support our generative replay based model.
Open AccessInforming generative replay for continual learning with long-term memory formation in the fruit fly
Continual learning without catastrophic forgetting is a challenge for artificial systems but it is done naturally across a range of biological systems, including in insects. A recurrent circuit has been identified in the fruit fly mushroom body to consolidate long term memories (LTM), but there is not currently an algorithmic understanding of this LTM formation. We hypothesize that generative replay is occurring to consolidate memories in this recurrent circuit, and find anatomical evidence in synapse-level connectivity that supports this hypothesis. Next, we introduce a computational model which combines an initial experience phase and LTM phase to perform generative replay based continual learning. When evaluated on a CIFAR-100 class-incremental continual learning task, the modeled LTM phase increases classification performance by 20% and approaches within 2% of the performance for a non-incremental upper bound baseline. Unique elements of the proposed generative replay model include: 1) coupling high dimensional sparse activation patterns with generative replay and 2) sampling and reconstructing higher level representations for training generative replay (as opposed to reconstructing sensory-level or processed sensory-level representations). Additionally, we make the experimentally testable prediction that a specific set of synapses would need to undergo experience-dependent plasticity during LTM formation to support our generative replay based model.
- Open Access
- 18 March 2022
- Published by Springer Nature in Scientific Reports
- Vol. 12 (1) , 1
Author Correction: Online learning for orientation estimation during translation in an insect ring attractor network
- 18 March 2022
- Published by Springer Nature in Scientific Reports
- Vol. 12 (1) , 1
Open AccessAuthor Correction: Online learning for orientation estimation during translation in an insect ring attractor network
- Open Access
- 25 February 2022
- Published by Springer Nature in Scientific Reports
- Vol. 12 (1) , 1-15
Online learning for orientation estimation during translation in an insect ring attractor network
Insect neural systems are a promising source of inspiration for new navigation algorithms, especially on low size, weight, and power platforms. There have been unprecedented recent neuroscience breakthroughs withDrosophilain behavioral and neural imaging experiments as well as the mapping of detailed connectivity of neural structures. General mechanisms for learning orientation in the central complex (CX) ofDrosophilahave been investigated previously; however, it is unclear how these underlying mechanisms extend to cases where there is translation through an environment (beyond only rotation), which is critical for navigation in robotic systems. Here, we develop a CX neural connectivity-constrained model that performs sensor fusion, as well as unsupervised learning of visual features for path integration; we demonstrate the viability of this circuit for use in robotic systems in simulated and physical environments. Furthermore, we propose a theoretical understanding of how distributed online unsupervised network weight modification can be leveraged for learning in a trajectory through an environment by minimizing orientation estimation error. Overall, our results may enable a new class of CX-derived low power robotic navigation algorithms and lead to testable predictions to inform future neuroscience experiments.
- 25 February 2022
- Published by Springer Nature in Scientific Reports
- Vol. 12 (1) , 1-15
Open AccessOnline learning for orientation estimation during translation in an insect ring attractor network
Insect neural systems are a promising source of inspiration for new navigation algorithms, especially on low size, weight, and power platforms. There have been unprecedented recent neuroscience breakthroughs withDrosophilain behavioral and neural imaging experiments as well as the mapping of detailed connectivity of neural structures. General mechanisms for learning orientation in the central complex (CX) ofDrosophilahave been investigated previously; however, it is unclear how these underlying mechanisms extend to cases where there is translation through an environment (beyond only rotation), which is critical for navigation in robotic systems. Here, we develop a CX neural connectivity-constrained model that performs sensor fusion, as well as unsupervised learning of visual features for path integration; we demonstrate the viability of this circuit for use in robotic systems in simulated and physical environments. Furthermore, we propose a theoretical understanding of how distributed online unsupervised network weight modification can be leveraged for learning in a trajectory through an environment by minimizing orientation estimation error. Overall, our results may enable a new class of CX-derived low power robotic navigation algorithms and lead to testable predictions to inform future neuroscience experiments.
- Open Access
Online learning for orientation estimation during translation in an insect ring attractor network
Insect neural systems are a promising source of inspiration for new algorithms for navigation, especially on low size, weight, and power platforms. There have been unprecedented recent neuroscience breakthroughs with Drosophila in behavioral and neural imaging experiments as well as the mapping of detailed connectivity of neural structures. General mechanisms for learning orientation in the central complex (CX) of Drosophila have been investigated previously; however, it is unclear how these underlying mechanisms extend to cases where there is translation through an environment (beyond only rotation), which is critical for navigation in robotic systems. Here, we develop a CX neural connectivity-constrained model that performs sensor fusion, as well as unsupervised learning of visual features for path integration; we demonstrate the viability of this circuit for use in robotic systems in simulated and physical environments. Furthermore, we propose a theoretical understanding of how distributed online unsupervised network weight modification can be leveraged for learning in a trajectory through an environment by minimizing of orientation estimation error. Overall, our results here may enable a new class of CX-derived low power robotic navigation algorithms and lead to testable predictions to inform future neuroscience experiments. Summary: An insect neural connectivity-constrained model performs sensor fusion and online learning for orientation estimation.
Open AccessOnline learning for orientation estimation during translation in an insect ring attractor network
Insect neural systems are a promising source of inspiration for new algorithms for navigation, especially on low size, weight, and power platforms. There have been unprecedented recent neuroscience breakthroughs with Drosophila in behavioral and neural imaging experiments as well as the mapping of detailed connectivity of neural structures. General mechanisms for learning orientation in the central complex (CX) of Drosophila have been investigated previously; however, it is unclear how these underlying mechanisms extend to cases where there is translation through an environment (beyond only rotation), which is critical for navigation in robotic systems. Here, we develop a CX neural connectivity-constrained model that performs sensor fusion, as well as unsupervised learning of visual features for path integration; we demonstrate the viability of this circuit for use in robotic systems in simulated and physical environments. Furthermore, we propose a theoretical understanding of how distributed online unsupervised network weight modification can be leveraged for learning in a trajectory through an environment by minimizing of orientation estimation error. Overall, our results here may enable a new class of CX-derived low power robotic navigation algorithms and lead to testable predictions to inform future neuroscience experiments. Summary: An insect neural connectivity-constrained model performs sensor fusion and online learning for orientation estimation.
- 30 August 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 26 (2) , 272-280
Sparse Large-Scale Nonlinear Dynamical Modeling of Human Hippocampus for Memory Prostheses
In order to build hippocampal prostheses for restoring memory functions, we build sparse multi-input, multi-output (MIMO) nonlinear dynamical models of the human hippocampus. Spike trains are recorded from hippocampal CA3 and CA1 regions of epileptic patients performing a variety of memory-dependent delayed match-to-sample (DMS) tasks. Using CA3 and CA1 spike trains as inputs and outputs respectively, sparse generalized Laguerre-Volterra models are estimated with group lasso and local coordinate descent methods to capture the nonlinear dynamics underlying the CA3-CA1 spike train transformations. These models can accurately predict the CA1 spike trains based on the ongoing CA3 spike trains during multiple memory events, e.g., sample presentation, sample response, match presentation and match response, of the DMS task, and thus will serve as the computational basis of human hippocampal memory prostheses.
- 30 August 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 26 (2) , 272-280
Sparse Large-Scale Nonlinear Dynamical Modeling of Human Hippocampus for Memory Prostheses
In order to build hippocampal prostheses for restoring memory functions, we build sparse multi-input, multi-output (MIMO) nonlinear dynamical models of the human hippocampus. Spike trains are recorded from hippocampal CA3 and CA1 regions of epileptic patients performing a variety of memory-dependent delayed match-to-sample (DMS) tasks. Using CA3 and CA1 spike trains as inputs and outputs respectively, sparse generalized Laguerre-Volterra models are estimated with group lasso and local coordinate descent methods to capture the nonlinear dynamics underlying the CA3-CA1 spike train transformations. These models can accurately predict the CA1 spike trains based on the ongoing CA3 spike trains during multiple memory events, e.g., sample presentation, sample response, match presentation and match response, of the DMS task, and thus will serve as the computational basis of human hippocampal memory prostheses.
- 2 October 2014
- Published by Elsevier in Journal of Neuroscience Methods
- Vol. 244, 123-135
Identification of functional synaptic plasticity from spiking activities using nonlinear dynamical modeling
- 2 October 2014
- Published by Elsevier in Journal of Neuroscience Methods
- Vol. 244, 123-135
Identification of functional synaptic plasticity from spiking activities using nonlinear dynamical modeling
- 1 August 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 2014 (1094687X) , 714-717
Implementing spiking neuron model and spike-timing-dependent plasticity with generalized Laguerre-Volterra models
To perform large-scale simulations of the brain or build biologically-inspired cognitive architectures, it is essential to have a succinct and flexible model of spiking neurons. The model should be able to capture the nonlinear dynamical properties of various types of neurons and the nonstationary properties such as the spike-timing-dependent plasticity (STDP). In this paper, we propose a generalized Laguerre-Volterra modeling approach for such a task. Due to its built-in nonlinear dynamical terms, the generalized Laguerre-Volterra model (GLVM) can capture various biological processes/mechanisms. Using Laguerre expansion of Volterra kernel technique, the model is fully represented with a small set of coefficients. The calculation of the model variables can be expressed recursively based on only the current and the one-step-before values and thus can be performed efficiently. In addition, we show that, using the same methodology, STDP can be implemented as a specific form of second-order Volterra kernel describing the causal relationship between pairs of input-output spikes and the changes of the feedforward kernels in the GLVMs.
- 1 August 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 2014 (1094687X) , 714-717
Implementing spiking neuron model and spike-timing-dependent plasticity with generalized Laguerre-Volterra models
To perform large-scale simulations of the brain or build biologically-inspired cognitive architectures, it is essential to have a succinct and flexible model of spiking neurons. The model should be able to capture the nonlinear dynamical properties of various types of neurons and the nonstationary properties such as the spike-timing-dependent plasticity (STDP). In this paper, we propose a generalized Laguerre-Volterra modeling approach for such a task. Due to its built-in nonlinear dynamical terms, the generalized Laguerre-Volterra model (GLVM) can capture various biological processes/mechanisms. Using Laguerre expansion of Volterra kernel technique, the model is fully represented with a small set of coefficients. The calculation of the model variables can be expressed recursively based on only the current and the one-step-before values and thus can be performed efficiently. In addition, we show that, using the same methodology, STDP can be implemented as a specific form of second-order Volterra kernel describing the causal relationship between pairs of input-output spikes and the changes of the feedforward kernels in the GLVMs.