Denis Turcu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Denis Turcu has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Bingni Wen Brunton
Department of Biology, University of Washington, Seattle, USA
2 shared publications Jackson K. Lancaster
Department of Biology, University of Washington, Seattle, USA
2 shared publications- John C. Tuthill
University of Washington
2 shared publications L. F. Abbott
Affiliation The Mortimer B. Zuckerman Mind, Brain and Behavior Institute, Department of Neuroscience, Columbia University, New York, New York, United States of America
2 shared publicationsL.F. Abbott
The Mortimer B. Zuckerman Mind, Brain and Behavior Institute, Columbia University, New York, NY 10027, USA; Department of Neuroscience, Columbia University, New York, NY 10027, USA; Kavli Institute for Brain Science, Columbia University, New York, NY 10027, USA
2 shared publications- Nathaniel B. Sawtell
Columbia University
2 shared publications
Latest Publications
- 28 October 2026
- Published by Springer Nature in Nature
- p. 1-9
Connectome simulations identify a central pattern generator for fly walking
Animal locomotion relies on rhythmic body movements driven by central pattern generators (CPGs): neural circuits that produce oscillating output without oscillating input. However, the circuit structure of a CPG for walking is not known in any animal1,2. To identify the cells and synapses that underlie rhythmic leg movement in walking flies, we developed dynamic simulations of the Drosophila ventral nerve cord connectomes3,4,5,6. A computational activation screen of descending neurons from the central brain identified DNg100—a known command neuron for walking—as the strongest driver of rhythmic leg motor activity. Simulated network pruning isolated a minimal rhythm-generating circuit consisting of one inhibitory and two excitatory interneurons; this three-neuron circuit was necessary and sufficient for motor rhythms across all six legs and in four connectome datasets. Simulations also predicted that a separate descending pathway (DNb08) drives rhythmic leg movements, which we confirmed experimentally using optogenetics in behaving flies. Our results reveal the cellular identity and synaptic structure of a putative CPG circuit for walking and other rhythmic leg movements in flies.
- 28 October 2026
- Published by Springer Nature in Nature
- p. 1-9
Connectome simulations identify a central pattern generator for fly walking
Animal locomotion relies on rhythmic body movements driven by central pattern generators (CPGs): neural circuits that produce oscillating output without oscillating input. However, the circuit structure of a CPG for walking is not known in any animal1,2. To identify the cells and synapses that underlie rhythmic leg movement in walking flies, we developed dynamic simulations of the Drosophila ventral nerve cord connectomes3,4,5,6. A computational activation screen of descending neurons from the central brain identified DNg100—a known command neuron for walking—as the strongest driver of rhythmic leg motor activity. Simulated network pruning isolated a minimal rhythm-generating circuit consisting of one inhibitory and two excitatory interneurons; this three-neuron circuit was necessary and sufficient for motor rhythms across all six legs and in four connectome datasets. Simulations also predicted that a separate descending pathway (DNb08) drives rhythmic leg movements, which we confirmed experimentally using optogenetics in behaving flies. Our results reveal the cellular identity and synaptic structure of a putative CPG circuit for walking and other rhythmic leg movements in flies.
- Open Access
Discrete synaptic states and context-modulated readouts support continual learning
Biological neural systems learn new behaviors while retaining earlier ones, whereas sequentially trained artificial networks often overwrite parameters and forget previously learned behaviors. We introduce the Context-modulated Synaptic-states Recurrent Neural Network (CoSyn-RNN), a new continual-learning model inspired by heterogeneous synaptic states and neuromodulatory control. CoSyn-RNN combines sparse recurrent allocation with a discrete protection rule: after each task, recurrent neurons whose synaptic weights cross a fixed threshold form a task-specific neuron group, and their associated parameters enter a protected state during later learning. A neuron-specific gain and a task-cued mask over a fixed base readout control how recurrent activity contributes to behavior. Across various task sequences consisting of many cognitive tasks, training and validation losses and accuracies were maintained throughout learning. The protection rule produced modular, task-ordered recurrent structure compatible with forward reuse of earlier computations. Pairwise recruitment measurements further defined minimum- and maximum-cost curricula that produced markedly different final recurrent structures and different remaining capacity for future learning. Thus, CoSyn-RNN provides a biologically inspired model in which protected synaptic states, sparse allocation, and context-dependent modulation support continual learning.
Open AccessDiscrete synaptic states and context-modulated readouts support continual learning
Biological neural systems learn new behaviors while retaining earlier ones, whereas sequentially trained artificial networks often overwrite parameters and forget previously learned behaviors. We introduce the Context-modulated Synaptic-states Recurrent Neural Network (CoSyn-RNN), a new continual-learning model inspired by heterogeneous synaptic states and neuromodulatory control. CoSyn-RNN combines sparse recurrent allocation with a discrete protection rule: after each task, recurrent neurons whose synaptic weights cross a fixed threshold form a task-specific neuron group, and their associated parameters enter a protected state during later learning. A neuron-specific gain and a task-cued mask over a fixed base readout control how recurrent activity contributes to behavior. Across various task sequences consisting of many cognitive tasks, training and validation losses and accuracies were maintained throughout learning. The protection rule produced modular, task-ordered recurrent structure compatible with forward reuse of earlier computations. Pairwise recruitment measurements further defined minimum- and maximum-cost curricula that produced markedly different final recurrent structures and different remaining capacity for future learning. Thus, CoSyn-RNN provides a biologically inspired model in which protected synaptic states, sparse allocation, and context-dependent modulation support continual learning.
- Open Access
Learning using switching synaptic plasticity rules
1 Hebbian-like learning has been repeatedly confirmed experimentally, yet computational models usually require non-local signals, such as backpropagating errors, to solve complex cognitive tasks. Recent cortical electron microscopy data suggests a model where synapses follow different plasticity rules depending on whether they are in a large or small state. Large synapses often include a spine apparatus, a calcium reservoir that influences synaptic dynamics and can alter rules of synaptic plasticity. Here, we test the computational outcomes of networks which compute with synapses switching their plasticity rules based on their strength. We designed a recurrent neural network (RNN) with synapses that switch between two learning rules: a Hebbian-like rule for weak synapses and a credit-assignment rule (backpropagation, BP) for strong synapses. We found that our plasticity-switching RNN (psRNN) learns cognitive tasks (e.g. working memory) in fewer trials than BP-only RNNs, despite fewer synapses using credit assignment. Three mechanisms underlie this advantage: BP samples multiple parameter configurations for better gradient estimation, Hebbian plasticity creates a dynamic task-relevant initialization, and the switching mechanism prevents Hebbian synapses from growing into unfavorable parameter regions. The interaction between rules also produces lower-rank, more feedforward recurrent structure, providing testable connectomic predictions and a framework for reconciling local learning rules observed in the brain with non-local rules used in computational models. Significance statement: Credit assignment is essential for learning in biological and artificial systems. Credit assignment can be achieved via complex synaptic plasticity rules, yet biological data points to plasticity rules at most synapses being simpler. Here we ask if networks with a combination of simple and complex plasticity rules could solve biorealistic tasks. Surprisingly, replacing the computationally expensive error-based learning with simple, experience-dependent, changes in a large portion of synapses improves learning in our models. This substitution changes several computational properties of the model, offering new hypotheses for computing in the brain.
Open AccessLearning using switching synaptic plasticity rules
1 Hebbian-like learning has been repeatedly confirmed experimentally, yet computational models usually require non-local signals, such as backpropagating errors, to solve complex cognitive tasks. Recent cortical electron microscopy data suggests a model where synapses follow different plasticity rules depending on whether they are in a large or small state. Large synapses often include a spine apparatus, a calcium reservoir that influences synaptic dynamics and can alter rules of synaptic plasticity. Here, we test the computational outcomes of networks which compute with synapses switching their plasticity rules based on their strength. We designed a recurrent neural network (RNN) with synapses that switch between two learning rules: a Hebbian-like rule for weak synapses and a credit-assignment rule (backpropagation, BP) for strong synapses. We found that our plasticity-switching RNN (psRNN) learns cognitive tasks (e.g. working memory) in fewer trials than BP-only RNNs, despite fewer synapses using credit assignment. Three mechanisms underlie this advantage: BP samples multiple parameter configurations for better gradient estimation, Hebbian plasticity creates a dynamic task-relevant initialization, and the switching mechanism prevents Hebbian synapses from growing into unfavorable parameter regions. The interaction between rules also produces lower-rank, more feedforward recurrent structure, providing testable connectomic predictions and a framework for reconciling local learning rules observed in the brain with non-local rules used in computational models. Significance statement: Credit assignment is essential for learning in biological and artificial systems. Credit assignment can be achieved via complex synaptic plasticity rules, yet biological data points to plasticity rules at most synapses being simpler. Here we ask if networks with a combination of simple and complex plasticity rules could solve biorealistic tasks. Surprisingly, replacing the computationally expensive error-based learning with simple, experience-dependent, changes in a large portion of synapses improves learning in our models. This substitution changes several computational properties of the model, offering new hypotheses for computing in the brain.
- Open Access
Connectome simulations identify a central pattern generator circuit for fly walking
Animal locomotion relies on rhythmic body movements driven by central pattern generators (CPGs): neural circuits that produce oscillating output without oscillating input. However, the circuit structure of a CPG for walking is not known in any animal. To identify the cells and synapses that underlie rhythmic leg movement in walking flies, we developed dynamic simulations of the Drosophila ventral nerve cord (VNC) connectomes. A computational activation screen of descending neurons from the central brain identified DNg100—a known command neuron for walking—as the top driver of rhythmic leg motor activity. Simulated network pruning isolated a minimal rhythm-generating circuit consisting of one inhibitory and two excitatory interneurons; this three-neuron circuit was necessary and sufficient for motor rhythms across all six legs and in four connectome datasets. Simulations also predicted that a separate descending pathway (DNb08) drives rhythmic leg movements, which we confirmed experimentally using optogenetics in behaving flies. Our results reveal the cellular identity and synaptic structure of a putative CPG circuit for walking and other rhythmic leg movements in flies.
Open AccessConnectome simulations identify a central pattern generator circuit for fly walking
Animal locomotion relies on rhythmic body movements driven by central pattern generators (CPGs): neural circuits that produce oscillating output without oscillating input. However, the circuit structure of a CPG for walking is not known in any animal. To identify the cells and synapses that underlie rhythmic leg movement in walking flies, we developed dynamic simulations of the Drosophila ventral nerve cord (VNC) connectomes. A computational activation screen of descending neurons from the central brain identified DNg100—a known command neuron for walking—as the top driver of rhythmic leg motor activity. Simulated network pruning isolated a minimal rhythm-generating circuit consisting of one inhibitory and two excitatory interneurons; this three-neuron circuit was necessary and sufficient for motor rhythms across all six legs and in four connectome datasets. Simulations also predicted that a separate descending pathway (DNb08) drives rhythmic leg movements, which we confirmed experimentally using optogenetics in behaving flies. Our results reveal the cellular identity and synaptic structure of a putative CPG circuit for walking and other rhythmic leg movements in flies.
- 23 April 2025
- Published by Elsevier in Current Biology
- Vol. 35 (10) , 2295-2306.e4
An end-to-end model of active electrosensation
- 23 April 2025
- Published by Elsevier in Current Biology
- Vol. 35 (10) , 2295-2306.e4
An end-to-end model of active electrosensation
- Open Access
An end-to-end model of active electrosensation
1 Weakly electric fish localize and identify objects by sensing distortions in a self-generated electric field. Fish can determine the resistance and capacitance of an object, for example, even though the field distortions being sensed are small and highly-dependent on object distance and size. Here we construct a model of the responses of the fish’s electroreceptors on the basis of experimental data, and we develop a model of the electric fields generated by the fish and the distortions due to objects of different resistances and capacitances. This provides us with an accurate and efficient method for generating large artificial data sets simulating fish interacting with a wide variety of objects. Using these sets, we train an artificial neural network (ANN), representing brain areas downstream of electroreceptors, to extract the 3D location, size, and electrical properties of objects. The model performs best if the ANN operates in two stages: first estimating object distance and size and then using this information to extract electrical properties. This suggests a specific form of modularity in the electrosensory system that can be tested experimentally and highlights the potential of end-to-end modeling for studies of sensory processing.
Open AccessAn end-to-end model of active electrosensation
1 Weakly electric fish localize and identify objects by sensing distortions in a self-generated electric field. Fish can determine the resistance and capacitance of an object, for example, even though the field distortions being sensed are small and highly-dependent on object distance and size. Here we construct a model of the responses of the fish’s electroreceptors on the basis of experimental data, and we develop a model of the electric fields generated by the fish and the distortions due to objects of different resistances and capacitances. This provides us with an accurate and efficient method for generating large artificial data sets simulating fish interacting with a wide variety of objects. Using these sets, we train an artificial neural network (ANN), representing brain areas downstream of electroreceptors, to extract the 3D location, size, and electrical properties of objects. The model performs best if the ANN operates in two stages: first estimating object distance and size and then using this information to extract electrical properties. This suggests a specific form of modularity in the electrosensory system that can be tested experimentally and highlights the potential of end-to-end modeling for studies of sensory processing.
- Open Access
- 14 December 2022
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 18 (12) , e1010759
Sparse RNNs can support high-capacity classification
Feedforward network models performing classification tasks rely on highly convergent output units that collect the information passed on by preceding layers. Although convergent output-unit like neurons may exist in some biological neural circuits, notably the cerebellar cortex, neocortical circuits do not exhibit any obvious candidates for this role; instead they are highly recurrent. We investigate whether a sparsely connected recurrent neural network (RNN) can perform classification in a distributed manner without ever bringing all of the relevant information to a single convergence site. Our model is based on a sparse RNN that performs classification dynamically. Specifically, the interconnections of the RNN are trained to resonantly amplify the magnitude of responses to some external inputs but not others. The amplified and non-amplified responses then form the basis for binary classification. Furthermore, the network acts as an evidence accumulator and maintains its decision even after the input is turned off. Despite highly sparse connectivity, learned recurrent connections allow input information to flow to every neuron of the RNN, providing the basis for distributed computation. In this arrangement, the minimum number of synapses per neuron required to reach maximum memory capacity scales only logarithmically with network size. The model is robust to various types of noise, works with different activation and loss functions and with both backpropagation- and Hebbian-based learning rules. The RNN can also be constructed with a split excitation-inhibition architecture with little reduction in performance.
- 14 December 2022
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 18 (12) , e1010759
Open AccessSparse RNNs can support high-capacity classification
Feedforward network models performing classification tasks rely on highly convergent output units that collect the information passed on by preceding layers. Although convergent output-unit like neurons may exist in some biological neural circuits, notably the cerebellar cortex, neocortical circuits do not exhibit any obvious candidates for this role; instead they are highly recurrent. We investigate whether a sparsely connected recurrent neural network (RNN) can perform classification in a distributed manner without ever bringing all of the relevant information to a single convergence site. Our model is based on a sparse RNN that performs classification dynamically. Specifically, the interconnections of the RNN are trained to resonantly amplify the magnitude of responses to some external inputs but not others. The amplified and non-amplified responses then form the basis for binary classification. Furthermore, the network acts as an evidence accumulator and maintains its decision even after the input is turned off. Despite highly sparse connectivity, learned recurrent connections allow input information to flow to every neuron of the RNN, providing the basis for distributed computation. In this arrangement, the minimum number of synapses per neuron required to reach maximum memory capacity scales only logarithmically with network size. The model is robust to various types of noise, works with different activation and loss functions and with both backpropagation- and Hebbian-based learning rules. The RNN can also be constructed with a split excitation-inhibition architecture with little reduction in performance.
- Open Access
Sparse RNNs can support high-capacity classification
Feedforward network models performing classification tasks rely on highly convergent output units that collect the information passed on by preceding layers. Although convergent output-unit like neurons may exist in some biological neural circuits, notably the cerebellar cortex, neocortical circuits do not exhibit any obvious candidates for this role; instead they are highly recurrent. We investigate whether a sparsely connected recurrent neural network (RNN) can perform classification in a distributed manner without ever bringing all of the relevant information to a single convergence site. Our model is based on a sparse RNN that performs classification dynamically. Specifically, the interconnections of the RNN are trained to resonantly amplify the magnitude of responses to some external inputs but not others. The amplified and non-amplified responses then form the basis for binary classification. Furthermore, the network acts as an evidence accumulator and maintains its decision even after the input is turned off. Despite highly sparse connectivity, learned recurrent connections allow input information to flow to every neuron of the RNN, providing the basis for distributed computation. In this arrangement, the minimum number of synapses per neuron required to reach maximum memory capacity scales only logarithmically with network size. The model is robust to various types of noise, works with different activation and loss functions and with both backpropagation- and Hebbian-based learning rules. The RNN can also be constructed with a split excitation-inhibition architecture with little reduction in performance.
Open AccessSparse RNNs can support high-capacity classification
Feedforward network models performing classification tasks rely on highly convergent output units that collect the information passed on by preceding layers. Although convergent output-unit like neurons may exist in some biological neural circuits, notably the cerebellar cortex, neocortical circuits do not exhibit any obvious candidates for this role; instead they are highly recurrent. We investigate whether a sparsely connected recurrent neural network (RNN) can perform classification in a distributed manner without ever bringing all of the relevant information to a single convergence site. Our model is based on a sparse RNN that performs classification dynamically. Specifically, the interconnections of the RNN are trained to resonantly amplify the magnitude of responses to some external inputs but not others. The amplified and non-amplified responses then form the basis for binary classification. Furthermore, the network acts as an evidence accumulator and maintains its decision even after the input is turned off. Despite highly sparse connectivity, learned recurrent connections allow input information to flow to every neuron of the RNN, providing the basis for distributed computation. In this arrangement, the minimum number of synapses per neuron required to reach maximum memory capacity scales only logarithmically with network size. The model is robust to various types of noise, works with different activation and loss functions and with both backpropagation- and Hebbian-based learning rules. The RNN can also be constructed with a split excitation-inhibition architecture with little reduction in performance.
- 1 January 2022
- Published by Springer Nature
- p. 3-14
Optimal Scheduling of the Leaves of a Tree and the SVO Frequencies of Languages
- 1 January 2022
- Published by Springer Nature
- p. 3-14
Optimal Scheduling of the Leaves of a Tree and the SVO Frequencies of Languages