Han Kheng Teoh
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Han Kheng Teoh has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
Itai Cohen
Cornell University
4 shared publications- B. Kemper Ludlow
Cornell University
3 shared publications - Itai Cohen
Department of Physics, Cornell University, Ithaca, New York 14853, USA
3 shared publications - Jesse Heymann Goldberg
Cornell University
3 shared publications - Noah J. Cowan
Department of Mechanical Engineering, Johns Hopkins University, Baltimore, MD 21218, USA
3 shared publications - Andrea Roeser
Department of Neurobiology and Behavior, Cornell University
2 shared publications
Latest Publications
- Open Access
- 1 September 2026
- Published by eLife Sciences Publications, Ltd in eLife preprints
- Vol. 15
A multi-muscular, redundant strategy for free-flight roll stability
Whether recovering after a gust of wind, or rapidly saccading away from an oncoming predator, fruit flies show remarkable aerial dexterity about their body roll axis. Here, we investigated the detailed wing kinematic changes during free-flight roll motion and probed the neuromuscular basis for such changes. Consistent with previous work, we observed that flies manipulated the stroke amplitude difference between their wings to control their roll angle. Here, we show that flies are capable of achieving such changes by altering the stroke amplitude of either or both of their wings. Further we found that during corrections flies can also take advantage of an aerodynamically significant change in the angle of attack of their uppermost wing. Curiously, these corrective wing changes cannot be eliminated when motor neurons hypothesized to be used during roll maneuvers (i1, i2, b1, b2, and b3) are individually inhibited. However, free-flight optogenetic manipulations and quasi-steady aerodynamic calculations show that each of these motor neurons individually can effect kinematic changes consistent with a roll correction. Combining this evidence with an analysis of haltere inputs found in the BANC connectome, we propose that the observed robustness could be the result of two sets of muscular redundancies that receive shared inputs from haltere sensory afferents: one set, containing b1 and b2, is able to increase the stroke amplitude of the lower wing; while the other set, containing i1, i2, and b3, is able to decrease the stroke amplitude and wing pitch angle of the upper wing. Because of the redundancy in the input sensory information and output wing motion in the muscles in each cluster, the fly is able to perform roll stability maneuvers even when one of the constituent motor neurons is inhibited. This framework proposes new ways fast aerial maneuverability can be implemented when dealing with the fly’s most unstable rotational degree of freedom.
- 1 September 2026
- Published by eLife Sciences Publications, Ltd in eLife preprints
- Vol. 15
Open AccessA multi-muscular, redundant strategy for free-flight roll stability
Whether recovering after a gust of wind, or rapidly saccading away from an oncoming predator, fruit flies show remarkable aerial dexterity about their body roll axis. Here, we investigated the detailed wing kinematic changes during free-flight roll motion and probed the neuromuscular basis for such changes. Consistent with previous work, we observed that flies manipulated the stroke amplitude difference between their wings to control their roll angle. Here, we show that flies are capable of achieving such changes by altering the stroke amplitude of either or both of their wings. Further we found that during corrections flies can also take advantage of an aerodynamically significant change in the angle of attack of their uppermost wing. Curiously, these corrective wing changes cannot be eliminated when motor neurons hypothesized to be used during roll maneuvers (i1, i2, b1, b2, and b3) are individually inhibited. However, free-flight optogenetic manipulations and quasi-steady aerodynamic calculations show that each of these motor neurons individually can effect kinematic changes consistent with a roll correction. Combining this evidence with an analysis of haltere inputs found in the BANC connectome, we propose that the observed robustness could be the result of two sets of muscular redundancies that receive shared inputs from haltere sensory afferents: one set, containing b1 and b2, is able to increase the stroke amplitude of the lower wing; while the other set, containing i1, i2, and b3, is able to decrease the stroke amplitude and wing pitch angle of the upper wing. Because of the redundancy in the input sensory information and output wing motion in the muscles in each cluster, the fly is able to perform roll stability maneuvers even when one of the constituent motor neurons is inhibited. This framework proposes new ways fast aerial maneuverability can be implemented when dealing with the fly’s most unstable rotational degree of freedom.
- Open Access
How tp1, an indirect wing steering muscle, stabilizes Drosophila’s flight
Flapping flight is inherently unstable, requiring Drosophila to fine-tune their wing motions on a milliseconds timescale. Pioneering studies have shown that the direct steering muscles, which attach to the wing hinge, are important for generating these rapid flight reflexes. Recent connectome data, however, reveal that indirect steering muscles or tension muscles, which alter the mechanics of the thorax, receive some of the same synaptic inputs from the sensory apparatus as the direct steering muscles. This discovery suggests that the indirect steering muscles may also be important for flight control. Here, we show that the indirect tergopleural muscles indeed contribute substantially to stabilization, particularly for large pitch perturbations. We find that for small perturbations (less than 1000 deg/s), flies modulate their wing stroke amplitude to generate lift-based corrective torques, a strategy that has been previously documented. For larger perturbations, however, we observe that Drosophila engage an additional wing degree of freedom—the wing pitch angle—to leverage additional lift and drag forces during the corrective maneuver. Quasi-steady aerodynamic simulations reveal that this strategy minimizes power consumption, anaolgous to how some mammals (including humans) adjust their steady-state gaits in a near energy-optimal manner. Using optogenetics and a control theory framework we demonstrate that the tergopleural muscle is activated by a proportional gain component of a nonlinear PI controller responsible for determining the wing pitch angle during large perturbations. A simplified torsional-spring model for the wing hinge captures the changes in the wing pitch dynamics observed during correction maneuvers by using the tergopleural muscle to adjust the rest angle of the wing. These findings provide a striking example of reflex strategy selection in time-critical behaviors and underscores the vital role of indirect steering muscles in flight stabilization.
Open AccessHow tp1, an indirect wing steering muscle, stabilizes Drosophila’s flight
Flapping flight is inherently unstable, requiring Drosophila to fine-tune their wing motions on a milliseconds timescale. Pioneering studies have shown that the direct steering muscles, which attach to the wing hinge, are important for generating these rapid flight reflexes. Recent connectome data, however, reveal that indirect steering muscles or tension muscles, which alter the mechanics of the thorax, receive some of the same synaptic inputs from the sensory apparatus as the direct steering muscles. This discovery suggests that the indirect steering muscles may also be important for flight control. Here, we show that the indirect tergopleural muscles indeed contribute substantially to stabilization, particularly for large pitch perturbations. We find that for small perturbations (less than 1000 deg/s), flies modulate their wing stroke amplitude to generate lift-based corrective torques, a strategy that has been previously documented. For larger perturbations, however, we observe that Drosophila engage an additional wing degree of freedom—the wing pitch angle—to leverage additional lift and drag forces during the corrective maneuver. Quasi-steady aerodynamic simulations reveal that this strategy minimizes power consumption, anaolgous to how some mammals (including humans) adjust their steady-state gaits in a near energy-optimal manner. Using optogenetics and a control theory framework we demonstrate that the tergopleural muscle is activated by a proportional gain component of a nonlinear PI controller responsible for determining the wing pitch angle during large perturbations. A simplified torsional-spring model for the wing hinge captures the changes in the wing pitch dynamics observed during correction maneuvers by using the tergopleural muscle to adjust the rest angle of the wing. These findings provide a striking example of reflex strategy selection in time-critical behaviors and underscores the vital role of indirect steering muscles in flight stabilization.
- Open Access
A multi-muscular, redundant strategy for free-flight roll stability
Whether recovering after a gust of wind, or rapidly saccading away from an oncoming predator, fruit flies show remarkable aerial dexterity about their body roll axis. Here, we investigated the detailed wing kinematic changes during free-flight roll motion and probed the neuromuscular basis for such changes. Consistent with previous work, we observed that flies manipulated the stroke amplitude difference between their wings to control their roll angle. Here, we show that flies are capable of achieving such changes by altering the stroke amplitude of either or both of their wings. Further we found that during corrections flies can also take advantage of an aerodynamically significant change in the angle of attack of their uppermost wing. Curiously, these corrective wing changes cannot be eliminated when motor neurons hypothesized to be used during roll maneuvers (i1, i2, b1, b2, and b3) are individually inhibited. However, free-flight optogenetic manipulations and quasi-steady aerodynamic calculations show that each of these motor neurons individually can effect kinematic changes consistent with a roll correction. Combining this evidence with an analysis of haltere inputs found in the BANC connectome, we propose that the observed robustness could be the result of two sets of muscular redundancies that receive shared inputs from haltere sensory afferents: one set, containing b1 and b2, is able to increase the stroke amplitude of the lower wing; while the other set, containing i1, i2, and b3, is able to decrease the stroke amplitude and wing pitch angle of the upper wing. Because of the redundancy in the input sensory information and output wing motion in the muscles in each cluster, the fly is able to perform roll stability maneuvers even when one of the constituent motor neurons is inhibited. This framework proposes new ways fast aerial maneuverability can be implemented when dealing with the fly’s most unstable rotational degree of freedom.
Open AccessA multi-muscular, redundant strategy for free-flight roll stability
Whether recovering after a gust of wind, or rapidly saccading away from an oncoming predator, fruit flies show remarkable aerial dexterity about their body roll axis. Here, we investigated the detailed wing kinematic changes during free-flight roll motion and probed the neuromuscular basis for such changes. Consistent with previous work, we observed that flies manipulated the stroke amplitude difference between their wings to control their roll angle. Here, we show that flies are capable of achieving such changes by altering the stroke amplitude of either or both of their wings. Further we found that during corrections flies can also take advantage of an aerodynamically significant change in the angle of attack of their uppermost wing. Curiously, these corrective wing changes cannot be eliminated when motor neurons hypothesized to be used during roll maneuvers (i1, i2, b1, b2, and b3) are individually inhibited. However, free-flight optogenetic manipulations and quasi-steady aerodynamic calculations show that each of these motor neurons individually can effect kinematic changes consistent with a roll correction. Combining this evidence with an analysis of haltere inputs found in the BANC connectome, we propose that the observed robustness could be the result of two sets of muscular redundancies that receive shared inputs from haltere sensory afferents: one set, containing b1 and b2, is able to increase the stroke amplitude of the lower wing; while the other set, containing i1, i2, and b3, is able to decrease the stroke amplitude and wing pitch angle of the upper wing. Because of the redundancy in the input sensory information and output wing motion in the muscles in each cluster, the fly is able to perform roll stability maneuvers even when one of the constituent motor neurons is inhibited. This framework proposes new ways fast aerial maneuverability can be implemented when dealing with the fly’s most unstable rotational degree of freedom.
- Open Access
- 20 May 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (5) , e0297947
Application of a variational autoencoder for clustering and analyzing in situ articular cartilage cellular response to mechanical stimuli
In various biological systems, analyzing how cell behaviors are coordinated over time would enable a deeper understanding of tissue-scale response to physiologic or superphysiologic stimuli. Such data is necessary for establishing both normal tissue function and the sequence of events after injury that lead to chronic disease. However, collecting and analyzing these large datasets presents a challenge—such systems are time-consuming to process, and the overwhelming scale of data makes it difficult to parse overall behaviors. This problem calls for an analysis technique that can quickly provide an overview of the groups present in the entire system and also produce meaningful categorization of cell behaviors. Here, we demonstrate the application of an unsupervised method—the Variational Autoencoder (VAE)—to learn the features of cells in cartilage tissue after impact-induced injury and identify meaningful clusters of chondrocyte behavior. This technique quickly generated new insights into the spatial distribution of specific cell behavior phenotypes and connected specific peracute calcium signaling timeseries with long term cellular outcomes, demonstrating the value of the VAE technique.
- 20 May 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (5) , e0297947
Open AccessApplication of a variational autoencoder for clustering and analyzing in situ articular cartilage cellular response to mechanical stimuli
In various biological systems, analyzing how cell behaviors are coordinated over time would enable a deeper understanding of tissue-scale response to physiologic or superphysiologic stimuli. Such data is necessary for establishing both normal tissue function and the sequence of events after injury that lead to chronic disease. However, collecting and analyzing these large datasets presents a challenge—such systems are time-consuming to process, and the overwhelming scale of data makes it difficult to parse overall behaviors. This problem calls for an analysis technique that can quickly provide an overview of the groups present in the entire system and also produce meaningful categorization of cell behaviors. Here, we demonstrate the application of an unsupervised method—the Variational Autoencoder (VAE)—to learn the features of cells in cartilage tissue after impact-induced injury and identify meaningful clusters of chondrocyte behavior. This technique quickly generated new insights into the spatial distribution of specific cell behavior phenotypes and connected specific peracute calcium signaling timeseries with long term cellular outcomes, demonstrating the value of the VAE technique.
- Open Access
- 12 March 2024
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 121 (12)
The training process of many deep networks explores the same low-dimensional manifold
We develop information-geometric techniques to analyze the trajectories of the predictions of deep networks during training. By examining the underlying high-dimensional probabilistic models, we reveal that the training process explores an effectively low-dimensional manifold. Networks with a wide range of architectures, sizes, trained using different optimization methods, regularization techniques, data augmentation techniques, and weight initializations lie on the same manifold in the prediction space. We study the details of this manifold to find that networks with different architectures follow distinguishable trajectories, but other factors have a minimal influence; larger networks train along a similar manifold as that of smaller networks, just faster; and networks initialized at very different parts of the prediction space converge to the solution along a similar manifold.
- 12 March 2024
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 121 (12)
Open AccessThe training process of many deep networks explores the same low-dimensional manifold
We develop information-geometric techniques to analyze the trajectories of the predictions of deep networks during training. By examining the underlying high-dimensional probabilistic models, we reveal that the training process explores an effectively low-dimensional manifold. Networks with a wide range of architectures, sizes, trained using different optimization methods, regularization techniques, data augmentation techniques, and weight initializations lie on the same manifold in the prediction space. We study the details of this manifold to find that networks with different architectures follow distinguishable trajectories, but other factors have a minimal influence; larger networks train along a similar manifold as that of smaller networks, just faster; and networks initialized at very different parts of the prediction space converge to the solution along a similar manifold.
- Open Access
- 24 August 2023
- Published by eLife Sciences Publications, Ltd in eLife preprints
- Vol. 12
The songbird lateral habenula projects to dopaminergic midbrain and is important for normal vocal development
Mistakes in performance feel disappointing, suggesting that brain pathways for aversive feedback may play a role in motor learning. Here we tested if the lateral habenula (LHb), an evolutionarily conserved part of the limbic system known in mammals to relay aversive feedback from ventral pallidum (VP) to ventral tegmental area (VTA) dopamine neurons, is involved in birdsong learning and production. By combining viral tract tracing and functional circuit mapping, we discovered that songbird LHb links VP and an auditory cortical area to singing-related DA neurons that signal song errors. As in mammals, VP stimulation activated LHb activity and LHb stimulation suppressed DA firing. To test this pathway’s role in learning we lesioned the LHb in juvenile zebra finches and recorded their songs in adulthood. Birds with the LHb lesioned as juveniles produced highly unusual vocalizations as adults, including prolonged high-pitch notes and species-atypical trills. These findings identify a songbird VP-LHb-VTA pathway with similar functional connectivity as mammals, expand the known territories of vocal learning circuits, and demonstrate that limbic circuits associated with disappointing outcomes are important for motor performance learning.
- 24 August 2023
- Published by eLife Sciences Publications, Ltd in eLife preprints
- Vol. 12
Open AccessThe songbird lateral habenula projects to dopaminergic midbrain and is important for normal vocal development
Mistakes in performance feel disappointing, suggesting that brain pathways for aversive feedback may play a role in motor learning. Here we tested if the lateral habenula (LHb), an evolutionarily conserved part of the limbic system known in mammals to relay aversive feedback from ventral pallidum (VP) to ventral tegmental area (VTA) dopamine neurons, is involved in birdsong learning and production. By combining viral tract tracing and functional circuit mapping, we discovered that songbird LHb links VP and an auditory cortical area to singing-related DA neurons that signal song errors. As in mammals, VP stimulation activated LHb activity and LHb stimulation suppressed DA firing. To test this pathway’s role in learning we lesioned the LHb in juvenile zebra finches and recorded their songs in adulthood. Birds with the LHb lesioned as juveniles produced highly unusual vocalizations as adults, including prolonged high-pitch notes and species-atypical trills. These findings identify a songbird VP-LHb-VTA pathway with similar functional connectivity as mammals, expand the known territories of vocal learning circuits, and demonstrate that limbic circuits associated with disappointing outcomes are important for motor performance learning.
- Open Access
The songbird lateral habenula projects to dopaminergic midbrain and is important for normal vocal development
Mistakes in performance feel disappointing, suggesting that brain pathways for aversive feedback may play a role in motor learning. Here we tested if the lateral habenula (LHb), an evolutionarily conserved part of the limbic system known in mammals to relay aversive feedback from ventral pallidum (VP) to ventral tegmental area (VTA) dopamine neurons, is involved in birdsong learning and production. By combining viral tract tracing and functional circuit mapping, we discovered that songbird LHb links VP and an auditory cortical area to singing-related DA neurons that signal song errors. As in mammals, VP stimulation activated LHb activity and LHb stimulation suppressed DA firing. To test this pathway’s role in learning we lesioned the LHb in juvenile zebra finches and recorded their songs in adulthood. Birds with the LHb lesioned as juveniles produced highly unusual vocalizations as adults, including prolonged high-pitch notes and species-atypical trills. These findings identify a songbird VP-LHb-VTA pathway with similar functional connectivity as mammals, expand the known territories of vocal learning circuits, and demonstrate that limbic circuits associated with disappointing outcomes are important for motor performance learning.
Open AccessThe songbird lateral habenula projects to dopaminergic midbrain and is important for normal vocal development
Mistakes in performance feel disappointing, suggesting that brain pathways for aversive feedback may play a role in motor learning. Here we tested if the lateral habenula (LHb), an evolutionarily conserved part of the limbic system known in mammals to relay aversive feedback from ventral pallidum (VP) to ventral tegmental area (VTA) dopamine neurons, is involved in birdsong learning and production. By combining viral tract tracing and functional circuit mapping, we discovered that songbird LHb links VP and an auditory cortical area to singing-related DA neurons that signal song errors. As in mammals, VP stimulation activated LHb activity and LHb stimulation suppressed DA firing. To test this pathway’s role in learning we lesioned the LHb in juvenile zebra finches and recorded their songs in adulthood. Birds with the LHb lesioned as juveniles produced highly unusual vocalizations as adults, including prolonged high-pitch notes and species-atypical trills. These findings identify a songbird VP-LHb-VTA pathway with similar functional connectivity as mammals, expand the known territories of vocal learning circuits, and demonstrate that limbic circuits associated with disappointing outcomes are important for motor performance learning.
- Open Access
Anterior forebrain pathway in parrots is necessary for producing learned vocalizations with individual signatures
Summary: Parrots have enormous vocal imitation capacities and produce individually unique vocal signatures. Like songbirds, parrots have a nucleated neural song system with distinct anterior (AFP) and posterior forebrain pathways (PFP). To test if song systems of parrots and songbirds, which diverged over 50 million years ago, have a similar functional organization, we first established a neuroscience-compatible call-and-response behavioral paradigm to elicit learned contact calls in budgerigars ( Melopsittacus undulatus ). Using variational autoencoder-based machine learning methods, we show that contact calls within affiliated groups converge but that individuals maintain unique acoustic features, or vocal signatures, even after call convergence. Next, we transiently inactivated the outputs of AFP to test if learned vocalizations can be produced by the PFP alone. As in songbirds, AFP inactivation had an immediate effect on vocalizations, consistent with a premotor role. But in contrast to songbirds, where the isolated PFP is sufficient to produce stereotyped and acoustically normal vocalizations, isolation of the budgerigar PFP caused a degradation of call acoustic structure, stereotypy, and individual uniqueness. Thus the contribution of AFP and the capacity of isolated PFP to produce learned vocalizations have diverged substantially between songbirds and parrots, likely driven by their distinct behavioral ecology and neural connectivity.
Open AccessAnterior forebrain pathway in parrots is necessary for producing learned vocalizations with individual signatures
Summary: Parrots have enormous vocal imitation capacities and produce individually unique vocal signatures. Like songbirds, parrots have a nucleated neural song system with distinct anterior (AFP) and posterior forebrain pathways (PFP). To test if song systems of parrots and songbirds, which diverged over 50 million years ago, have a similar functional organization, we first established a neuroscience-compatible call-and-response behavioral paradigm to elicit learned contact calls in budgerigars ( Melopsittacus undulatus ). Using variational autoencoder-based machine learning methods, we show that contact calls within affiliated groups converge but that individuals maintain unique acoustic features, or vocal signatures, even after call convergence. Next, we transiently inactivated the outputs of AFP to test if learned vocalizations can be produced by the PFP alone. As in songbirds, AFP inactivation had an immediate effect on vocalizations, consistent with a premotor role. But in contrast to songbirds, where the isolated PFP is sufficient to produce stereotyped and acoustically normal vocalizations, isolation of the budgerigar PFP caused a degradation of call acoustic structure, stereotypy, and individual uniqueness. Thus the contribution of AFP and the capacity of isolated PFP to produce learned vocalizations have diverged substantially between songbirds and parrots, likely driven by their distinct behavioral ecology and neural connectivity.