Demba Ba
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
Demba Ba has a total of 61 co-authors. Here are the top co-authors by number of shared publications:
- Bahareh Tolooshams
Computing + Mathematical Sciences, California Institute of Technology
9 shared publications - Andrew H. Song
Department of Pathology, Mass General Brigham, Harvard Medical School, Boston, USA
6 shared publications - Paul Masset
Affiliations Department of Psychology, McGill University, Montréal, Québec, Canada, Mila - Quebec Artificial Intelligence Institute, Montréal, Québec, Canada
5 shared publications - Abiy Tasissa
Department of Mathematics, Tufts University, Medford, MA, USA
4 shared publications - Venkatesh N. Murthy
Harvard University
4 shared publications - Simona Temereanca
Carney Institute for Brain Science, Brown University, Providence, RI 02906, USA
3 shared publications
Latest Publications
- 20 May 2026
- Published by MIT Press in Neural Computation
- Vol. 38 (6) , 897-951
Implicit Generative Modeling by Kernel Similarity Matching
Understanding how the brain encodes stimuli has been a fundamental problem in computational neuroscience. Insights into this problem have led to the design and development of artificial neural networks that learn representations by incorporating brain-like learning abilities. Recently, learning representations by capturing similarity among input samples has been studied (Pehlevan et al., 2018) to tackle this problem. This approach, however, has thus far been used only to learn downstream features from an input and has not been studied in the context of a generative paradigm, where one can map the representations back to the input space, incorporating not only bottom-up interactions (stimuli→latent) but also learning features in a top-down manner (latent→stimuli). We investigate a kernel similarity matching framework for generative modeling. Starting with a modified sparse coding objective for learning representations proposed in prior works (Olshausen & Field, 1996; Tolooshams & Ba, 2022), we demonstrate that representation learning in this context is equivalent to maximizing similarity between the input kernel and a latent kernel. We show that an implicit generative model arises from learning the kernel structure in the latent space and show how the framework can be adapted to learn manifold structures, potentially providing insights as to how task representations can be encoded in the brain. To solve the objective, we propose a novel alternate direction method of multipliers (ADMM)–based algorithm and discuss the interpretation of the optimization process. Finally, we discuss how this representation learning problem can lead toward a biologically plausible architecture to learn the model parameters that ties together representation learning using similarity matching (a bottom-up approach) with predictive coding (a top-down approach).
- 20 May 2026
- Published by MIT Press in Neural Computation
- Vol. 38 (6) , 897-951
Implicit Generative Modeling by Kernel Similarity Matching
Understanding how the brain encodes stimuli has been a fundamental problem in computational neuroscience. Insights into this problem have led to the design and development of artificial neural networks that learn representations by incorporating brain-like learning abilities. Recently, learning representations by capturing similarity among input samples has been studied (Pehlevan et al., 2018) to tackle this problem. This approach, however, has thus far been used only to learn downstream features from an input and has not been studied in the context of a generative paradigm, where one can map the representations back to the input space, incorporating not only bottom-up interactions (stimuli→latent) but also learning features in a top-down manner (latent→stimuli). We investigate a kernel similarity matching framework for generative modeling. Starting with a modified sparse coding objective for learning representations proposed in prior works (Olshausen & Field, 1996; Tolooshams & Ba, 2022), we demonstrate that representation learning in this context is equivalent to maximizing similarity between the input kernel and a latent kernel. We show that an implicit generative model arises from learning the kernel structure in the latent space and show how the framework can be adapted to learn manifold structures, potentially providing insights as to how task representations can be encoded in the brain. To solve the objective, we propose a novel alternate direction method of multipliers (ADMM)–based algorithm and discuss the interpretation of the optimization process. Finally, we discuss how this representation learning problem can lead toward a biologically plausible architecture to learn the model parameters that ties together representation learning using similarity matching (a bottom-up approach) with predictive coding (a top-down approach).
- Open Access
- 29 April 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Perception and neural representation of intermittent odor stimuli in mice
Odor cues in nature are sparse and fluctuating due to turbulent transport. To investigate how animals perceive these intermittent cues, we developed a behavioral task in which mice made binary decisions based on the total number of discrete odor pulses presented stochastically over several seconds. Mice quickly learned this task, placing higher perceptual weight to stimuli arriving during inhalation than exhalation, a phase dependency that strongly correlated with the magnitude of responses in olfactory sensory neurons. Neurons in the anterior piriform cortex responded to odor pulses with varying degrees of dependence on respiration phase. Single cortical neurons responded stochastically and transiently to odor pulses, leading to a representation that carries signatures of sensory evidence, but not its accumulation. Our study reveals that mice can integrate intermittent odor signals across dozens of breaths and that respiratory modulation imposes limits on sensory information acquisition that cortical circuits cannot overcome to improve behavior.
- 29 April 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessPerception and neural representation of intermittent odor stimuli in mice
Odor cues in nature are sparse and fluctuating due to turbulent transport. To investigate how animals perceive these intermittent cues, we developed a behavioral task in which mice made binary decisions based on the total number of discrete odor pulses presented stochastically over several seconds. Mice quickly learned this task, placing higher perceptual weight to stimuli arriving during inhalation than exhalation, a phase dependency that strongly correlated with the magnitude of responses in olfactory sensory neurons. Neurons in the anterior piriform cortex responded to odor pulses with varying degrees of dependence on respiration phase. Single cortical neurons responded stochastically and transiently to odor pulses, leading to a representation that carries signatures of sensory evidence, but not its accumulation. Our study reveals that mice can integrate intermittent odor signals across dozens of breaths and that respiratory modulation imposes limits on sensory information acquisition that cortical circuits cannot overcome to improve behavior.
- Open Access
Perception and neural representation of intermittent odor stimuli in mice
Odor cues in nature are sparse and highly fluctuating due to turbulent transport. To investigate how animals perceive these intermittent cues, we developed a behavioral task in which head-restrained mice made binary decisions based on the total number of discrete odor pulses presented stochastically over several seconds. Mice readily learned this task, and their performance was well-described by widely used decision models. Logistic regression of binary choices against the timing of odor pulses within the respiratory cycle revealed that mice placed higher perceptual weight to stimuli arriving during inhalation than exhalation, a phase dependency that strongly correlated with the magnitude of responses in olfactory sensory neurons. The population response of anterior piriform cortex (APCx) neurons to odor pulses was also modulated by respiration phase, although individual neurons displayed varying levels of phase-dependence. Single APCx neurons responded stochastically and transiently to odor pulses, leading to a representation that carries signatures of sensory evidence, but not its accumulation. Our study reveals that mice can integrate intermittent odor signals across dozens of breaths, but respiratory modulation of sensory inputs imposes limits on information acquisition that cortical circuits cannot overcome to improve behavior.
Open AccessPerception and neural representation of intermittent odor stimuli in mice
Odor cues in nature are sparse and highly fluctuating due to turbulent transport. To investigate how animals perceive these intermittent cues, we developed a behavioral task in which head-restrained mice made binary decisions based on the total number of discrete odor pulses presented stochastically over several seconds. Mice readily learned this task, and their performance was well-described by widely used decision models. Logistic regression of binary choices against the timing of odor pulses within the respiratory cycle revealed that mice placed higher perceptual weight to stimuli arriving during inhalation than exhalation, a phase dependency that strongly correlated with the magnitude of responses in olfactory sensory neurons. The population response of anterior piriform cortex (APCx) neurons to odor pulses was also modulated by respiration phase, although individual neurons displayed varying levels of phase-dependence. Single APCx neurons responded stochastically and transiently to odor pulses, leading to a representation that carries signatures of sensory evidence, but not its accumulation. Our study reveals that mice can integrate intermittent odor signals across dozens of breaths, but respiratory modulation of sensory inputs imposes limits on information acquisition that cortical circuits cannot overcome to improve behavior.
- Open Access
Learning Artistic Signatures: Symmetry Discovery and Style Transfer
Despite nearly a decade of literature on style transfer, there is no undisputed definition of artistic style. State-of-the-art models produce impressive results but are difficult to interpret since, without a coherent definition of style, the problem of style transfer is inherently ill-posed. Early work framed style-transfer as an optimization problem but treated style as a measure only of texture. This led to artifacts in the outputs of early models where content features from the style image sometimes bled into the output image. Conversely, more recent work with diffusion models offers compelling empirical results but provides little theoretical grounding. To address these issues, we propose an alternative definition of artistic style. We suggest that style should be thought of as a set of global symmetries that dictate the arrangement of local textures. We validate this perspective empirically by learning the symmetries of a large dataset of paintings and showing that symmetries are predictive of the artistic movement to which each painting belongs. Finally, we show that by considering both local and global features, using both Lie generators and traditional measures of texture, we can quantitatively capture the stylistic similarity between artists better than with either set of features alone. This approach not only aligns well with art historians' consensus but also offers a robust framework for distinguishing nuanced stylistic differences, allowing for a more interpretable, theoretically grounded approach to style transfer.
Open AccessLearning Artistic Signatures: Symmetry Discovery and Style Transfer
Despite nearly a decade of literature on style transfer, there is no undisputed definition of artistic style. State-of-the-art models produce impressive results but are difficult to interpret since, without a coherent definition of style, the problem of style transfer is inherently ill-posed. Early work framed style-transfer as an optimization problem but treated style as a measure only of texture. This led to artifacts in the outputs of early models where content features from the style image sometimes bled into the output image. Conversely, more recent work with diffusion models offers compelling empirical results but provides little theoretical grounding. To address these issues, we propose an alternative definition of artistic style. We suggest that style should be thought of as a set of global symmetries that dictate the arrangement of local textures. We validate this perspective empirically by learning the symmetries of a large dataset of paintings and showing that symmetries are predictive of the artistic movement to which each painting belongs. Finally, we show that by considering both local and global features, using both Lie generators and traditional measures of texture, we can quantitatively capture the stylistic similarity between artists better than with either set of features alone. This approach not only aligns well with art historians' consensus but also offers a robust framework for distinguishing nuanced stylistic differences, allowing for a more interpretable, theoretically grounded approach to style transfer.
- Open Access
An unbiased method to partition diverse neuronal responses into functional ensembles reveals interpretable population dynamics during innate social behavior
In neuroscience, understanding how single-neuron firing contributes to distributed neural ensembles is crucial. Traditional methods of analysis have been limited to descriptions of whole population activity, or, when analyzing individual neurons, criteria for response categorization varied significantly across experiments. Current methods lack scalability for large datasets, fail to capture temporal changes and rely on parametric assumptions. There’s a need for a robust, scalable, and non-parametric functional clustering approach to capture interpretable dynamics. To address this challenge, we developed a model-based, statistical framework for unsupervised clustering of multiple time series datasets that exhibit nonlinear dynamics into an a-priori- unknown number of parameterized ensembles called Functional Encoding Units (FEUs). FEU outperforms existing techniques in accuracy and benchmark scores. Here, we apply this FEU formalism to single-unit recordings collected during social behaviors in rodents and primates and demonstrate its hypothesis-generating and testing capacities. This novel pipeline serves as an analytic bridge, translating neural ensemble codes across model systems.
Open AccessAn unbiased method to partition diverse neuronal responses into functional ensembles reveals interpretable population dynamics during innate social behavior
In neuroscience, understanding how single-neuron firing contributes to distributed neural ensembles is crucial. Traditional methods of analysis have been limited to descriptions of whole population activity, or, when analyzing individual neurons, criteria for response categorization varied significantly across experiments. Current methods lack scalability for large datasets, fail to capture temporal changes and rely on parametric assumptions. There’s a need for a robust, scalable, and non-parametric functional clustering approach to capture interpretable dynamics. To address this challenge, we developed a model-based, statistical framework for unsupervised clustering of multiple time series datasets that exhibit nonlinear dynamics into an a-priori- unknown number of parameterized ensembles called Functional Encoding Units (FEUs). FEU outperforms existing techniques in accuracy and benchmark scores. Here, we apply this FEU formalism to single-unit recordings collected during social behaviors in rodents and primates and demonstrate its hypothesis-generating and testing capacities. This novel pipeline serves as an analytic bridge, translating neural ensemble codes across model systems.
- Open Access
Interpretable deep learning for deconvolutional analysis of neural signals
The widespread adoption of deep learning to build models that capture the dynamics of neural populations is typically based on “black-box” approaches that lack an interpretable link between neural activity and network parameters. Here, we propose to apply algorithm unrolling, a method for interpretable deep learning, to design the architecture of sparse deconvolutional neural networks and obtain a direct interpretation of network weights in relation to stimulus-driven single-neuron activity through a generative model. We characterize our method, referred to as deconvolutional unrolled neural learning (DUNL), and show its versatility by applying it to deconvolve single-trial local signals across multiple brain areas and recording modalities. To exemplify use cases of our decomposition method, we uncover multiplexed salience and reward prediction error signals from midbrain dopamine neurons in an unbiased manner, perform simultaneous event detection and characterization in somatosensory thalamus recordings, and characterize the heterogeneity of neural responses in the piriform cortex and in the striatum during unstructured, naturalistic experiments. Our work leverages the advances in interpretable deep learning to gain a mechanistic understanding of neural activity.
Open AccessInterpretable deep learning for deconvolutional analysis of neural signals
The widespread adoption of deep learning to build models that capture the dynamics of neural populations is typically based on “black-box” approaches that lack an interpretable link between neural activity and network parameters. Here, we propose to apply algorithm unrolling, a method for interpretable deep learning, to design the architecture of sparse deconvolutional neural networks and obtain a direct interpretation of network weights in relation to stimulus-driven single-neuron activity through a generative model. We characterize our method, referred to as deconvolutional unrolled neural learning (DUNL), and show its versatility by applying it to deconvolve single-trial local signals across multiple brain areas and recording modalities. To exemplify use cases of our decomposition method, we uncover multiplexed salience and reward prediction error signals from midbrain dopamine neurons in an unbiased manner, perform simultaneous event detection and characterization in somatosensory thalamus recordings, and characterize the heterogeneity of neural responses in the piriform cortex and in the striatum during unstructured, naturalistic experiments. Our work leverages the advances in interpretable deep learning to gain a mechanistic understanding of neural activity.
- 12 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 71 (1053587X) , 3741-3754
K-Deep Simplex: Manifold Learning via Local Dictionaries
We propose K- Deep Simplex (KDS) which, given a set of data points, learns a dictionary comprising synthetic landmarks, along with representation coefficients supported on a simplex. KDS employs a local weighted ℓ 1 penalty that encourages each data point to represent itself as a convex combination of nearby landmarks. We solve the proposed optimization program using alternating minimization and design an efficient, interpretable autoencoder using algorithm unrolling. We theoretically analyse the proposed program by relating the weighted ℓ 1 penalty in KDS to a weighted ℓ 0 program. Assuming that the data are generated from a Delaunay triangulation, we prove the equivalence of the weighted ℓ 1 and weighted ℓ 0 programs. We further show the stability of the representation coefficients under mild geometrical assumptions. If the representation coefficients are fixed, we prove that the sub-problem of minimizing over the dictionary yields a unique solution. Further, we show that low-dimensional representations can be efficiently obtained from the covariance of the coefficient matrix. Experiments show that the algorithm is highly efficient and performs competitively on synthetic and real data sets.
- 12 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 71 (1053587X) , 3741-3754
K-Deep Simplex: Manifold Learning via Local Dictionaries
We propose K- Deep Simplex (KDS) which, given a set of data points, learns a dictionary comprising synthetic landmarks, along with representation coefficients supported on a simplex. KDS employs a local weighted ℓ 1 penalty that encourages each data point to represent itself as a convex combination of nearby landmarks. We solve the proposed optimization program using alternating minimization and design an efficient, interpretable autoencoder using algorithm unrolling. We theoretically analyse the proposed program by relating the weighted ℓ 1 penalty in KDS to a weighted ℓ 0 program. Assuming that the data are generated from a Delaunay triangulation, we prove the equivalence of the weighted ℓ 1 and weighted ℓ 0 programs. We further show the stability of the representation coefficients under mild geometrical assumptions. If the representation coefficients are fixed, we prove that the sub-problem of minimizing over the dictionary yields a unique solution. Further, we show that low-dimensional representations can be efficiently obtained from the covariance of the coefficient matrix. Experiments show that the algorithm is highly efficient and performs competitively on synthetic and real data sets.
- Open Access
SHAPER: Can You Hear the Shape of a Jet?
The identification of interesting substructures within jets is an important tool for searching for new physics and probing the Standard Model at colliders. Many of these substructure tools have previously been shown to take the form of optimal transport problems, in particular the Energy Mover's Distance (EMD). In this work, we show that the EMD is in fact the natural structure for comparing collider events, which accounts for its recent success in understanding event and jet substructure. We then present a Shape Hunting Algorithm using Parameterized Energy Reconstruction (SHAPER), which is a general framework for defining and computing shape-based observables. SHAPER generalizes N-jettiness from point clusters to any extended, parametrizable shape. This is accomplished by efficiently minimizing the EMD between events and parameterized manifolds of energy flows representing idealized shapes, implemented using the dual-potential Sinkhorn approximation of the Wasserstein metric. We show how the geometric language of observables as manifolds can be used to define novel observables with built-in infrared-and-collinear safety. We demonstrate the efficacy of the SHAPER framework by performing empirical jet substructure studies using several examples of new shape-based observables.
Open AccessSHAPER: Can You Hear the Shape of a Jet?
The identification of interesting substructures within jets is an important tool for searching for new physics and probing the Standard Model at colliders. Many of these substructure tools have previously been shown to take the form of optimal transport problems, in particular the Energy Mover's Distance (EMD). In this work, we show that the EMD is in fact the natural structure for comparing collider events, which accounts for its recent success in understanding event and jet substructure. We then present a Shape Hunting Algorithm using Parameterized Energy Reconstruction (SHAPER), which is a general framework for defining and computing shape-based observables. SHAPER generalizes N-jettiness from point clusters to any extended, parametrizable shape. This is accomplished by efficiently minimizing the EMD between events and parameterized manifolds of energy flows representing idealized shapes, implemented using the dual-potential Sinkhorn approximation of the Wasserstein metric. We show how the geometric language of observables as manifolds can be used to define novel observables with built-in infrared-and-collinear safety. We demonstrate the efficacy of the SHAPER framework by performing empirical jet substructure studies using several examples of new shape-based observables.
- Open Access
- 29 June 2023
- Published by Springer Nature in Journal of High Energy Physics
- Vol. 2023 (6) , 1-58
SHAPER: can you hear the shape of a jet?
The identification of interesting substructures within jets is an important tool for searching for new physics and probing the Standard Model at colliders. Many of these substructure tools have previously been shown to take the form of optimal transport problems, in particular the Energy Mover’s Distance (EMD). In this work, we show that the EMD is in fact the natural structure for comparing collider events, which accounts for its recent success in understanding event and jet substructure. We then present a Shape Hunting Algorithm using Parameterized Energy Reconstruction (Shaper), which is a general framework for defining and computing shape-based observables. Shaper generalizes N-jettiness from point clusters to any extended, parametrizable shape. This is accomplished by efficiently minimizing the EMD between events and parameterized manifolds of energy flows representing idealized shapes, implemented using the dual-potential Sinkhorn approximation of the Wasserstein metric. We show how the geometric language of observables as manifolds can be used to define novel observables with built-in infrared-and-collinear safety. We demonstrate the efficacy of the Shaper framework by performing empirical jet substructure studies using several examples of new shape-based observables.
- 29 June 2023
- Published by Springer Nature in Journal of High Energy Physics
- Vol. 2023 (6) , 1-58
Open AccessSHAPER: can you hear the shape of a jet?
The identification of interesting substructures within jets is an important tool for searching for new physics and probing the Standard Model at colliders. Many of these substructure tools have previously been shown to take the form of optimal transport problems, in particular the Energy Mover’s Distance (EMD). In this work, we show that the EMD is in fact the natural structure for comparing collider events, which accounts for its recent success in understanding event and jet substructure. We then present a Shape Hunting Algorithm using Parameterized Energy Reconstruction (Shaper), which is a general framework for defining and computing shape-based observables. Shaper generalizes N-jettiness from point clusters to any extended, parametrizable shape. This is accomplished by efficiently minimizing the EMD between events and parameterized manifolds of energy flows representing idealized shapes, implemented using the dual-potential Sinkhorn approximation of the Wasserstein metric. We show how the geometric language of observables as manifolds can be used to define novel observables with built-in infrared-and-collinear safety. We demonstrate the efficacy of the Shaper framework by performing empirical jet substructure studies using several examples of new shape-based observables.