Publications

All organisms make temporal predictions, and their evolutionary fitness level depends on the accuracy of these predictions. In the context of visual perception, the motions of both the observer and objects in the scene structure the dynamics of sensory signals, allowing for partial prediction of future signals based on past ones. Here, we propose a self-supervised representation-learning framework that extracts and exploits the regularities of natural videos to compute accurate predictions. We motivate the polar architecture by appealing to the Fourier shift theorem and its group-theoretic generalization, and we optimize its parameters on next-frame prediction. Through controlled experiments, we demonstrate that this approach can discover the representation of simple transformation groups acting in data. When trained on natural video datasets, our framework achieves better prediction performance than traditional motion compensation and rivals conventional deep networks, while maintaining interpretability and speed. Furthermore, the polar computations can be restructured into components resembling normalized simple and direction-selective complex cell models of primate V1 neurons. Thus, polar prediction offers a principled framework for understanding how the visual system represents sensory inputs in a form that simplifies temporal prediction.

The efficient coding hypothesis posits that sensory systems are adapted to the statistics of their inputs, maximizing mutual information between environmental signals and their representations, subject to biological constraints. While elegant, information theoretic quantities are notoriously difficult to measure or optimize, and most research on the hypothesis employs approximations, bounds, or substitutes (e.g., reconstruction error). A recently developed measure of coding efficiency, the "manifold capacity", quantifies the number of object categories that may be represented in a linearly separable fashion, but its calculation relies on a computationally intensive iterative procedure that precludes its use as an objective. Here, we simplify this measure to a form that facilitates direct optimization, use it to learn Maximum Manifold Capacity Representations (MMCRs), and demonstrate that these are competitive with state-of-the-art results on current self-supervised learning (SSL) recognition benchmarks. Empirical analyses reveal important differences between MMCRs and the representations learned by other SSL frameworks, and suggest a mechanism by which manifold compression gives rise to class separability. Finally, we evaluate a set of SSL methods on a suite of neural predictivity benchmarks, and find MMCRs are highly competitive as models of the primate ventral stream.

Internal representations are not uniquely identifiable from perceptual measurements: different representations can generate identical perceptual predictions, and similar representations may predict dissimilar percepts. Here, we generalize a previous method (``Eigendistortions'' -- Berardino et al., 2017) to enable comparison of models based on their metric tensors, which can be verified perceptually. Metric tensors characterize sensitivity to stimulus perturbations, reflecting both the geometric and stochastic properties of the representation, and providing an explicit prediction of perceptual discriminability. Brute force comparison of model-predicted metric tensors would require estimation of human perceptual thresholds along an infeasibly large set of stimulus directions. To circumvent this ``perceptual curse of dimensionality'', we compute and measure discrimination capabilities for a small set of most-informative perturbations, reducing the measurement cost from thousands of hours (a conservative estimate) to a single trial. We show that this single measurement, made for a variety of different test stimuli, is sufficient to differentiate models, select models that better match human perception, or generate new models that combine the advantages of existing models. We demonstrate the power of this method in comparison of (1) two models for trichromatic color representation, with differing internal noise; and (2) two autoencoders trained with different regularizers.

Sensory-guided behavior requires reliable encoding of stimulus information in neural populations, and flexible, task-specific readout. The former has been studied extensively, but the latter remains poorly understood. We introduce a theory for adaptive sensory processing based on functionally-targeted stochastic modulation. We show that responses of neurons in area V1 of monkeys performing a visual discrimination task exhibit low-dimensional, rapidly fluctuating gain modulation, which is stronger in task-informative neurons and can be used to decode from neural activity after few training trials, consistent with observed behavior. In a simulated hierarchical neural network model, such labels are learned quickly and can be used to adapt downstream readout, even after several intervening processing stages. Consistently, we find the modulatory signal estimated in V1 is also present in the activity of simultaneously recorded MT units, and is again strongest in task-informative neurons. These results support the idea that co-modulation facilitates task-adaptive hierarchical information routing.

The decision to stop growing and mature into an adult is a critical point in development that determines adult body size, impacting multiple aspects of an adult’s biology. In many animals, growth cessation is a consequence of hormone release that appears to be tied to the attainment of a particular body size or condition. Nevertheless, the size-sensing mechanism animals use to initiate hormone synthesis is poorly understood. Here, we develop a simple mathematical model of growth cessation in Drosophila melanogaster, which is ostensibly triggered by the attainment of a critical weight (CW) early in the last instar. Attainment of CW is correlated with the synthesis of the steroid hormone ecdysone, which causes a larva to stop growing, pupate, and metamorphose into the adult form. Our model suggests that, contrary to expectation, the size-sensing mechanism that initiates metamorphosis occurs before the larva reaches CW; that is, the critical-weight phenomenon is a downstream consequence of an earlier size-dependent developmental decision, not a decision point itself. Further, this size-sensing mechanism does not require a direct assessment of body size but emerges from the interactions between body size, ecdysone, and nutritional signaling. Because many aspects of our model are evolutionarily conserved among all animals, the model may provide a general framework for understanding how animals commit to maturing from their juvenile to adult form.

During vertebrate organ morphogenesis, large collectives of cells robustly self-organize to form architectural units (bones, villi, follicles) whose form persists into adulthood. Over the past few decades, mechanisms of organ morphogenesis have been developed predominantly through molecular, genetic, and cellular frameworks. More recently, there has been a resurgence of interest in collective cell and tissue mechanics during organ formation. This approach has amplified the need to clarify and unambiguously link events across biological length scales. Doing so may require reassessing canonical models that continue to guide the field. The most recognized model for organ formation centers around morphogens as determinants of gene expression and morphological patterns. The classical view of a morphogen is that morphogen gradients specify differential gene expression in a distinct spatial order. Because morphogen expression colocalizes with emerging feather and hair follicles, the skin has served as a paradigmatic example of such morphogen prepatterning mechanisms.

Molecular electronics break-junction experiments are widely used to investigate fundamental physics and chemistry at the nanoscale. Reproducibility in these experiments relies on measuring conductance on thousands of freshly formed molecular junctions, yielding a broad histogram of conductance events. Experiments typically focus on the most probable conductance, while the information content of the conductance histogram has remained unclear. Here we develop a microscopic theory for the conductance histogram by merging the theory of force-spectroscopy with molecular conductance. The procedure yields analytical equations that accurately fit the conductance histogram of a wide range of molecular junctions and augments the information content that can be extracted from them. Our formulation captures contributions to the conductance dispersion due to conductance changes during the mechanical elongation inherent to the experiments. In turn, the histogram shape is determined by the non-equilibrium stochastic features of junction rupture and formation. The microscopic parameters in the theory capture the junction’s electromechanical properties and can be isolated from separate conductance and rupture force (or junction-lifetime) measurements. The predicted behavior can be used to test the range of validity of the theory, understand the conductance histograms, design molecular junction experiments with enhanced resolution and molecular devices with more reproducible conductance properties.

We study bacterial diffusion in disordered porous media. Interactions with obstacles, at unknown locations, make this problem challenging. We approach it by abstracting the environment to cell states with memoryless transitions. With this, we derive an effective diffusivity that agrees well with simulations in explicit geometries. The diffusivity is non-monotonic, and we solve the optimal run length. We also find a rescaling that causes all of the theory and simulations to collapse. Our results indicate that a small set of microscopic features captures bacterial diffusion in disordered media.

This paper is associated with a poster winner of a 2022 American Physical Society's Division of Fluid Dynamics (DFD) Milton van Dyke Award for work presented at the DFD Gallery of Fluid Motion. The original poster is available online at the Gallery of Fluid Motion.

A growing list of diverse biological systems and their equally diverse functionalities provides realizations of a paradigm of emergent behavior. In each of these biological systems, pervasive ensembles of weak, short-lived, spatially local interactions act autonomously to convey functionalities at larger spatial and temporal scales. In this article, a range of diverse systems and functionalities are presented in a cursory manner with literature citations for further details. Then two systems and their properties are discussed in more detail: yeast chromosome biology and human respiratory mucus