Optimization in Visual Motion Estimation.

Sighted animals use visual signals to discern directional motion in their environment. Motion is not directly detected by visual neurons, and it must instead be computed from light signals that vary over space and time. This makes visual motion estimation a near universal neural computation, and decades of research have revealed much about the algorithms and mechanisms that generate directional signals. The idea that sensory systems are optimized for performance in natural environments has deeply impacted this research. In this article, we review the many ways that optimization has been used to quantitatively model visual motion estimation and reveal its underlying principles. We emphasize that no single optimization theory has dominated the literature. Instead, researchers have adeptly incorporated different computational demands and biological constraints that are pertinent to the specific brain system and animal model under study. The successes and failures of the resulting optimization models have thereby provided insights into how computational demands and biological constraints together shape neural computation.

Reward expectations direct learning and drive operant matching in Drosophila.

Foraging animals must use decision-making strategies that dynamically adapt to the changing availability of rewards in the environment. A wide diversity of animals do this by distributing their choices in proportion to the rewards received from each option, Herrnstein's operant matching law. Theoretical work suggests an elegant mechanistic explanation for this ubiquitous behavior, as operant matching follows automatically from simple synaptic plasticity rules acting within behaviorally relevant neural circuits. However, no past work has mapped operant matching onto plasticity mechanisms in the brain, leaving the biological relevance of the theory unclear. Here, we discovered operant matching in Drosophila and showed that it requires synaptic plasticity that acts in the mushroom body and incorporates the expectation of reward. We began by developing a dynamic foraging paradigm to measure choices from individual flies as they learn to associate odor cues with probabilistic rewards. We then built a model of the fly mushroom body to explain each fly's sequential choice behavior using a family of biologically realistic synaptic plasticity rules. As predicted by past theoretical work, we found that synaptic plasticity rules could explain fly matching behavior by incorporating stimulus expectations, reward expectations, or both. However, by optogenetically bypassing the representation of reward expectation, we abolished matching behavior and showed that the plasticity rule must specifically incorporate reward expectations. Altogether, these results reveal the first synapse-level mechanisms of operant matching and provide compelling evidence for the role of reward expectation signals in the fly brain.

Organizing memories for generalization in complementary learning systems.

Memorization and generalization are complementary cognitive processes that jointly promote adaptive behavior. For example, animals should memorize safe routes to specific water sources and generalize from these memories to discover environmental features that predict new ones. These functions depend on systems consolidation mechanisms that construct neocortical memory traces from hippocampal precursors, but why systems consolidation only applies to a subset of hippocampal memories is unclear. Here we introduce a new neural network formalization of systems consolidation that reveals an overlooked tension-unregulated neocortical memory transfer can cause overfitting and harm generalization in an unpredictable world. We resolve this tension by postulating that memories only consolidate when it aids generalization. This framework accounts for partial hippocampal-cortical memory transfer and provides a normative principle for reconceptualizing numerous observations in the field. Generalization-optimized systems consolidation thus provides new insight into how adaptive behavior benefits from complementary learning systems specialized for memorization and generalization.

Surface viscosities of lipid bilayers determined from equilibrium molecular dynamics simulations.

Lipid membrane viscosity is critical to biological function. Bacterial cells grown in different environments alter their lipid composition in order to maintain a specific viscosity, and membrane viscosity has been linked to the rate of cellular respiration. To understand the factors that determine the viscosity of a membrane, we ran equilibrium all-atom simulations of single component lipid bilayers and calculated their viscosities. The viscosity was calculated via a Green-Kubo relation, with the stress-tensor autocorrelation function modeled by a stretched exponential function. By simulating a series of lipids at different temperatures, we establish the dependence of viscosity on several aspects of lipid chemistry, including hydrocarbon chain length, unsaturation, and backbone structure. Sphingomyelin is found to have a remarkably high viscosity, roughly 20 times that of DPPC. Furthermore, we find that inclusion of the entire range of the dispersion interaction increases viscosity by up to 140%. The simulated viscosities are similar to experimental values obtained from the rotational dynamics of small chromophores and from the diffusion of integral membrane proteins but significantly lower than recent measurements based on the deformation of giant vesicles.

Exact learning dynamics of deep linear networks with prior knowledge.

Learning in deep neural networks is known to depend critically on the knowledge embedded in the initial network weights. However, few theoretical results have precisely linked prior knowledge to learning dynamics. Here we derive exact solutions to the dynamics of learning with rich prior knowledge in deep linear networks by generalising Fukumizu's matrix Riccati solution (Fukumizu 1998 Gen 1 1E-03). We obtain explicit expressions for the evolving network function, hidden representational similarity, and neural tangent kernel over training for a broad class of initialisations and tasks. The expressions reveal a class of task-independent initialisations that radically alter learning dynamics from slow non-linear dynamics to fast exponential trajectories while converging to a global optimum with identical representational similarity, dissociating learning trajectories from the structure of initial internal representations. We characterise how network weights dynamically align with task structure, rigorously justifying why previous solutions successfully described learning from small initial weights without incorporating their fine-scale structure. Finally, we discuss the implications of these findings for continual learning, reversal learning and learning of structured knowledge. Taken together, our results provide a mathematical toolkit for understanding the impact of prior knowledge on deep learning.

A brainstem integrator for self-location memory and positional homeostasis in zebrafish.

To track and control self-location, animals integrate their movements through space. Representations of self-location are observed in the mammalian hippocampal formation, but it is unknown if positional representations exist in more ancient brain regions, how they arise from integrated self-motion, and by what pathways they control locomotion. Here, in a head-fixed, fictive-swimming, virtual-reality preparation, we exposed larval zebrafish to a variety of involuntary displacements. They tracked these displacements and, many seconds later, moved toward their earlier location through corrective swimming ("positional homeostasis"). Whole-brain functional imaging revealed a network in the medulla that stores a memory of location and induces an error signal in the inferior olive to drive future corrective swimming. Optogenetically manipulating medullary integrator cells evoked displacement-memory behavior. Ablating them, or downstream olivary neurons, abolished displacement corrections. These results reveal a multiregional hindbrain circuit in vertebrates that integrates self-motion and stores self-location to control locomotor behavior.

Geometric framework to predict structure from function in neural networks.

Neural computation in biological and artificial networks relies on the nonlinear summation of many inputs. The structural connectivity matrix of synaptic weights between neurons is a critical determinant of overall network function, but quantitative links between neural network structure and function are complex and subtle. For example, many networks can give rise to similar functional responses, and the same network can function differently depending on context. Whether certain patterns of synaptic connectivity are required to generate specific network-level computations is largely unknown. Here we introduce a geometric framework for identifying synaptic connections required by steady-state responses in recurrent networks of threshold-linear neurons. Assuming that the number of specified response patterns does not exceed the number of input synapses, we analytically calculate the solution space of all feedforward and recurrent connectivity matrices that can generate the specified responses from the network inputs. A generalization accounting for noise further reveals that the solution space geometry can undergo topological transitions as the allowed error increases, which could provide insight into both neuroscience and machine learning. We ultimately use this geometric characterization to derive certainty conditions guaranteeing a nonzero synapse between neurons. Our theoretical framework could thus be applied to neural activity data to make rigorous anatomical predictions that follow generally from the model architecture.

Theoretical principles for illuminating sensorimotor processing with brain-wide neuronal recordings.

Modern recording techniques now permit brain-wide sensorimotor circuits to be observed at single neuron resolution in small animals. Extracting theoretical understanding from these recordings requires principles that organize findings and guide future experiments. Here we review theoretical principles that shed light onto brain-wide sensorimotor processing. We begin with an analogy that conceptualizes principles as streetlamps that illuminate the empirical terrain, and we illustrate the analogy by showing how two familiar principles apply in new ways to brain-wide phenomena. We then focus the bulk of the review on describing three more principles that have wide utility for mapping brain-wide neural activity, making testable predictions from highly parameterized mechanistic models, and investigating the computational determinants of neuronal response patterns across the brain.

A Neural Representation of Naturalistic Motion-Guided Behavior in the Zebrafish Brain.

All animals must transform ambiguous sensory data into successful behavior. This requires sensory representations that accurately reflect the statistics of natural stimuli and behavior. Multiple studies show that visual motion processing is tuned for accuracy under naturalistic conditions, but the sensorimotor circuits extracting these cues and implementing motion-guided behavior remain unclear. Here we show that the larval zebrafish retina extracts a diversity of naturalistic motion cues, and the retinorecipient pretectum organizes these cues around the elements of behavior. We find that higher-order motion stimuli, gliders, induce optomotor behavior matching expectations from natural scene analyses. We then image activity of retinal ganglion cell terminals and pretectal neurons. The retina exhibits direction-selective responses across glider stimuli, and anatomically clustered pretectal neurons respond with magnitudes matching behavior. Peripheral computations thus reflect natural input statistics, whereas central brain activity precisely codes information needed for behavior. This general principle could organize sensorimotor transformations across animal species.

Correcting for physical distortions in visual stimuli improves reproducibility in zebrafish neuroscience.

Optical refraction causes light to bend at interfaces between optical media. This phenomenon can significantly distort visual stimuli presented to aquatic animals in water, yet refraction has often been ignored in the design and interpretation of visual neuroscience experiments. Here we provide a computational tool that transforms between projected and received stimuli in order to detect and control these distortions. The tool considers the most commonly encountered interface geometry, and we show that this and other common configurations produce stereotyped distortions. By correcting these distortions, we reduced discrepancies in the literature concerning stimuli that evoke escape behavior, and we expect this tool will help reconcile other confusing aspects of the literature. This tool also aids experimental design, and we illustrate the dangers that uncorrected stimuli pose to receptive field mapping experiments.

Asymmetric ON-OFF processing of visual motion cancels variability induced by the structure of natural scenes.

Animals detect motion using a variety of visual cues that reflect regularities in the natural world. Experiments in animals across phyla have shown that motion percepts incorporate both pairwise and triplet spatiotemporal correlations that could theoretically benefit motion computation. However, it remains unclear how visual systems assemble these cues to build accurate motion estimates. Here, we used systematic behavioral measurements of fruit fly motion perception to show how flies combine local pairwise and triplet correlations to reduce variability in motion estimates across natural scenes. By generating synthetic images with statistics controlled by maximum entropy distributions, we show that the triplet correlations are useful only when images have light-dark asymmetries that mimic natural ones. This suggests that asymmetric ON-OFF processing is tuned to the particular statistics of natural scenes. Since all animals encounter the world's light-dark asymmetries, many visual systems are likely to use asymmetric ON-OFF processing to improve motion estimation.

Threshold-Based Ordering of Sequential Actions during Drosophila Courtship.

Goal-directed animal behaviors are typically composed of sequences of motor actions whose order and timing are critical for a successful outcome. Although numerous theoretical models for sequential action generation have been proposed, few have been supported by the identification of control neurons sufficient to elicit a sequence. Here, we identify a pair of descending neurons that coordinate a stereotyped sequence of engagement actions during Drosophila melanogaster male courtship behavior. These actions are initiated sequentially but persist cumulatively, a feature not explained by existing models of sequential behaviors. We find evidence consistent with a ramp-to-threshold mechanism, in which increasing neuronal activity elicits each action independently at successively higher activity thresholds.

The Neuronal Basis of an Illusory Motion Percept Is Explained by Decorrelation of Parallel Motion Pathways.

Both vertebrates and invertebrates perceive illusory motion, known as "reverse-phi," in visual stimuli that contain sequential luminance increments and decrements. However, increment (ON) and decrement (OFF) signals are initially processed by separate visual neurons, and parallel elementary motion detectors downstream respond selectively to the motion of light or dark edges, often termed ON- and OFF-edges. It remains unknown how and where ON and OFF signals combine to generate reverse-phi motion signals. Here, we show that each of Drosophila's elementary motion detectors encodes motion by combining both ON and OFF signals. Their pattern of responses reflects combinations of increments and decrements that co-occur in natural motion, serving to decorrelate their outputs. These results suggest that the general principle of signal decorrelation drives the functional specialization of parallel motion detection channels, including their selectivity for moving light or dark edges.

Long-Term Consolidation of Ensemble Neural Plasticity Patterns in Hippocampal Area CA1.

Neural network remodeling underpins the ability to remember life experiences, but little is known about the long-term plasticity of neural populations. To study how the brain encodes episodic events, we used time-lapse two-photon microscopy and a fluorescent reporter of neural plasticity based on an enhanced form of the synaptic activity-responsive element (E-SARE) within the Arc promoter to track thousands of CA1 hippocampal pyramidal cells over weeks in mice that repeatedly encountered different environments. Each environment evokes characteristic patterns of ensemble neural plasticity, but with each encounter, the set of activated cells gradually evolves. After repeated exposures, the plasticity patterns evoked by an individual environment progressively stabilize. Compared with young adults, plasticity patterns in aged mice are less specific to individual environments and less stable across repeat experiences. Long-term consolidation of hippocampal plasticity patterns may support long-term memory formation, whereas weaker consolidation in aged subjects might reflect declining memory function.

From Whole-Brain Data to Functional Circuit Models: The Zebrafish Optomotor Response.

Detailed descriptions of brain-scale sensorimotor circuits underlying vertebrate behavior remain elusive. Recent advances in zebrafish neuroscience offer new opportunities to dissect such circuits via whole-brain imaging, behavioral analysis, functional perturbations, and network modeling. Here, we harness these tools to generate a brain-scale circuit model of the optomotor response, an orienting behavior evoked by visual motion. We show that such motion is processed by diverse neural response types distributed across multiple brain regions. To transform sensory input into action, these regions sequentially integrate eye- and direction-specific sensory streams, refine representations via interhemispheric inhibition, and demix locomotor instructions to independently drive turning and forward swimming. While experiments revealed many neural response types throughout the brain, modeling identified the dimensions of functional connectivity most critical for the behavior. We thus reveal how distributed neurons collaborate to generate behavior and illustrate a paradigm for distilling functional circuit models from whole-brain data.

Nonlinear circuits for naturalistic visual motion estimation.

Many animals use visual signals to estimate motion. Canonical models suppose that animals estimate motion by cross-correlating pairs of spatiotemporally separated visual signals, but recent experiments indicate that humans and flies perceive motion from higher-order correlations that signify motion in natural environments. Here we show how biologically plausible processing motifs in neural circuits could be tuned to extract this information. We emphasize how known aspects of Drosophila's visual circuitry could embody this tuning and predict fly behavior. We find that segregating motion signals into ON/OFF channels can enhance estimation accuracy by accounting for natural light/dark asymmetries. Furthermore, a diversity of inputs to motion detecting neurons can provide access to more complex higher-order correlations. Collectively, these results illustrate how non-canonical computations improve motion estimation with naturalistic inputs. This argues that the complexity of the fly's motion computations, implemented in its elaborate circuits, represents a valuable feature of its visual motion estimator.

Impermanence of dendritic spines in live adult CA1 hippocampus.

The mammalian hippocampus is crucial for episodic memory formation and transiently retains information for about 3-4 weeks in adult mice and longer in humans. Although neuroscientists widely believe that neural synapses are elemental sites of information storage, there has been no direct evidence that hippocampal synapses persist for time intervals commensurate with the duration of hippocampal-dependent memory. Here we tested the prediction that the lifetimes of hippocampal synapses match the longevity of hippocampal memory. By using time-lapse two-photon microendoscopy in the CA1 hippocampal area of live mice, we monitored the turnover dynamics of the pyramidal neurons' basal dendritic spines, postsynaptic structures whose turnover dynamics are thought to reflect those of excitatory synaptic connections. Strikingly, CA1 spine turnover dynamics differed sharply from those seen previously in the neocortex. Mathematical modelling revealed that the data best matched kinetic models with a single population of spines with a mean lifetime of approximately 1-2 weeks. This implies ∼100% turnover in ∼2-3 times this interval, a near full erasure of the synaptic connectivity pattern. Although N-methyl-d-aspartate (NMDA) receptor blockade stabilizes spines in the neocortex, in CA1 it transiently increased the rate of spine loss and thus lowered spine density. These results reveal that adult neocortical and hippocampal pyramidal neurons have divergent patterns of spine regulation and quantitatively support the idea that the transience of hippocampal-dependent memory directly reflects the turnover dynamics of hippocampal synapses.

Whole-brain activity mapping onto a zebrafish brain atlas.

In order to localize the neural circuits involved in generating behaviors, it is necessary to assign activity onto anatomical maps of the nervous system. Using brain registration across hundreds of larval zebrafish, we have built an expandable open-source atlas containing molecular labels and definitions of anatomical regions, the Z-Brain. Using this platform and immunohistochemical detection of phosphorylated extracellular signal­regulated kinase (ERK) as a readout of neural activity, we have developed a system to create and contextualize whole-brain maps of stimulus- and behavior-dependent neural activity. This mitogen-activated protein kinase (MAP)-mapping assay is technically simple, and data analysis is completely automated. Because MAP-mapping is performed on freely swimming fish, it is applicable to studies of nearly any stimulus or behavior. Here we demonstrate our high-throughput approach using pharmacological, visual and noxious stimuli, as well as hunting and feeding. The resultant maps outline hundreds of areas associated with behaviors.

Flies and humans share a motion estimation strategy that exploits natural scene statistics.

Sighted animals extract motion information from visual scenes by processing spatiotemporal patterns of light falling on the retina. The dominant models for motion estimation exploit intensity correlations only between pairs of points in space and time. Moving natural scenes, however, contain more complex correlations. We found that fly and human visual systems encode the combined direction and contrast polarity of moving edges using triple correlations that enhance motion estimation in natural environments. Both species extracted triple correlations with neural substrates tuned for light or dark edges, and sensitivity to specific triple correlations was retained even as light and dark edge motion signals were combined. Thus, both species separately process light and dark image contrasts to capture motion signatures that can improve estimation accuracy. This convergence argues that statistical structures in natural scenes have greatly affected visual processing, driving a common computational strategy over 500 million years of evolution.

Photon shot noise limits on optical detection of neuronal spikes and estimation of spike timing.

Optical approaches for tracking neural dynamics are of widespread interest, but a theoretical framework quantifying the physical limits of these techniques has been lacking. We formulate such a framework by using signal detection and estimation theory to obtain physical bounds on the detection of neural spikes and the estimation of their occurrence times as set by photon counting statistics (shot noise). These bounds are succinctly expressed via a discriminability index that depends on the kinetics of the optical indicator and the relative fluxes of signal and background photons. This approach facilitates quantitative evaluations of different indicators, detector technologies, and data analyses. Our treatment also provides optimal filtering techniques for optical detection of spikes. We compare various types of Ca(2+) indicators and show that background photons are a chief impediment to voltage sensing. Thus, voltage indicators that change color in response to membrane depolarization may offer a key advantage over those that change intensity. We also examine fluorescence resonance energy transfer indicators and identify the regimes in which the widely used ratiometric analysis of signals is substantially suboptimal. Overall, by showing how different optical factors interact to affect signal quality, our treatment offers a valuable guide to experimental design and provides measures of confidence to assess optically extracted traces of neural activity.