Metric learning seeks a transformation of the feature space that enhances prediction quality for a given task. In this work we provide PAC-style sample complexity rates for supervised metric learning. We give matching lower- and upper-bounds showing that sample complexity scales with the representation dimension when no assumptions are made about the underlying data distribution. In addition, by leveraging the structure of the data distribution, we provide rates fine-tuned to a specific notion of the intrinsic complexity of a given dataset, allowing us to relax the dependence on representation dimension. We show both theoretically and empirically that augmenting the metric learning optimization criterion with a simple norm-based regularization is important and can help adapt to a dataset’s intrinsic complexity yielding better generalization, thus partly explaining the empirical success of similar regularizations reported in previous works.

Animals rely on visual motion for navigating the world, and research in flies has clarified how neural circuits extract information from moving visual scenes. However, the major pathways connecting these patterns of optic flow to behavior remain poorly understood. Using a high-throughput quantitative assay of visually guided behaviors and genetic neuronal silencing, we discovered a region in Drosophila’s protocerebrum critical for visual motion following. We used neuronal silencing, calcium imaging, and optogenetics to identify a single cell type, LPC1, that innervates this region, detects translational optic flow, and plays a key role in regulating forward walking. Moreover, the population of LPC1s can estimate the travelling direction, such as when gaze direction diverges from body heading. By linking specific cell types and their visual computations to specific behaviors, our findings establish a foundation for understanding how the nervous system uses vision to guide navigation.

Animals are often bombarded with visual information and must prioritize specific visual features based on their current needs. The neuronal circuits that detect and relay visual features have been well studied. Much less is known about how an animal adjusts its visual attention as its goals or environmental conditions change. During social behaviours, flies need to focus on nearby flies. Here we study how the flow of visual information is altered when female Drosophila enter an aggressive state. From the connectome, we identify three state-dependent circuit motifs poised to modify the response of an aggressive female to fly-sized visual objects: convergence of excitatory inputs from neurons conveying select visual features and internal state; dendritic disinhibition of select visual feature detectors; and a switch that toggles between two visual feature detectors. Using cell-type-specific genetic tools, together with behavioural and neurophysiological analyses, we show that each of these circuit motifs is used during female aggression. We reveal that features of this same switch operate in male Drosophila during courtship pursuit, suggesting that disparate social behaviours may share circuit mechanisms. Our study provides a compelling example of using the connectome to infer circuit mechanisms that underlie dynamic processing of sensory signals.

New work on innate escape behavior shows that mice spontaneously form a spatially precise memory of the location of shelter, which is laid down quickly and updated continuously.

An approaching predator and self-motion toward an object can generate similar looming patterns on the retina, but these situations demand different rapid responses. How central circuits flexibly process visual cues to activate appropriate, fast motor pathways remains unclear. Here we identify two descending neuron (DN) types that control landing and contribute to visuomotor flexibility in Drosophila. For each, silencing impairs visually evoked landing, activation drives landing, and spike rate determines leg extension amplitude. Critically, visual responses of both DNs are severely attenuated during non-flight periods, effectively decoupling visual stimuli from the landing motor pathway when landing is inappropriate. The flight-dependence mechanism differs between DN types. Octopamine exposure mimics flight effects in one, whereas the other probably receives neuronal feedback from flight motor circuits. Thus, this sensorimotor flexibility arises from distinct mechanisms for gating action-specific descending pathways, such that sensory and motor networks are coupled or decoupled according to the behavioral state.

In the fruit fly, Drosophila melanogaster, connectome data and genetic tools provide a unique opportunity to study complex behaviors including navigation, mating, aggression, and grooming in an organism with a tractable nervous system of 140,000 neurons. Here we present the Fly Disco, a flexible system for high quality video collection, optogenetic manipulation, and fine-grained behavioral analysis of freely walking and socializing fruit fly groups. The data collection hardware and software automates the collection of videos synced to programmable optogenetic stimuli. Key pipeline features include behavioral analysis based on trajectories of 21 keypoints and optogenetic-specific summary statistics and data visualization. We created the multifly dataset for pose estimation that includes 9701 examples enriched in complex behaviors. All hardware designs, software, and the multifly dataset are freely available.

We present a particle filtering algorithm for robustly tracking the contours of multiple deformable objects through severe occlusions. Our algorithm combines a multiple blob tracker with a contour tracker in a manner that keeps the required number of samples small. This is a natural combination because both algorithms have complementary strengths. The multiple blob tracker uses a natural multi-target model and searches a smaller and simpler space. On the other hand, contour tracking gives more fine-tuned results and relies on cues that are available during severe occlusions. Our choice of combination of these two algorithms accentuates the advantages of each. We demonstrate good performance on challenging video of three identical mice that contains multiple instances of severe occlusion.

Modern supervised learning algorithms can learn very accurate and complex discriminating functions. But when these classifiers fail, this complexity can also be a drawback because there is no easy, intuitive way to diagnose why they are failing and remedy the problem. This important question has received little attention. To address this problem, we propose a novel method to analyze and understand a classifier's errors. Our method centers around a measure of how much influence a training example has on the classifier's prediction for a test example. To understand why a classifier is mispredicting the label of a given test example, the user can find and review the most influential training examples that caused this misprediction, allowing them to focus their attention on relevant areas of the data space. This will aid the user in determining if and how the training data is inconsistently labeled or lacking in diversity, or if the feature representation is insufficient. As computing the influence of each training example is computationally impractical, we propose a novel distance metric to approximate influence for boosting classifiers that is fast enough to be used interactively. We also show several novel use paradigms of our distance metric. Through experiments, we show that it can be used to find incorrectly or inconsistently labeled training examples, to find specific areas of the data space that need more training data, and to gain insight into which features are missing from the current representation. 

Code is available at https://github.com/kristinbranson/InfluentialNeighbors.

The body of an animal influences how the nervous system produces behavior. Therefore, detailed modeling of the neural control of sensorimotor behavior requires a detailed model of the body. Here we contribute an anatomically-detailed biomechanical whole-body model of the fruit fly Drosophila melanogaster in the MuJoCo physics engine. Our model is general-purpose, enabling the simulation of diverse fly behaviors, both on land and in the air. We demonstrate the generality of our model by simulating realistic locomotion, both flight and walking. To support these behaviors, we have extended MuJoCo with phenomenological models of fluid forces and adhesion forces. Through data-driven end-to-end reinforcement learning, we demonstrate that these advances enable the training of neural network controllers capable of realistic locomotion along complex trajectories based on high-level steering control signals. We demonstrate the use of visual sensors and the re-use of a pre-trained general-purpose flight controller by training the model to perform visually guided flight tasks. Our project is an open-source platform for modeling neural control of sensorimotor behavior in an embodied context.Competing Interest StatementThe authors have declared no competing interest.