Search for: All records

Although deep Reinforcement Learning (RL) has proven successful in a wide range of tasks, one challenge it faces is interpretability when applied to real-world problems. Saliency maps are frequently used to provide interpretability for deep neural networks. However, in the RL domain, existing saliency map approaches are either computationally expensive and thus cannot satisfy the real-time requirement of real-world scenarios or cannot produce interpretable saliency maps for RL policies. In this work, we propose an approach of Distillation with selective Input Gradient Regularization (DIGR) which uses policy distillation and input gradient regularization to produce new policies that achieve both high interpretability and computation efficiency in generating saliency maps. Our approach is also found to improve the robustness of RL policies to multiple adversarial attacks. We conduct experiments on three tasks, MiniGrid (Fetch Object), Atari (Breakout) and CARLA Autonomous Driving, to demonstrate the importance and effectiveness of our approach. 

Humans and other animals have a remarkable capacity to translate their position from one spatial frame of reference to another. The ability to seamlessly move between top-down and first-person views is important for navigation, memory formation, and other cognitive tasks. Evidence suggests that the medial temporal lobe and other cortical regions contribute to this function. To understand how a neural system might carry out these computations, we used variational autoencoders (VAEs) to reconstruct the first-person view from the top-down view of a robot simulation, and vice versa. Many latent variables in the VAEs had similar responses to those seen in neuron recordings, including location-specific activity, head direction tuning, and encoding of distance to local objects. Place-specific responses were prominent when reconstructing a first-person view from a top-down view, but head direction–specific responses were prominent when reconstructing a top-down view from a first-person view. In both cases, the model could recover from perturbations without retraining, but rather through remapping. These results could advance our understanding of how brain regions support viewpoint linkages and transformations. 

In their book “How the Body Shapes the Way We Think: A New View of Intelligence,” Pfeifer and Bongard put forth an embodied approach to cognition. Because of this position, many of their robot examples demonstrated “intelligent” behavior despite limited neural processing. It is our belief that neurorobots should attempt to follow many of these principles. In this article, we discuss a number of principles to consider when designing neurorobots and experiments using robots to test brain theories. These principles are strongly inspired by Pfeifer and Bongard, but build on their design principles by grounding them in neuroscience and by adding principles based on neuroscience research. Our design principles fall into three categories. First, organisms must react quickly and appropriately to events. Second, organisms must have the ability to learn and remember over their lifetimes. Third, organisms must weigh options that are crucial for survival. We believe that by following these design principles a robot's behavior will be more naturalistic and more successful. 

Neuromorphic computing systems execute machine learning tasks designed with spiking neural networks. These systems are embracing non-volatile memory to implement high-density and low-energy synaptic storage. Elevated voltages and currents needed to operate non-volatile memories cause aging of CMOS-based transistors in each neuron and synapse circuit in the hardware, drifting the transistor’s parameters from their nominal values. If these circuits are used continuously for too long, the parameter drifts cannot be reversed, resulting in permanent degradation of circuit performance over time, eventually leading to hardware faults. Aggressive device scaling increases power density and temperature, which further accelerates the aging, challenging the reliable operation of neuromorphic systems. Existing reliability-oriented techniques periodically de-stress all neuron and synapse circuits in the hardware at fixed intervals, assuming worst-case operating conditions, without actually tracking their aging at run-time. To de-stress these circuits, normal operation must be interrupted, which introduces latency in spike generation and propagation, impacting the inter-spike interval and hence, performance (e.g., accuracy). We observe that in contrast to long-term aging, which permanently damages the hardware, short-term aging in scaled CMOS transistors is mostly due to bias temperature instability. The latter is heavily workload-dependent and, more importantly, partially reversible. We propose a new architectural technique to mitigate the aging-related reliability problems in neuromorphic systems by designing an intelligent run-time manager (NCRTM), which dynamically de-stresses neuron and synapse circuits in response to the short-term aging in their CMOS transistors during the execution of machine learning workloads, with the objective of meeting a reliability target. NCRTM de-stresses these circuits only when it is absolutely necessary to do so, otherwise reducing the performance impact by scheduling de-stress operations off the critical path. We evaluate NCRTM with state-of-the-art machine learning workloads on a neuromorphic hardware. Our results demonstrate that NCRTM significantly improves the reliability of neuromorphic hardware, with marginal impact on performance. 

Individuals vary in the way they navigate through space. Some take novel shortcuts, while others rely on known routes to find their way around. We wondered how and why there is so much variation in the population. To address this, we first compared the trajectories of 368 human subjects navigating a virtual maze with simulated trajectories. The simulated trajectories were generated by strategy‐based path planning algorithms from robotics. Based on the similarities between human trajectories and different strategy‐based simulated trajectories, we found that there is a variation in the type of strategy individuals apply to navigate space, as well as variation within individuals on a trial‐by‐trial basis. Moreover, we observed variation within a trial when subjects occasionally switched the navigation strategies halfway through a trajectory. In these cases, subjects started with a route strategy, in which they followed a familiar path, and then switched to a survey strategy, in which they took shortcuts by considering the layout of the environment. Then we simulated a second set of trajectories using five different but comparable artificial maps. These trajectories produced the similar pattern of strategy variation within and between trials. Furthermore, we varied the relative cost, that is, the assumed mental effort or required timesteps to choose a learned route over alternative paths. When the learned route was relatively costly, the simulated agents tended to take shortcuts. Conversely, when the learned route was less costly, the simulated agents showed preference toward a route strategy. We suggest that cost or assumed mental effort may be the reason why in previous studies, subjects used survey knowledge when instructed to take the shortest path. We suggest that this variation we observe in humans may be beneficial for robotic swarms or collections of autonomous agents during information gathering.

 

Disentangling the sources of visual motion in a dynamic scene during self-movement or ego motion is important for autonomous navigation and tracking. In the dynamic image segments of a video frame containing independently moving objects, optic flow relative to the next frame is the sum of the motion fields generated due to camera and object motion. The traditional ego-motion estimation methods assume the scene to be static, and the recent deep learning-based methods do not separate pixel velocities into object- and ego-motion components. We propose a learning-based approach to predict both ego-motion parameters and object-motion field (OMF) from image sequences using a convolutional autoencoder while being robust to variations due to the unconstrained scene depth. This is achieved by: 1) training with continuous ego-motion constraints that allow solving for ego-motion parameters independently of depth and 2) learning a sparsely activated overcomplete ego-motion field (EMF) basis set, which eliminates the irrelevant components in both static and dynamic segments for the task of ego-motion estimation. In order to learn the EMF basis set, we propose a new differentiable sparsity penalty function that approximates the number of nonzero activations in the bottleneck layer of the autoencoder and enforces sparsity more effectively than L1- and L2-norm-based penalties. Unlike the existing direct ego-motion estimation methods, the predicted global EMF can be used to extract OMF directly by comparing it against the optic flow. Compared with the state-of-the-art baselines, the proposed model performs favorably on pixelwise object- and ego-motion estimation tasks when evaluated on real and synthetic data sets of dynamic scenes.