Reward function specification can be difficult, even in simple environments. Realistic environments contain millions of states. Rewarding the agent for making a widget may be easy, but penalizing the multitude of possible negative side effects is hard. In toy environments, Attainable Utility Preservation (AUP) avoids side effects by penalizing shifts in the ability to achieve randomly generated goals. We scale this approach to large, randomly generated environments based on Conway's Game of Life. By preserving optimal value for a single randomly generated reward function, AUP incurs modest overhead, completes the specified task, and avoids side effects.

It has long been recognized that multi-agent reinforcement learning (MARL) faces significant scalability issues due to the fact that the size of the state and action spaces are exponentially large in the number of agents. In this paper, we identify a rich class of networked MARL problems where the model exhibits a local dependence structure that allows it to be solved in a scalable manner. Specifically, we propose a Scalable Actor-Critic (SAC) method that can learn a near optimal localized policy for optimizing the average reward with complexity scaling with the state-action space size of local neighborhoods, as opposed to the entire network. Our result centers around identifying and exploiting an exponential decay property that ensures the effect of agents on each other decays exponentially fast in their graph distance.

In this paper, we consider the problem of actor-critic reinforcement learning. Firstly, we extend the actor-critic architecture to actor-critic-N architecture by introducing more critics beyond rewards. Secondly, we combine the reward-based critic with a potential-field-based critic to formulate the proposed potential field guided actor-critic reinforcement learning approach (actor-critic-2). This can be seen as a combination of the model-based gradients and the model-free gradients in policy improvement. State with large potential field often contains a strong prior information, such as pointing to the target at a long distance or avoiding collision by the side of an obstacle. In this situation, we should trust potential-field-based critic more as policy evaluation to accelerate policy improvement, where action policy tends to be guided. For example, in practical application, learning to avoid obstacles should be guided rather than learned by trial and error. State with small potential filed is often lack of information, for example, at the local minimum point or around the moving target. At this time, we should trust reward-based critic as policy evaluation more to evaluate the long-term return. In this case, action policy tends to explore. In addition, potential field evaluation can be combined with planning to estimate a better state value function. In this way, reward design can focus more on the final stage of reward, rather than reward shaping or phased reward. Furthermore, potential field evaluation can make up for the lack of communication in multi-agent cooperation problem, i.e., multi-agent each has a reward-based critic and a relative unified potential-field-based critic with prior information. Thirdly, simplified experiments on predator-prey game demonstrate the effectiveness of the proposed approach.

How the human vision system addresses the object identity-preserving recognition problem is largely unknown. Here, we use a vision recognition-reconstruction network (RRN) to investigate the development, recognition, learning and forgetting mechanisms, and achieve similar characteristics to electrophysiological measurements in monkeys. First, in network development study, the RRN also experiences critical developmental stages characterized by specificities in neuron types, synapse and activation patterns, and visual task performance from the early stage of coarse salience map recognition to mature stage of fine structure recognition. In digit recognition study, we witness that the RRN could maintain object invariance representation under various viewing conditions by coordinated adjustment of responses of population neurons. And such concerted population responses contained untangled object identity and properties information that could be accurately extracted via high-level cortices or even a simple weighted summation decoder. In the learning and forgetting study, novel structure recognition is implemented by adjusting entire synapses in low magnitude while pattern specificities of original synaptic connectivity are preserved, which guaranteed a learning process without disrupting the existing functionalities. This work benefits the understanding of the human visual processing mechanism and the development of human-like machine intelligence.

Acquiring multiple skills has commonly involved collecting a large number of expert demonstrations per task or engineering custom reward functions. Recently it has been shown that it is possible to acquire a diverse set of skills by self-supervising control on top of human teleoperated play data. Play is rich in state space coverage and a policy trained on this data can generalize to specific tasks at test time outperforming policies trained on individual expert task demonstrations. In this work, we explore the question of whether robots can learn to play to autonomously generate play data that can ultimately enhance performance. By training a behavioral cloning policy on a relatively small quantity of human play, we autonomously generate a large quantity of cloned play data that can be used as additional training. We demonstrate that a general purpose goal-conditioned policy trained on this augmented dataset substantially outperforms one trained only with the original human data on 18 difficult user-specified manipulation tasks in a simulated robotic tabletop environment. A video example of a robot imitating human play can be seen here: https://learning-to-play.github.io/videos/undirected_play1.mp4

Learning-enabled controllers used in cyber-physical systems (CPS) are known to be susceptible to adversarial attacks. Such attacks manifest as perturbations to the states generated by the controller's environment in response to its actions. We consider state perturbations that encompass a wide variety of adversarial attacks and describe an attack scheme for discovering adversarial states. To be useful, these attacks need to be natural, yielding states in which the controller can be reasonably expected to generate a meaningful response. We consider shield-based defenses as a means to improve controller robustness in the face of such perturbations. Our defense strategy allows us to treat the controller and environment as black-boxes with unknown dynamics. We provide a two-stage approach to construct this defense and show its effectiveness through a range of experiments on realistic continuous control domains such as the navigation control-loop of an F16 aircraft and the motion control system of humanoid robots.

Clustering is a ubiquitous task in data science. Compared to the commonly used $k$-means clustering algorithm, $k$-medoids clustering algorithms require the cluster centers to be actual data points and support arbitrary distance metrics, allowing for greater interpretability and the clustering of structured objects. Current state-of-the-art $k$-medoids clustering algorithms, such as Partitioning Around Medoids (PAM), are iterative and are quadratic in the dataset size $n$ for each iteration, being prohibitively expensive for large datasets. We propose Bandit-PAM, a randomized algorithm inspired by techniques from multi-armed bandits, that significantly improves the computational efficiency of PAM. We theoretically prove that Bandit-PAM reduces the complexity of each PAM iteration from $O(n^2)$ to $O(n \log n)$ and returns the same results with high probability, under assumptions on the data that often hold in practice. We empirically validate our results on several large-scale real-world datasets, including a coding exercise submissions dataset from Code.org, the 10x Genomics 68k PBMC single-cell RNA sequencing dataset, and the MNIST handwritten digits dataset. We observe that Bandit-PAM returns the same results as PAM while performing up to 200x fewer distance computations. The improvements demonstrated by Bandit-PAM enable $k$-medoids clustering on a wide range of applications, including identifying cell types in large-scale single-cell data and providing scalable feedback for students learning computer science online. We also release Python and C++ implementations of our algorithm.

We explore the problem of step-wise explaining how to solve constraint satisfaction problems, with a use case on logic grid puzzles. More specifically, we study the problem of explaining the inference steps that one can take during propagation, in a way that is easy to interpret for a person. Thereby, we aim to give the constraint solver explainable agency, which can help in building trust in the solver by being able to understand and even learn from the explanations. The main challenge is that of finding a sequence of simple explanations, where each explanation should aim to be as cognitively easy as possible for a human to verify and understand. This contrasts with the arbitrary combination of facts and constraints that the solver may use when propagating. We propose the use of a cost function to quantify how simple an individual explanation of an inference step is, and identify the explanation-production problem of finding the best sequence of explanations of a CSP. Our approach is agnostic of the underlying constraint propagation mechanisms, and can provide explanations even for inference steps resulting from combinations of constraints. In case multiple constraints are involved, we also develop a mechanism that allows to break the most difficult steps up and thus gives the user the ability to zoom in on specific parts of the explanation. Our proposed algorithm iteratively constructs the explanation sequence by using an optimistic estimate of the cost function to guide the search for the best explanation at each step. Our experiments on logic grid puzzles show the feasibility of the approach in terms of the quality of the individual explanations and the resulting explanation sequences obtained.

Reservoir computing (RC) is a machine learning algorithm that can learn complex time series from data very rapidly based on the use of high-dimensional dynamical systems, such as random networks of neurons, called "reservoirs." To implement RC in edge computing, it is highly important to reduce the amount of computational resources that RC requires. In this study, we propose methods that reduce the size of the reservoir by inputting the past or drifting states of the reservoir to the output layer at the current time step. These proposed methods are analyzed based on information processing capacity, which is a performance measure of RC proposed by Dambre et al. (2012). In addition, we evaluate the effectiveness of the proposed methods on time-series prediction tasks: the generalized Henon-map and NARMA. On these tasks, we found that the proposed methods were able to reduce the size of the reservoir up to one tenth without a substantial increase in regression error. Because the applications of the proposed methods are not limited to a specific network structure of the reservoir, the proposed methods could further improve the energy efficiency of RC-based systems, such as FPGAs and photonic systems.

We present simple differentially private estimators for the mean and covariance of multivariate sub-Gaussian data that are accurate at small sample sizes. We demonstrate the effectiveness of our algorithms both theoretically and empirically using synthetic and real-world datasets---showing that their asymptotic error rates match the state-of-the-art theoretical bounds, and that they concretely outperform all previous methods. Specifically, previous estimators either have weak empirical accuracy at small sample sizes, perform poorly for multivariate data, or require the user to provide strong a priori estimates for the parameters.