Safe control of neural network dynamic models (NNDMs) is important to robotics and many applications. However, it remains challenging to compute an optimal safe control in real time for NNDM. To enable real-time computation, we propose to use a sound approximation of the NNDM in the control synthesis. In particular, we propose Bernstein over-approximated neural dynamics (BOND) based on the Bernstein polynomial over-approximation (BPO) of ReLU activation functions in NNDM. To mitigate the errors introduced by the approximation and to ensure persistent feasibility of the safe control problems, we synthesize a worst-case safety index using the most unsafe approximated state within the BPO relaxation of NNDM offline. For the online real-time optimization, we formulate the first-order Taylor approximation of the nonlinear worst-case safety constraint as an additional linear layer of NNDM with the l2 bounded bias term for the higher-order remainder. Comprehensive experiments with different neural dynamics and safety constraints show that with safety guaranteed, our NNDMs with sound approximation are 10-100 times faster than the safe control baseline that uses mixed integer programming (MIP), validating the effectiveness of the worst-case safety index and scalability of the proposed BOND in real-time large-scale settings. The code is available at https://github.com/intelligent-control-lab/BOND.

This study examines the generalization ability of algorithm performance prediction models across various benchmark suites. Comparing the statistical similarity between the problem collections with the accuracy of performance prediction models that are based on exploratory landscape analysis features, we observe that there is a positive correlation between these two measures. Specifically, when the high-dimensional feature value distributions between training and testing suites lack statistical significance, the model tends to generalize well, in the sense that the testing errors are in the same range as the training errors. Two experiments validate these findings: one involving the standard benchmark suites, the BBOB and CEC collections, and another using five collections of affine combinations of BBOB problem instances.

In statistical mechanics, computing the partition function is generally difficult. An approximation method using a variational autoregressive network (VAN) has been proposed recently. This approach offers the advantage of directly calculating the generation probabilities while obtaining a significantly large number of samples. The present study introduces a novel approximation method that employs samples derived from quantum annealing machines in conjunction with VAN, which are empirically assumed to adhere to the Gibbs-Boltzmann distribution. When applied to the finite-size Sherrington-Kirkpatrick model, the proposed method demonstrates enhanced accuracy compared to the traditional VAN approach and other approximate methods, such as the widely utilized naive mean field.

From its inception, AI has had a rather ambivalent relationship with humans -- swinging between their augmentation and replacement. Now, as AI technologies enter our everyday lives at an ever increasing pace, there is a greater need for AI systems to work synergistically with humans. One critical requirement for such synergistic human-AI interaction is that the AI systems be explainable to the humans in the loop. To do this effectively, AI agents need to go beyond planning with their own models of the world, and take into account the mental model of the human in the loop. Drawing from several years of research in our lab, we will discuss how the AI agent can use these mental models to either conform to human expectations, or change those expectations through explanatory communication. While the main focus of the book is on cooperative scenarios, we will point out how the same mental models can be used for obfuscation and deception. Although the book is primarily driven by our own research in these areas, in every chapter, we will provide ample connections to relevant research from other groups.

Reinforcement Learning (RL) for control has become increasingly popular due to its ability to learn rich feedback policies that take into account uncertainty and complex representations of the environment. When considering safety constraints, constrained optimization approaches, where agents are penalized for constraint violations, are commonly used. In such methods, if agents are initialized in, or must visit, states where constraint violation might be inevitable, it is unclear how much they should be penalized. We address this challenge by formulating a constraint on the counterfactual harm of the learned policy compared to a default, safe policy. In a philosophical sense this formulation only penalizes the learner for constraint violations that it caused; in a practical sense it maintains feasibility of the optimal control problem. We present simulation studies on a rover with uncertain road friction and a tractor-trailer parking environment that demonstrate our constraint formulation enables agents to learn safer policies than contemporary constrained RL methods.

Many classical and modern machine learning algorithms require solving optimization tasks under orthogonal constraints. Solving these tasks often require calculating retraction-based gradient descent updates on the corresponding Riemannian manifold, which can be computationally expensive. Recently Ablin et al. proposed an infeasible retraction-free algorithm, which is significantly more efficient. In this paper, we study the decentralized non-convex optimization task over a network of agents on the Stiefel manifold with retraction-free updates. We propose \textbf{D}ecentralized \textbf{R}etraction-\textbf{F}ree \textbf{G}radient \textbf{T}racking (DRFGT) algorithm, and show that DRFGT exhibits ergodic $\mathcal{O}(1/K)$ convergence rate, the same rate of convergence as the centralized, retraction-based methods. We also provide numerical experiments demonstrating that DRFGT performs on par with the state-of-the-art retraction based methods with substantially reduced computational overhead.

The efficient scheduling of permanent and temporary workers is crucial for Improving the efficiency of sortation center management optimizing the efficiency of the logistics depot while has a direct impact on the fulfillment efficiency and minimizing labor usage. The study begins by establishing operational costs of the entire logistics network. Staff a 0-1 integer linear programming model, with decision management in sortation centers is a key challenge. Staffing needs to be adjusted according to the forecasted shipment variables determining the scheduling of permanent and volume to ensure a sufficient workforce to handle the flow of temporary workers for each time slot on a given day. The goods during peak hours while avoiding the wastage of excess objective function aims to minimize person-days, while manpower during low-demand times. Staff scheduling based constraints ensure fulfillment of hourly labor on effective solution algorithms becomes one of the key requirements, limit workers to one time slot per day, cap strategies to improve the efficiency of the sorting center. By consecutive working days for permanent workers, and reasonably allocating regular and temporary workers, the maintain non-negativity and integer constraints. The sorting speed and accuracy can be improved, thus reducing the model is then solved using genetic algorithms and overall logistics cost and improving customer satisfaction.

When optimizing machine learning models, there are various scenarios where gradient computations are challenging or even infeasible. Furthermore, in reinforcement learning (RL), preference-based RL that only compares between options has wide applications, including reinforcement learning with human feedback in large language models. In this paper, we systematically study optimization of a smooth function $f\colon\mathbb{R}^n\to\mathbb{R}$ only assuming an oracle that compares function values at two points and tells which is larger. When $f$ is convex, we give two algorithms using $\tilde{O}(n/\epsilon)$ and $\tilde{O}(n^{2})$ comparison queries to find an $\epsilon$-optimal solution, respectively. When $f$ is nonconvex, our algorithm uses $\tilde{O}(n/\epsilon^2)$ comparison queries to find an $\epsilon$-approximate stationary point. All these results match the best-known zeroth-order algorithms with function evaluation queries in $n$ dependence, thus suggest that \emph{comparisons are all you need for optimizing smooth functions using derivative-free methods}. In addition, we also give an algorithm for escaping saddle points and reaching an $\epsilon$-second order stationary point of a nonconvex $f$, using $\tilde{O}(n^{1.5}/\epsilon^{2.5})$ comparison queries.

Range-based localization is ubiquitous: global navigation satellite systems (GNSS) power mobile phone-based navigation, and autonomous mobile robots can use range measurements from a variety of modalities including sonar, radar, and even WiFi signals. Many of these localization systems rely on fixed anchors or beacons with known positions acting as transmitters or receivers. In this work, we answer a fundamental question: given a set of positions we would like to localize, how should beacons be placed so as to minimize localization error? Specifically, we present an information theoretic method for optimally selecting an arrangement consisting of a few beacons from a large set of candidate positions. By formulating localization as maximum a posteriori (MAP) estimation, we can cast beacon arrangement as a submodular set function maximization problem. This approach is probabilistically rigorous, simple to implement, and extremely flexible. Furthermore, we prove that the submodular structure of our problem formulation ensures that a greedy algorithm for beacon arrangement has suboptimality guarantees. We compare our method with a number of benchmarks on simulated data and release an open source Python implementation of our algorithm and experiments.

Statistical parity metrics have been widely studied and endorsed in the AI community as a means of achieving fairness, but they suffer from at least two weaknesses. They disregard the actual welfare consequences of decisions and may therefore fail to achieve the kind of fairness that is desired for disadvantaged groups. In addition, they are often incompatible with each other, and there is no convincing justification for selecting one rather than another. This paper explores whether a broader conception of social justice, based on optimizing a social welfare function (SWF), can be useful for assessing various definitions of parity. We focus on the well-known alpha fairness SWF, which has been defended by axiomatic and bargaining arguments over a period of 70 years. We analyze the optimal solution and show that it can justify demographic parity or equalized odds under certain conditions, but frequently requires a departure from these types of parity. In addition, we find that predictive rate parity is of limited usefulness. These results suggest that optimization theory can shed light on the intensely discussed question of how to achieve group fairness in AI.