However, it has been shown that existing attacks are perspective, where the sets are the vulnerability regions of each classifier not well suited for this kind of classifiers. In this paper, we discuss of the mixture. We then show that the problem of attacking a the problem of attacking a mixture in a principled way and introduce mixture can be seen as the problem of exploring a lattice. Using this two desirable properties of attacks based on a geometrical analysis of perspective, we identify a series of desirable properties, and devise a the problem (effectiveness and maximality). We then show that existing new attack that satisfies these properties and is efficient in practice.

We investigate the challenge of parametrizing policies for reinforcement learning (RL) in high-dimensional continuous action spaces. Our objective is to develop a multimodal policy that overcomes limitations inherent in the commonly-used Gaussian parameterization. To achieve this, we propose a principled framework that models the continuous RL policy as a generative model of optimal trajectories. By conditioning the policy on a latent variable, we derive a novel variational bound as the optimization objective, which promotes exploration of the environment. We then present a practical model-based RL method, called Reparameterized Policy Gradient (RPG), which leverages the multimodal policy parameterization and learned world model to achieve strong exploration capabilities and high data efficiency. Empirical results demonstrate that our method can help agents evade local optima in tasks with dense rewards and solve challenging sparse-reward environments by incorporating an object-centric intrinsic reward. Our method consistently outperforms previous approaches across a range of tasks. Code and supplementary materials are available on the project page https://haosulab.github.io/RPG/

This work proposes a new analysis technique for gradient descent, establishing provably better convergence rates for smooth, convex optimization than the prior state-of-art textbook proofs. Our theory allows for nonconstant stepsize policies, periodically taking larger steps that may violate the monotone decrease in objective value typically needed by analysis. In fact, contrary to the common intuition, we show periodic long steps, which may increase the objective value in the short term, provably speed up convergence in the long term, with increasingly large gains as longer and longer steps are periodically included. This bears a similarity to accelerated momentum methods, which also depart from ensuring a monotone objective decrease at every iteration. Establishing this requires a proof technique capable of analyzing the overall effect of many iterations at once rather than the typical (naive) one-iteration inductions used in most first-order method analyses. Our proofs are based on the Performance Estimation Problem (PEP) ideas of [1-3], which cast computing/bounding the worst-case problem instance of a given algorithm as a Semidefinite Program (SDP). We show that the existence of a feasible solution to a related SDP proves a descent guarantee after applying a corresponding pattern of nonconstant stepsizes, from which faster convergence guarantees follow.

Federated learning (FL) has emerged as a key technique for distributed machine learning (ML). Most literature on FL has focused on ML model training for (i) a single task/model, with (ii) a synchronous scheme for uplink/downlink transfer of model parameters, and (iii) a static data distribution setting across devices. These assumptions are often not well representative of conditions encountered in practical FL environments. To address this, we develop DMA-FL, which considers dynamic FL with multiple downstream tasks to be trained over an asynchronous model transmission architecture. We first characterize the convergence of ML model training under DMA-FL via introducing a family of scheduling tensors and rectangular functions to capture the scheduling of devices. Our convergence analysis sheds light on the impact of resource allocation, device scheduling, and individual model states on the performance of ML models. We then formulate a non-convex mixed integer optimization problem for jointly configuring the resource allocation and device scheduling to strike an efficient trade-off between energy consumption and ML performance. We develop a solution methodology employing successive convex approximations with convergence guarantee to a stationary point. Through numerical simulations, we reveal the advantages of DMA-FL in terms of model performance and network resource savings.

The motion objectives of a planning as inference problem are formulated as a joint distribution over coupled random variables on a factor graph. Leveraging optimization-inference duality, a fast solution to the maximum a posteriori estimation of the factor graph can be obtained via least-squares optimization. The computational efficiency of this approach can be used in competitive autonomous racing for finding the minimum curvature raceline. Finding the raceline is classified as a global planning problem that entails the computation of a minimum curvature path for a racecar which offers highest cornering speed for a given racetrack resulting in reduced lap time. This work introduces a novel methodology for formulating the minimum curvature raceline planning problem as probabilistic inference on a factor graph. By exploiting the tangential geometry and structural properties inherent in the minimum curvature planning problem, we represent it on a factor graph, which is subsequently solved via sparse least-squares optimization. The results obtained by performing comparative analysis with the quadratic programming-based methodology, the proposed approach demonstrated the superior computing performance, as it provides comparable lap time reduction while achieving fourfold improvement in computational efficiency.

Learning-based optimal control algorithms control unknown systems using past trajectory data and a learned model of the system dynamics. These controllers use either a linear approximation of the learned dynamics, trading performance for faster computation, or nonlinear optimization methods, which typically perform better but can limit real-time applicability. In this work, we present a novel nonlinear controller that exploits differential flatness to achieve similar performance to state-of-the-art learning-based controllers but with significantly less computational effort. Differential flatness is a property of dynamical systems whereby nonlinear systems can be exactly linearized through a nonlinear input mapping. Here, the nonlinear transformation is learned as a Gaussian process and is used in a safety filter that guarantees, with high probability, stability as well as input and flat state constraint satisfaction. This safety filter is then used to refine inputs from a flat model predictive controller to perform constrained nonlinear learning-based optimal control through two successive convex optimizations. We compare our method to state-of-the-art learning-based control strategies and achieve similar performance, but with significantly better computational efficiency, while also respecting flat state and input constraints, and guaranteeing stability.

This study proposes the GOOSE algorithm as a novel metaheuristic algorithm based on the goose's behavior during rest and foraging. The goose stands on one leg and keeps his balance to guard and protect other individuals in the flock. The GOOSE algorithm is benchmarked on 19 well-known benchmark test functions, and the results are verified by a comparative study with genetic algorithm (GA), particle swarm optimization (PSO), dragonfly algorithm (DA), and fitness dependent optimizer (FDO). In addition, the proposed algorithm is tested on 10 modern benchmark functions, and the gained results are compared with three recent algorithms, such as the dragonfly algorithm, whale optimization algorithm (WOA), and salp swarm algorithm (SSA). Moreover, the GOOSE algorithm is tested on 5 classical benchmark functions, and the obtained results are evaluated with six algorithms, such as fitness dependent optimizer (FDO), FOX optimizer, butterfly optimization algorithm (BOA), whale optimization algorithm, dragonfly algorithm, and chimp optimization algorithm (ChOA). The achieved findings attest to the proposed algorithm's superior performance compared to the other algorithms that were utilized in the current study. The technique is then used to optimize Welded beam design and Economic Load Dispatch Problem, three renowned real-world engineering challenges, and the Pathological IgG Fraction in the Nervous System. The outcomes of the engineering case studies illustrate how well the suggested approach can optimize issues that arise in the real-world.

Differentiable Search Index is a recently proposed paradigm for document retrieval, that encodes information about a corpus of documents within the parameters of a neural network and directly maps queries to corresponding documents. These models have achieved state-of-the-art performances for document retrieval across many benchmarks. These kinds of models have a significant limitation: it is not easy to add new documents after a model is trained. We propose IncDSI, a method to add documents in real time (about 20-50ms per document), without retraining the model on the entire dataset (or even parts thereof). Instead we formulate the addition of documents as a constrained optimization problem that makes minimal changes to the network parameters. Although orders of magnitude faster, our approach is competitive with re-training the model on the whole dataset and enables the development of document retrieval systems that can be updated with new information in real-time. Our code for IncDSI is available at https://github.com/varshakishore/IncDSI.

The paradigm of federated learning (FL) to address data privacy concerns by locally training parameters on resource-constrained clients in a distributed manner has garnered significant attention. Nonetheless, FL is not applicable when not all clients within the coverage of the FL server are registered with the FL network. To bridge this gap, this paper proposes joint learner referral aided federated client selection (LRef-FedCS), along with communications and computing resource scheduling, and local model accuracy optimization (LMAO) methods. These methods are designed to minimize the cost incurred by the worst-case participant and ensure the long-term fairness of FL in hierarchical Internet of Things (HieIoT) networks. Utilizing the Lyapunov optimization technique, we reformulate the original problem into a stepwise joint optimization problem (JOP). Subsequently, to tackle the mixed-integer non-convex JOP, we separatively and iteratively address LRef-FedCS and LMAO through the centralized method and self-adaptive global best harmony search (SGHS) algorithm, respectively. To enhance scalability, we further propose a distributed LRef-FedCS approach based on a matching game to replace the centralized method described above. Numerical simulations and experimental results on the MNIST/CIFAR-10 datasets demonstrate that our proposed LRef-FedCS approach could achieve a good balance between pursuing high global accuracy and reducing cost.

In this paper, a Priority-based Dynamic REsource Allocation with decentralized Multi-task assignment (P-DREAM) approach is presented to protect a territory from highly manoeuvring intruders. In the first part, static optimization problems are formulated to compute the following parameters of the perimeter defense problem; the number of reserve stations, their locations, the priority region, the monitoring region, and the minimum number of defenders required for the monitoring purpose. The concept of a prioritized intruder is proposed here to identify and handle those critical intruders (computed based on the velocity ratio and location) to be tackled on a priority basis. The computed priority region helps to assign reserve defenders sufficiently earlier such that they can neutralize the prioritized intruders. The monitoring region defines the minimum region to be monitored and is sufficient enough to handle the intruders. In the second part, the earlier developed DREAM approach is modified to incorporate the priority of an intruder. The proposed P-DREAM approach assigns the defenders to the prioritized intruders as the first task. A convex territory protection problem is simulated to illustrate the P-DREAM approach. It involves the computation of static parameters and solving the prioritized task assignments with dynamic resource allocation. Monte-Carlo results were conducted to verify the performance of P-DREAM, and the results clearly show that the P-DREAM approach can protect the territory with consistent performance against highly manoeuvring intruders.