In this paper, we consider the problem of sensor selection for parameter estimation with correlated measurement noise. We seek optimal activations by formulating an optimization problem, in which error, given trace inverse Bayesian Fisher information matrix, is minimized subject to energy constraints. has been widely used as effective criterion. However, existing information-based methods are limited case uncorrelated noise or weakly due use approximate metrics. By contrast, here derive...

Pretrained models from self-supervision are prevalently used in fine-tuning downstream tasks faster or for better accuracy. However, gaining robustness pretraining is left unexplored. We introduce adversarial training into self-supervision, to provide general-purpose robust pretrained the first time. find these can benefit subsequent two ways: i) boosting final model robustness; ii) saving computation cost, if proceeding towards fine-tuning. conduct extensive experiments demonstrate that...

Abstract The production of hydrogen peroxide (H 2 O ) through two‐electron oxygen reduction reaction (2e − ORR) has emerged as a more environmentally friendly alternative to the traditional anthraquinone method. Although oxidized carbon catalysts have intensive developed due their high selectivity and activity, yield conversion rate H under overpotential still limited. produced was rapidly consumed by increased intensity reduction, which could ascribe decomposition radicals voltage in...

In this paper, we are interested in learning the underlying graph structure behind training data. Solving basic problem is essential to carry out any signal processing or machine task. To realize this, assume that data smooth with respect topology, and parameterize topology using an edge sampling function. That is, Laplacian expressed terms of a sparse selection vector, which provides explicit handle control sparsity level graph. We solve given some both noiseless noisy settings. Given true...

It is well known that deep neural networks (DNNs) are vulnerable to adversarial attacks, which implemented by adding crafted perturbations onto benign examples. Min-max robust optimization based training can provide a notion of security against attacks. However, robustness requires significantly larger capacity the network than for natural with only This paper proposes framework concurrent and weight pruning enables model compression while still preserving essentially tackles dilemma...

This paper studies a class of adaptive gradient based momentum algorithms that update the search directions and learning rates simultaneously using past gradients. class, which we refer to as "Adam-type", includes popular such Adam, AMSGrad AdaGrad. Despite their popularity in training deep neural networks, convergence these for solving nonconvex problems remains an open question. provides set mild sufficient conditions guarantee Adam-type methods. We prove under our derived conditions,...

Zeroth-order (ZO) optimization is a subset of gradient-free that emerges in many signal processing and machine learning (ML) applications. It used for solving problems similarly to gradient-based methods. However, it does not require the gradient, using only function evaluations. Specifically, ZO iteratively performs three major steps: gradient estimation, descent direction computation, solution update. In this article, we provide comprehensive review optimization, with an emphasis on...

Despite the record-breaking performance in Text-to-Image (T2I) generation by Stable Diffusion, less research attention is paid to its adversarial robustness. In this work, we study problem of attack for Diffusion and ask if an text prompt can be obtained even absence end-to-end model queries. We call resulting 'query-free generation'. To resolve problem, show that vulnerability T2I models rooted lack robustness encoders, e.g., CLIP encoder used attacking Diffusion. Based on such insight,...

We revisit and advance visual prompting (VP), an input technique for vision tasks. VP can reprogram a fixed, pre-trained source model to accomplish downstream tasks in the target domain by simply incorporating universal prompts (in terms of perturbation patterns) into data points. Yet, it remains elusive why stays effective even given ruleless label mapping (LM) between classes classes. Inspired above, we ask: How is LM interrelated with VP? And how exploit such relationship improve its...

We consider the problem of finding optimal time-periodic sensor schedules for estimating state discrete-time dynamical systems. assume that {multiple} sensors have been deployed and are subject to resource constraints, which limits number times each can be activated over one period periodic schedule. seek an algorithm strikes a balance between estimation accuracy total activations period. make correspondence active nonzero columns estimator gain. formulate optimization in we minimize trace...

The computer vision world has been re-gaining enthusiasm in various pre-trained models, including both classical ImageNet supervised pre-training and recently emerged self-supervised such as simCLR [10] MoCo [40]. Pre-trained weights often boost a wide range of downstream tasks classification, detection, segmentation. Latest studies suggest that benefits from gigantic model capacity [11]. We are hereby curious ask: after pre-training, does indeed have to stay large for its transferability?...

Heavy metals (HMs) have become a major environmental pollutant threatening ecosystems and human health. Although hyperaccumulators provide viable alternative for the bioremediation of HMs, potential phytoremediation is often limited by small biomass slow growth rate HM toxicity to plants. Here, plant growth-promoting bacteria (PGPB)-assisted was used enhance HM-contaminated soils. A PGPB with HM-tolerant (HMT-PGPB), Bacillus sp. PGP15 isolated from rhizosphere cadmium (Cd) hyperaccumulator,...

In this paper, we study the problem of temporal video grounding (TVG), which aims to predict starting/ending time points moments described by a text sentence within long untrimmed video. Benefiting from fine-grained 3D visual features, TVG techniques have achieved remarkable progress in recent years. However, high complexity convolutional neural networks (CNNs) makes extracting dense features time-consuming, calls for intensive memory and computing resources. Towards efficient TVG, propose...

Abstract Highly active oxygen evolution reaction (OER) electrocatalysts, such as those containing Fe, often face the challenge of severe dissolution elements. Addressing this concern through establishment a dynamically stable interface during OER presents promising strategy, achieved by manipulation catalyst components. Herein, findings reveal that Fe loss predominantly occurs initial activation phase, marked irreversible structural distortion disrupts interfacial dynamical stability. By...

In the context of distributed estimation, we consider problem sensor collaboration, which refers to act sharing measurements with neighboring sensors prior transmission a fusion center. While incorporating cost aim find optimal sparse collaboration schemes subject certain information or energy constraint. Two types problems are studied: minimum an constraint; and maximum To solve resulting problems, present tractable optimization formulations propose efficient methods that render...

In this paper, we design and analyze a new zeroth-order online algorithm, namely, the alternating direction method of multipliers (ZOO-ADMM), which enjoys dual advantages being gradient-free operation employing ADMM to accommodate complex structured regularizers. Compared first-order gradient-based show that ZOO-ADMM requires $\sqrt{m}$ times more iterations, leading convergence rate $O(\sqrt{m}/\sqrt{T})$, where $m$ is number optimization variables, $T$ iterations. To accelerate ZOO-ADMM,...

Memristors have recently received significant attention as device-level components for building a novel generation of computing systems. These devices many promising features, such non-volatility, low power consumption, high density, and excellent scalability. The ability to control modify biasing voltages at memristor terminals make them candidates efficiently perform matrix-vector multiplications solve systems linear equations. In this article, we discuss how networks memristors arranged...

Robust machine learning is currently one of the most prominent topics which could potentially help shaping a future advanced AI platforms that not only perform well in average cases but also worst or adverse situations. Despite long-term vision, however, existing studies on black-box adversarial attacks are still restricted to very specific settings threat models (e.g., single distortion metric and restrictive assumption target model's feedback queries) and/or suffer from prohibitively high...