Current Transferable Adversarial Examples (TAE) are primarily generated by adding Adversarial Noise (AN). Recent studies emphasize the importance of optimizing Data Augmentation (DA) parameters along with AN, which poses a greater threat to real-world AI applications. However, existing DA-based strategies often struggle to find optimal solutions due to the challenging DA search procedure without proper guidance. In this work, we propose a novel DA-based attack algorithm, GADT. GADT identifies suitable DA parameters through iterative antagonism and uses posterior estimates to update AN based on these parameters. We uniquely employ a differentiable DA operation library to identify adversarial DA parameters and introduce a new loss function as a metric during DA optimization. This loss term enhances adversarial effects while preserving the original image content, maintaining attack crypticity. Extensive experiments on public datasets with various networks demonstrate that GADT can be integrated with existing transferable attack methods, updating their DA parameters effectively while retaining their AN formulation strategies. Furthermore, GADT can be utilized in other black-box attack scenarios, e.g., query-based attacks, offering a new avenue to enhance attacks on real-world AI applications in both research and industrial contexts.

Stochastic decentralized optimization algorithms often suffer from issues such as synchronization overhead and intermittent communication. This paper proposes a $\underline{\rm F}$ully $\underline{\rm S}$tochastic $\underline{\rm P}$rimal $\underline{\rm D}$ual gradient $\underline{\rm A}$lgorithm (FSPDA) that suggests an asynchronous decentralized procedure with (i) sparsified non-blocking communication on random undirected graphs and (ii) local stochastic gradient updates. FSPDA allows multiple local gradient steps to accelerate convergence to stationarity while finding a consensual solution with stochastic primal-dual updates. For problems with smooth (possibly non-convex) objective function, we show that FSPDA converges to an $\mathrm{\mathcal{O}( {\it \sigma /\sqrt{nT}} )}$-stationary solution after $\mathrm{\it T}$ iterations without assuming data heterogeneity. The performance of FSPDA is on par with state-of-the-art algorithms whose convergence depend on static graph and synchronous updates. To our best knowledge, FSPDA is the first asynchronous algorithm that converges exactly under the non-convex setting. Numerical experiments are presented to show the benefits of FSPDA.

Federated learning heavily relies on distributed gradient descent techniques. In the situation where gradient information is not available, the gradients need to be estimated from zeroth-order information, which typically involves computing finite-differences along isotropic random directions. This method suffers from high estimation errors, as the geometric features of the objective landscape may be overlooked during the isotropic sampling. In this work, we propose a non-isotropic sampling method to improve the gradient estimation procedure. Gradients in our method are estimated in a subspace spanned by historical trajectories of solutions, aiming to encourage the exploration of promising regions and hence improve the convergence. The proposed method uses a covariance matrix for sampling which is a convex combination of two parts. The first part is a thin projection matrix containing the basis of the subspace which is designed to improve the exploitation ability. The second part is the historical trajectories. We implement this method in zeroth-order federated settings, and show that the convergence rate aligns with existing ones while introducing no significant overheads in communication or local computation. The effectiveness of our proposal is verified on several numerical experiments in comparison to several commonly-used zeroth-order federated optimization algorithms.

In this paper, we give an in-depth analysis on the mathematical problem formulations and the probabilistic optimization explorations for some of the key components in Transformer model [33] in the field of generative AI. We explore and discuss some potential further enhancement for current state of the art methods for some key underlying technologies of generative AI models from algorithmic and probabilistic optimization perspective. In particular, we present an optimal solution for sub-word encoding (SWE) based on similar initial settings as that of byte-pair encoding (BPE) algorithm in [9] with similar objectives as that of WordPiece approach in [28, 31] to maximize the likelihood of the training data. We also present cross entropy optimization method to optimize hyperparameters for word2vec model [17]. In addition, we propose a factored combination of rotary positional encoding (RoPE) [32] and attention with linear biases (ALiBi) [23] with a harmonic series. We also present a probabilistic FlashAttention [6, 7] (PrFlashAttention) method with a probability distribution over block distances in the matrix to decide which block is likely to participate in a given round of attention computation while maintaining the lower triangle shape of the tensor for autoregressive language models by re-shaping the tensors. Finally, we present staircase adaptive quantization (SAQ) of key-value (KV) cache for multi-query attention (MQA) based on the framework presented in [16] to have gradual quantization degradation while achieving reasonable model quality and cost savings.

Many Bayesian statistical inference problems come down to computing a maximum a-posteriori (MAP) assignment of latent variables. Yet, standard methods for estimating the MAP assignment do not have a finite time guarantee that the algorithm has converged to a fixed point. Previous research has found that MAP inference can be represented in dual form as a linear programming problem with a non-polynomial number of constraints. A Lagrangian relaxation of the dual yields a statistical inference algorithm as a linear programming problem. However, the decision as to which constraints to remove in the relaxation is often heuristic. We present a method for maximum a-posteriori inference in general Bayesian factor models that sequentially adds constraints to the fully relaxed dual problem using Benders' decomposition. Our method enables the incorporation of expressive integer and logical constraints in clustering problems such as must-link, cannot-link, and a minimum number of whole samples allocated to each cluster. Using this approach, we derive MAP estimation algorithms for the Bayesian Gaussian mixture model and latent Dirichlet allocation. Empirical results show that our method produces a higher optimal posterior value compared to Gibbs sampling and variational Bayes methods for standard data sets and provides certificate of convergence.

Meta-learning is a general approach to equip machine learning models with the ability to handle few-shot scenarios when dealing with many tasks. Most existing meta-learning methods work based on the assumption that all tasks are of equal importance. However, real-world applications often present heterogeneous tasks characterized by varying difficulty levels, noise in training samples, or being distinctively different from most other tasks. In this paper, we introduce a novel meta-learning method designed to effectively manage such heterogeneous tasks by employing rank-based task-level learning objectives, Heterogeneous Tasks Robust Meta-learning (HeTRoM). HeTRoM is proficient in handling heterogeneous tasks, and it prevents easy tasks from overwhelming the meta-learner. The approach allows for an efficient iterative optimization algorithm based on bi-level optimization, which is then improved by integrating statistical guidance. Our experimental results demonstrate that our method provides flexibility, enabling users to adapt to diverse task settings and enhancing the meta-learner's overall performance.

Ensuring safety is crucial to promote the application of robot manipulators in open workspace. Factors such as sensor errors or unpredictable collisions make the environment full of uncertainties. In this work, we investigate these potential safety challenges on redundant robot manipulators, and propose a task-oriented planning and control framework to achieve multi-layered safety while maintaining efficient task execution. Our approach consists of two main parts: a task-oriented trajectory planner based on multiple-shooting model predictive control method, and a torque controller that allows safe and efficient collision reaction using only proprioceptive data. Through extensive simulations and real-hardware experiments, we demonstrate that the proposed framework can effectively handle uncertain static or dynamic obstacles, and perform disturbance resistance in manipulation tasks when unforeseen contacts occur. All code will be open-sourced to benefit the community.

Continuum robots, known for their high flexibility and adaptability, offer immense potential for applications such as medical surgery, confined-space inspections, and wearable devices. However, their non-linear elastic nature and complex kinematics present significant challenges in digital modeling and visualization. Identifying the modal shape coefficients of specific robot configuration often requires plenty of physical experiments, which is time-consuming and robot-specific. To address this issue, this research proposes a computational framework that utilizes evolutionary algorithm (EA) to simplify the coefficient identification process. Our method starts by generating datasets using Lie group kinematics and physics-based simulations, defining both ideal configurations and models to be optimized. With the deployment of EA solver, the deviations were iteratively minimized through two fitness objectives \textemdash mean square error of shape deviation (\(\text{MSE}_1\)) and tool center point (TCP) vector deviation (\(\text{MSE}_2\)) \textemdash to align the robot's backbone curve with the desired configuration. Built on the Computer-Aided Design (CAD) platform Grasshopper, this framework provides real-time visualization suitable for development of continuum robots. Results show that this integrated method achieves precise alignment and effective identification. Overall, the objective of this research aims to reduce the modeling complexity of continuum robots, enabling precise, efficient virtual simulation before robot programming and implementation.

We formulate a Markov potential game with final-time reach-avoid objectives by integrating potential game theory with stochastic reach-avoid control. Our focus is on multi-player trajectory planning where players maximize the same multi-player reach-avoid objective: the probability of all participants reaching their designated target states by a specified time, while avoiding collisions with one another. Existing approaches require centralized computation of actions via a global policy, which may have prohibitively expensive communication costs. Instead, we focus on approximations of the global policy via local state feedback policies. First, we adapt the recursive single player reach-avoid value iteration to the multi-player framework with local policies, and show that the same recursion holds on the joint state space. To find each player's optimal local policy, the multi-player reach-avoid value function is projected from the joint state to the local state using the other players' occupancy measures. Then, we propose an iterative best response scheme for the multi-player value iteration to converge to a pure Nash equilibrium. We demonstrate the utility of our approach in finding collision-free policies for multi-player motion planning in simulation.

Adapter-based vision language models were an early attempt to augment large language models (LLMs) with image inputs (Liu et al., 2024). They use a pretrained image embedding model, like CLIP (Radford et al., 2021), and train adapters to map image embeddings directly into the embedding space of a pretrained LLM. However, separate input spaces can limit multimodal understanding and do not support native generation of images. In contrast, early-fusion multimodal models have been introduced as a more general approach that supports unlimited modalities as both input and output (Chameleon Team, 2024; Gemini Team, 2023; OpenAI, 2024). These models project all modalities into a shared tokenized space and are pretrained from scratch on multimodal inputs. In this work, we will refer to early-fusion multimodal models as multimodal fusion models. Just like LLMs, most vision language models are trained to behave safely and reject harmful requests (Bai et al., 2022). Carlini et al. (2024) demonstrated that bypassing safeguards in adapter-based vision language models is easy because input images can be continuously optimized to maximize harmful outputs. This is in contrast to text input optimization, which requires less efficient discrete optimization methods (Zou et al., 2023).