This paper investigates neural networks-based state-dependent intermittent control (SDIC) from a data-driven perspective, considering unknown continuous-time linear systems subject to external disturbances. Instead of relying on precise system models, the proposed approach utilizes offline-collected data to develop a data-driven SDIC scheme. An ℓ1-norm-based convex optimization method is employed to estimate the domain of attraction (DOA), enabling the partitioning of the state space into certified control regions. This facilitates state-dependent control updates that reduce communication burden while preserving system stability. Compared with existing methods, the proposed scheme offers greater flexibility, as its triggering mechanism does not depend on prior model knowledge or a small estimated DOA. Furthermore, it is practically implementable: a neural feedback controller is constructed to satisfy Lyapunov-based stability conditions using only a data-driven linear matrix inequality (LMI), without requiring complex additional assumptions. The effectiveness of the proposed strategy is demonstrated through simulations on a real HVAC system. The learned DOA-based region partitioning allows the controller to adapt to varying environmental conditions while ensuring stability and energy efficiency. Comprehensive simulation results validate the practicality and performance of the proposed control framework.

Reinforcement learning (RL) has become a trending paradigm for training large language models (LLMs), particularly for reasoning tasks. Effective RL for LLMs requires massive parallelization and poses an urgent need for efficient training systems. Most existing large-scale RL systems for LLMs are synchronous by alternating generation and training in a batch setting, where the rollouts in each training batch are generated by the same (or latest) model. This stabilizes RL training but suffers from severe system-level inefficiency. Generation must wait until the longest output in the batch is completed before model update, resulting in GPU underutilization. We present AReaL, a fully asynchronous RL system that completely decouples generation from training. Rollout workers in AReaL continuously generate new outputs without waiting, while training workers update the model whenever a batch of data is collected. AReaL also incorporates a collection of system-level optimizations, leading to substantially higher GPU utilization. To stabilize RL training, AReaL balances the workload of rollout and training workers to control data staleness, and adopts a staleness-enhanced PPO variant to better handle outdated training samples. Extensive experiments on math and code reasoning benchmarks show that AReaL achieves up to 2.77x training speedup compared to synchronous systems with the same number of GPUs and matched or even improved final performance. The code of AReaL is available at https://github.com/inclusionAI/AReaL/.

In this article, we study the optimal feedback control problems of knowledge dissemination processes in multilayer complex networks. First, a node-based model is established in multilayer complex networks and two collaborative control strategies are exerted to increase the scope and speed of knowledge dissemination, forming a closed-loop control system. Then, we develop a two-layer optimal control framework. At the upper level, the optimal solution of the control system is solved and sent to the lower layer. At the lower level, a model predictive controller (MPC) receives input information from the upper level and is formulated to decide on the network and then transmits it to its heterogeneous networks which can reduce control resources and computation complexity. Finally, numerical simulations are conducted to confirm the theoretical results.

This article investigates a robust guaranteed cost finite-time control for coupled neural networks with parametric uncertainties. The parameter uncertainties are assumed to be time-varying norm bounded, which appears on the system state and input matrices. The robust guaranteed cost control laws presented in this article include both continuous feedback controllers and intermittent feedback controllers, which were rarely found in the literature. The proposed guaranteed cost finite-time control is designed in terms of a set of linear-matrix inequalities (LMIs) to steer the coupled neural networks to achieve finite-time synchronization with an upper bound of a guaranteed cost function. Furthermore, open-loop optimization problems are formulated to minimize the upper bound of the quadratic cost function and convergence time, it can obtain the optimal guaranteed cost periodically intermittent and continuous feedback control parameters. Finally, the proposed guaranteed cost periodically intermittent and continuous feedback control schemes are verified by simulations.

In this paper, we present VisRel, a deployed large-scale media search system that leverages text understanding, media understanding, and multimodal technologies to deliver a modern multimedia search experience. We share our insight on developing image and video understanding models for content retrieval, training efficient and effective media-to-query relevance models, and refining online and offline metrics to measure the success of one of the largest media search databases in the industry. We summarize our learnings gathered from hundreds of A/B test experiments and describe the most effective technical approaches. The techniques presented in this work have contributed 34% (abs.) improvement to media-to-query relevance and 10% improvement to user engagement. We believe that this work can provide practical solutions and insights for engineers who are interested in applying media understanding technologies to empower multimedia search systems that operate at Facebook scale.

Sum-product networks (SPNs) are a class of probabilistic graphical models that allow tractable marginal inference. However, the maximum a posteriori (MAP) inference in SPNs is NP-hard. We investigate MAP inference in SPNs from both theoretical and algorithmic perspectives. For the theoretical part, we reduce general MAP inference to its special case without evidence and hidden variables; we also show that it is NP-hard to approximate the MAP problem to 2nε for fixed 0 ≤ ε < 1, where n is the input size. For the algorithmic part, we first present an exact MAP solver that runs reasonably fast and could handle SPNs with up to 1k variables and 150k arcs in our experiments. We then present a new approximate MAP solver with a good balance between speed and accuracy, and our comprehensive experiments on real-world datasets show that it has better overall performance than existing approximate solvers.