Most existing heuristic optimizers are found to be restricted to problems of moderate dimensionality, and their performance suffers when solving high-dimensional or large-scale optimization tasks. In this paper, we transform the high-dimensional optimization into online decision making problems and propose a stochastic online decisioning hyper-heuristic framework, by considering multi-armed bandits with temporal reward estimation as our essential backbone. The multi-armed bandit problem simulates an agent which tries to balance exploration and exploitation simultaneously. Specifically, we introduce 1) a sliding time window to assign temporal credit for differing heuristics, and 2) boltzmann exploration for balancing the exploration-exploitation tradeoff. The proposed method is well suited for real-world applications, with flexible compatibility for versatile cost definitions, easy interfaces for heuristics as well as fewer hyper-parameters for consistent generalization performance. Experimental studies on the benchmarks results verify the efficacy and significance of the proposed framework, i.e., when considering three differing heuristics, our method reported consistently competitive performance on benchmark problems with a dimensionality up to 10,000.

In this paper, we scale evolutionary algorithms to high-dimensional optimization problems that deceptively possess a low effective dimensionality (certain dimensions do not significantly affect the objective function). To this end, an instantiation of the multiform optimization paradigm is presented, where multiple low-dimensional counterparts of a target high-dimensional task are generated via random embeddings. Since the exact relationship between the auxiliary (low-dimensional) tasks and the target is a priori unknown, a multiform evolutionary algorithm is developed for unifying all formulations into a single multi-task setting. The resultant joint optimization enables the target task to efficiently reuse solutions evolved across various low-dimensional searches via cross-form genetic transfers, hence speeding up overall convergence characteristics. To validate the overall efficacy of our proposed algorithmic framework, comprehensive experimental studies are carried out on well-known continuous benchmark functions as well as a set of practical problems in the hyper-parameter tuning of machine learning models and deep learning models in classification tasks and Predator-Prey games, respectively.

Recent developments in reinforcement learning have been able to derive optimal policies for sophisticated and capable agents, and shown to achieve human-level performance on a number of challenging tasks. Unfortunately, when it comes to multi-agent systems (MASs), complexities such as non-stationarity and partial observability bring new challenges to the field. Building a flexible and efficient multi-agent reinforcement learning algorithm capable of handling complex tasks has to date remained an open challenge. This paper presents an Multi-Agent learning system with the evolution of Social Roles (eSRMA). The main interest is placed on solving the key issues in the definition and evolution of suitable roles, and optimizing the policies accompanied by social roles in MAS efficiently. Specifically, eSRMA incorporates and cultivates role division awareness of agents to improve the ability to deal with complex cooperative tasks. Each agent is assigned a role module, which can dynamically generate roles based on the individuals’ local observations. A novel multi-agent reinforcement learning algorithm (MARL) is designed as the principal driving force that governs the role-policy learning process by a role-attention credit assignment mechanism. Moreover, a role evolution process is developed to help agents dynamically choose appropriate roles in decision-making. Comprehensive experiments on the StarCraft II micromanagement benchmark have demonstrated that eSRMA exhibits superiority in achieving higher learning capability and efficiency for multiple agents compared to the state-of-the-art MARL methods.

In multi-agent reinforcement learning (MARL), the uncertainty of state change and the inconsistency between agents’ local observation and global information are always the main obstacles of cooperative multi-agent exploration. To address these challenges, we propose a novel MARL exploration method by combining surprise minimization and social influence maximization. Considering state entropy as a measure of surprise, surprise minimization is achieved by rewarding the individual’s intrinsic motivation (or rewards) for coping with more stable and familiar situations, hence promoting the policy learning. Furthermore, we introduce mutual information between agents’ actions as a regularizer to maximize the social influence via optimizing a tractable variational estimation. In this way, the agents are guided to interact positively with one another by navigating between states that favor cooperation.

There is a growing demand of automatically synthesizing co-speech gestures for virtual characters. However, it remains a challenge due to the complex relationship between input speeches and target gestures. Most existing works focus on predicting the next gesture that fits the data best, however, such methods are myopic and lack the ability to plan for future gestures. In this paper, we propose a novel reinforcement learning (RL) framework called RACER to generate sequences of gestures that maximize the overall satisfactory. RACER employs a vector quantized variational autoencoder to learn compact representations of gestures and a GPT-based policy architecture to generate coherent sequence of gestures autoregressively. In particular, we propose a contrastive pre-training approach to calculate the rewards, which integrates contextual information into action evaluation and successfully captures the complex relationships between multi-modal speech-gesture data. Experimental results show that our method significantly outperforms existing baselines in terms of both objective metrics and subjective human judgements. Demos can be found at https://github.com/RLracer/RACER.git.

One important component of developing autonomous agents lies in the accurate prediction of their opponents’ behaviors when the agents interact with others in an uncertain environment. Most recent study focuses on first constructing predictive types (or models) of the opponents, considering their various properties of interest, and subsequently using these models to predict their behaviors accordingly. However, as the possible type space can be rather large, it is time-consuming, and sometimes even infeasible, to predict the actual behaviors of opponents with all candidate types. Thus, in this paper a tractable opponent behavior reasoning approach is proposed that facilitates ( $a$ ) extraction of a small yet representative summary of all candidates using sub-modular-type maximization, and accordingly, ( $b$ ) identification of the most appropriate type for real-time behavior prediction based on multi-armed bandits. In addition, we propose a knowledge-transfer scheme through demonstration learning to synchronize subject agents’ knowledge about their opponents’ behaviors. This further reduces the burden of reasoning with all models of their opponents from the perspective of individual subject agents. We integrate the new behavior prediction and reasoning method into a state-of-the-art evolutionary multi-agent framework, namely a memetic multi-agent system (MeMAS), and demonstrate empirical performance in two problem domains.

Memetic multi-agent system (MeMAS) is recently proposed as an enhanced version that integrates meme concept into multi-agent system (MAS) wherein all meme-inspired agents have an improvement in learning performance via meme evolution independently or social interaction. In the process of solving the black box optimization problem, the potential advantages of MeMAS have not been utilized well, which makes it a fertile area for further exploration. This paper presents a memetic multi-agent optimization paradigm through coordinate rotation (MeMAO-R) to combine MeMAS with evolutionary algorithms (EAs) to improve optimization efficiency. Based on MeMAS, the particular interest of MeMAO-R is placed on assisting original complex optimization task with new tasks generated by coordinate rotation. Further, MeMAO-R constructs the social interaction mechanism which facilitates to improve their convergence speed for solving the target optimization problem by utilizing meaningful information transferred across multiple agents with differing views of the target problem. Besides, MeMAO-R employs one or more classical EAs as the fundamental population based evolutionary solvers for multiple agents to optimize multiple tasks in a multi-agent scenario. Lastly, to testify the efficacy of the proposed MeMAO-R, comprehensive empirical studies on basic optimization problems are provided.