This paper introduces the Adaptive Learning Path Navigation (ALPN) system, a novel approach for enhancing E-learning platforms by providing highly adaptive learning paths for students. The ALPN system integrates the Attentive Knowledge Tracing (AKT) model, which assesses students' knowledge states, with the proposed Entropy-enhanced Proximal Policy Optimization (EPPO) algorithm. This new algorithm optimizes the recommendation of learning materials. By harmonizing these models, the ALPN system tailors the learning path to students' needs, significantly increasing learning effectiveness. Experimental results demonstrate that the ALPN system outperforms previous research by 8.2% in maximizing learning outcomes and provides a 10.5% higher diversity in generating learning paths. The proposed system marks a significant advancement in adaptive E-learning, potentially transforming the educational landscape in the digital era.

Safe reinforcement learning (RL) trains a constraint satisfaction policy by interacting with the environment. We aim to tackle a more challenging problem: learning a safe policy from an offline dataset. We study the offline safe RL problem from a novel multi-objective optimization perspective and propose the $\epsilon$-reducible concept to characterize problem difficulties. The inherent trade-offs between safety and task performance inspire us to propose the constrained decision transformer (CDT) approach, which can dynamically adjust the trade-offs during deployment. Extensive experiments show the advantages of the proposed method in learning an adaptive, safe, robust, and high-reward policy. CDT outperforms its variants and strong offline safe RL baselines by a large margin with the same hyperparameters across all tasks, while keeping the zero-shot adaptation capability to different constraint thresholds, making our approach more suitable for real-world RL under constraints. The code is available at https://github.com/liuzuxin/OSRL.

Reinforcement Learning (RL) has opened up new opportunities to enhance existing smart systems that generally include a complex decision-making process. However, modern RL algorithms, e.g., Deep Q-Networks (DQN), are based on deep neural networks, resulting in high computational costs. In this paper, we propose QHD, an off-policy value-based Hyperdimensional Reinforcement Learning, that mimics brain properties toward robust and real-time learning. QHD relies on a lightweight brain-inspired model to learn an optimal policy in an unknown environment. On both desktop and power-limited embedded platforms, QHD achieves significantly better overall efficiency than DQN while providing higher or comparable rewards. QHD is also suitable for highly-efficient reinforcement learning with great potential for online and real-time learning. Our solution supports a small experience replay batch size that provides 12.3 times speedup compared to DQN while ensuring minimal quality loss. Our evaluation shows QHD capability for real-time learning, providing 34.6 times speedup and significantly better quality of learning than DQN.

In real scenarios, state observations that an agent observes may contain measurement errors or adversarial noises, misleading the agent to take suboptimal actions or even collapse while training. In this paper, we study the training robustness of distributional Reinforcement Learning (RL), a class of state-of-the-art methods that estimate the whole distribution, as opposed to only the expectation, of the total return. Firstly, we validate the contraction of distributional Bellman operators in the State-Noisy Markov Decision Process (SN-MDP), a typical tabular case that incorporates both random and adversarial state observation noises. In the noisy setting with function approximation, we then analyze the vulnerability of least squared loss in expectation-based RL with either linear or nonlinear function approximation. By contrast, we theoretically characterize the bounded gradient norm of distributional RL loss based on the categorical parameterization equipped with the KL divergence. The resulting stable gradients while the optimization in distributional RL accounts for its better training robustness against state observation noises. Finally, extensive experiments on the suite of environments verified that distributional RL is less vulnerable against both random and adversarial noisy state observations compared with its expectation-based counterpart.

The ensemble method is a promising way to mitigate the overestimation issue in Q-learning, where multiple function approximators are used to estimate the action values. It is known that the estimation bias hinges heavily on the ensemble size (i.e., the number of Q-function approximators used in the target), and that determining the `right' ensemble size is highly nontrivial, because of the time-varying nature of the function approximation errors during the learning process. To tackle this challenge, we first derive an upper bound and a lower bound on the estimation bias, based on which the ensemble size is adapted to drive the bias to be nearly zero, thereby coping with the impact of the time-varying approximation errors accordingly. Motivated by the theoretic findings, we advocate that the ensemble method can be combined with Model Identification Adaptive Control (MIAC) for effective ensemble size adaptation. Specifically, we devise Adaptive Ensemble Q-learning (AdaEQ), a generalized ensemble method with two key steps: (a) approximation error characterization which serves as the feedback for flexibly controlling the ensemble size, and (b) ensemble size adaptation tailored towards minimizing the estimation bias. Extensive experiments are carried out to show that AdaEQ can improve the learning performance than the existing methods for the MuJoCo benchmark.

Inspired by cognitive theories of creativity, this paper introduces a computational model (AIGenC) that lays down the necessary components to enable artificial agents to learn, use and generate transferable representations. Unlike machine representation learning, which relies exclusively on raw sensory data, biological representations incorporate relational and associative information that embeds rich and structured concept spaces. The AIGenC model poses a hierarchical graph architecture with various levels and types of representations procured by different components. The first component, Concept Processing, extracts objects and affordances from sensory input and encodes them into a concept space. The resulting representations are stored in a dual memory system and enriched with goal-directed and temporal information acquired through reinforcement learning, creating a higher-level of abstraction. Two additional components work in parallel to detect and recover relevant concepts and create new ones, respectively, in a process akin to cognitive Reflective Reasoning and Blending. The Reflective Reasoning unit detects and recovers from memory concepts relevant to the task by means of a matching process that calculates a similarity value between the current state and memory graph structures. Once the matching interaction ends, rewards and temporal information are added to the graph, building further abstractions. If the reflective reasoning processing fails to offer a suitable solution, a blending operation comes into place, creating new concepts by combining past information. We discuss the model's capability to yield better out-of-distribution generalisation in artificial agents, thus advancing toward Artificial General Intelligence.

Model-based reinforcement learning (MBRL) is a sample efficient technique to obtain control policies, yet unavoidable modeling errors often lead performance deterioration. The model in MBRL is often solely fitted to reconstruct dynamics, state observations in particular, while the impact of model error on the policy is not captured by the training objective. This leads to a mismatch between the intended goal of MBRL, enabling good policy and value learning, and the target of the loss function employed in practice, future state prediction. Naive intuition would suggest that value-aware model learning would fix this problem and, indeed, several solutions to this objective mismatch problem have been proposed based on theoretical analysis. However, they tend to be inferior in practice to commonly used maximum likelihood (MLE) based approaches. In this paper we propose the Value-gradient weighted Model Learning (VaGraM), a novel method for value-aware model learning which improves the performance of MBRL in challenging settings, such as small model capacity and the presence of distracting state dimensions. We analyze both MLE and value-aware approaches and demonstrate how they fail to account for exploration and the behavior of function approximation when learning value-aware models and highlight the additional goals that must be met to stabilize optimization in the deep learning setting. We verify our analysis by showing that our loss function is able to achieve high returns on the Mujoco benchmark suite while being more robust than maximum likelihood based approaches.

We study the problem of computing an optimal policy of an infinite-horizon discounted constrained Markov decision process (constrained MDP). Despite the popularity of Lagrangian-based policy search methods used in practice, the oscillation of policy iterates in these methods has not been fully understood, bringing out issues such as violation of constraints and sensitivity to hyper-parameters. To fill this gap, we employ the Lagrangian method to cast a constrained MDP into a constrained saddle-point problem in which max/min players correspond to primal/dual variables, respectively, and develop two single-time-scale policy-based primal-dual algorithms with non-asymptotic convergence of their policy iterates to an optimal constrained policy. Specifically, we first propose a regularized policy gradient primal-dual (RPG-PD) method that updates the policy using an entropy-regularized policy gradient, and the dual via a quadratic-regularized gradient ascent, simultaneously. We prove that the policy primal-dual iterates of RPG-PD converge to a regularized saddle point with a sublinear rate, while the policy iterates converge sublinearly to an optimal constrained policy. We further instantiate RPG-PD in large state or action spaces by including function approximation in policy parametrization, and establish similar sublinear last-iterate policy convergence. Second, we propose an optimistic policy gradient primal-dual (OPG-PD) method that employs the optimistic gradient method to update primal/dual variables, simultaneously. We prove that the policy primal-dual iterates of OPG-PD converge to a saddle point that contains an optimal constrained policy, with a linear rate. To the best of our knowledge, this work appears to be the first non-asymptotic policy last-iterate convergence result for single-time-scale algorithms in constrained MDPs.

Network slicing enables operators to efficiently support diverse applications on a common physical infrastructure. The ever-increasing densification of network deployment leads to complex and non-trivial inter-cell interference, which requires more than inaccurate analytic models to dynamically optimize resource management for network slices. In this paper, we develop a DIRP algorithm with multiple deep reinforcement learning (DRL) agents to cooperatively optimize resource partition in individual cells to fulfill the requirements of each slice, based on two alternative reward functions. Nevertheless, existing DRL approaches usually tie the pretrained model parameters to specific network environments with poor transferability, which raises practical deployment concerns in large-scale mobile networks. Hence, we design a novel transfer learning-aided DIRP (TL-DIRP) algorithm to ease the transfer of DIRP agents across different network environments in terms of sample efficiency, model reproducibility, and algorithm scalability. The TL-DIRP algorithm first centrally trains a generalized model and then transfers the "generalist" to each local agent as "specialist" with distributed finetuning and execution. TL-DIRP consists of two steps: 1) centralized training of a generalized distributed model, 2) transferring the "generalist" to each "specialist" with distributed finetuning and execution. The numerical results show that not only DIRP outperforms existing baseline approaches in terms of faster convergence and higher reward, but more importantly, TL-DIRP significantly improves the service performance, with reduced exploration cost, accelerated convergence rate, and enhanced model reproducibility. As compared to a traffic-aware baseline, TL-DIRP provides about 15% less violation ratio of the quality of service (QoS) for the worst slice service and 8.8% less violation on the average service QoS.

Evolutionary Algorithms and Deep Reinforcement Learning have both successfully solved control problems across a variety of domains. Recently, algorithms have been proposed which combine these two methods, aiming to leverage the strengths and mitigate the weaknesses of both approaches. In this paper we introduce a new Evolutionary Reinforcement Learning model which combines a particular family of Evolutionary algorithm called Evolutionary Strategies with the off-policy Deep Reinforcement Learning algorithm TD3. The framework utilises a multi-buffer system instead of using a single shared replay buffer. The multi-buffer system allows for the Evolutionary Strategy to search freely in the search space of policies, without running the risk of overpopulating the replay buffer with poorly performing trajectories which limit the number of desirable policy behaviour examples thus negatively impacting the potential of the Deep Reinforcement Learning within the shared framework. The proposed algorithm is demonstrated to perform competitively with current Evolutionary Reinforcement Learning algorithms on MuJoCo control tasks, outperforming the well known state-of-the-art CEM-RL on 3 of the 4 environments tested.