Policy-based algorithms are among the most widely adopted techniques in model-free RL, thanks to their strong theoretical groundings and good properties in continuous action spaces. Unfortunately, these methods require precise and problem-specific hyperparameter tuning to achieve good performance, and tend to struggle when asked to accomplish a series of heterogeneous tasks. In particular, the selection of the step size has a crucial impact on their ability to learn a highly performing policy, affecting the speed and the stability of the training process, and often being the main culprit for poor results. In this paper, we tackle these issues with a Meta Reinforcement Learning approach, by introducing a new formulation, known as meta-MDP, that can be used to solve any hyperparameter selection problem in RL with contextual processes. After providing a theoretical Lipschitz bound to the difference of performance in different tasks, we adopt the proposed framework to train a batch RL algorithm to dynamically recommend the most adequate step size for different policies and tasks. In conclusion, we present an experimental campaign to show the advantages of selecting an adaptive learning rate in heterogeneous environments.
相關內容
The vast increase of Internet of Things (IoT) technologies and the ever-evolving attack vectors have increased cyber-security risks dramatically. A common approach to implementing AI-based Intrusion Detection systems (IDSs) in distributed IoT systems is in a centralised manner. However, this approach may violate data privacy and prohibit IDS scalability. Therefore, intrusion detection solutions in IoT ecosystems need to move towards a decentralised direction. Federated Learning (FL) has attracted significant interest in recent years due to its ability to perform collaborative learning while preserving data confidentiality and locality. Nevertheless, most FL-based IDS for IoT systems are designed under unrealistic data distribution conditions. To that end, we design an experiment representative of the real world and evaluate the performance of an FL-based IDS. For our experiments, we rely on TON-IoT, a realistic IoT network traffic dataset, associating each IP address with a single FL client. Additionally, we explore pre-training and investigate various aggregation methods to mitigate the impact of data heterogeneity. Lastly, we benchmark our approach against a centralised solution. The comparison shows that the heterogeneous nature of the data has a considerable negative impact on the model's performance when trained in a distributed manner. However, in the case of a pre-trained initial global FL model, we demonstrate a performance improvement of over 20% (F1-score) compared to a randomly initiated global model.
In this paper, we explore two fundamental first-order algorithms in convex optimization, namely, gradient descent (GD) and proximal gradient method (ProxGD). Our focus is on making these algorithms entirely adaptive by leveraging local curvature information of smooth functions. We propose adaptive versions of GD and ProxGD that are based on observed gradient differences and, thus, have no added computational costs. Moreover, we prove convergence of our methods assuming only local Lipschitzness of the gradient. In addition, the proposed versions allow for even larger stepsizes than those initially suggested in [MM20].
Text-to-motion generation has gained increasing attention, but most existing methods are limited to generating short-term motions that correspond to a single sentence describing a single action. However, when a text stream describes a sequence of continuous motions, the generated motions corresponding to each sentence may not be coherently linked. Existing long-term motion generation methods face two main issues. Firstly, they cannot directly generate coherent motions and require additional operations such as interpolation to process the generated actions. Secondly, they generate subsequent actions in an autoregressive manner without considering the influence of future actions on previous ones. To address these issues, we propose a novel approach that utilizes a past-conditioned diffusion model with two optional coherent sampling methods: Past Inpainting Sampling and Compositional Transition Sampling. Past Inpainting Sampling completes subsequent motions by treating previous motions as conditions, while Compositional Transition Sampling models the distribution of the transition as the composition of two adjacent motions guided by different text prompts. Our experimental results demonstrate that our proposed method is capable of generating compositional and coherent long-term 3D human motions controlled by a user-instructed long text stream. The code is available at \href{//github.com/yangzhao1230/PCMDM}{//github.com/yangzhao1230/PCMDM}.
This paper proposes a locally differentially private federated learning algorithm for strongly convex but possibly nonsmooth problems that protects the gradients of each worker against an honest but curious server. The proposed algorithm adds artificial noise to the shared information to ensure privacy and dynamically allocates the time-varying noise variance to minimize an upper bound of the optimization error subject to a predefined privacy budget constraint. This allows for an arbitrarily large but finite number of iterations to achieve both privacy protection and utility up to a neighborhood of the optimal solution, removing the need for tuning the number of iterations. Numerical results show the superiority of the proposed algorithm over state-of-the-art methods.
Vision transformers have achieved leading performance on various visual tasks yet still suffer from high computational complexity. The situation deteriorates in dense prediction tasks like semantic segmentation, as high-resolution inputs and outputs usually imply more tokens involved in computations. Directly removing the less attentive tokens has been discussed for the image classification task but can not be extended to semantic segmentation since a dense prediction is required for every patch. To this end, this work introduces a Dynamic Token Pruning (DToP) method based on the early exit of tokens for semantic segmentation. Motivated by the coarse-to-fine segmentation process by humans, we naturally split the widely adopted auxiliary-loss-based network architecture into several stages, where each auxiliary block grades every token's difficulty level. We can finalize the prediction of easy tokens in advance without completing the entire forward pass. Moreover, we keep $k$ highest confidence tokens for each semantic category to uphold the representative context information. Thus, computational complexity will change with the difficulty of the input, akin to the way humans do segmentation. Experiments suggest that the proposed DToP architecture reduces on average $20\% - 35\%$ of computational cost for current semantic segmentation methods based on plain vision transformers without accuracy degradation.
Trust-aware human-robot interaction (HRI) has received increasing research attention, as trust has been shown to be a crucial factor for effective HRI. Research in trust-aware HRI discovered a dilemma -- maximizing task rewards often leads to decreased human trust, while maximizing human trust would compromise task performance. In this work, we address this dilemma by formulating the HRI process as a two-player Markov game and utilizing the reward-shaping technique to improve human trust while limiting performance loss. Specifically, we show that when the shaping reward is potential-based, the performance loss can be bounded by the potential functions evaluated at the final states of the Markov game. We apply the proposed framework to the experience-based trust model, resulting in a linear program that can be efficiently solved and deployed in real-world applications. We evaluate the proposed framework in a simulation scenario where a human-robot team performs a search-and-rescue mission. The results demonstrate that the proposed framework successfully modifies the robot's optimal policy, enabling it to increase human trust at a minimal task performance cost.
Energy optimization leveraging artificially intelligent algorithms has been proven effective. However, when buildings are commissioned, there is no historical data that could be used to train these algorithms. On-line Reinforcement Learning (RL) algorithms have shown significant promise, but their deployment carries a significant risk, because as the RL agent initially explores its action space it could cause significant discomfort to the building residents. In this paper we present ReLBOT - a new technique that uses transfer learning in conjunction with deep RL to transfer knowledge from an existing, optimized and instrumented building, to the newly commissioning smart building, to reduce the adverse impact of the reinforcement learning agent's warm-up period. We demonstrate improvements of up to 6.2 times in the duration, and up to 132 times in prediction variance, for the reinforcement learning agent's warm-up period.
We present SEIF, a methodology that combines static analysis with symbolic execution to verify and explicate information flow paths in a hardware design. SEIF begins with a statically built model of the information flow through a design and uses guided symbolic execution to recognize and eliminate non-flows with high precision or to find corresponding paths through the design state for true flows. We evaluate SEIF on two open-source CPUs, an AES core, and the AKER access control module. SEIF can exhaustively explore 10-12 clock cycles deep in 4-6 seconds on average, and can automatically account for 86-90% of the paths in the statically built model. Additionally, SEIF can be used to find multiple violating paths for security properties, providing a new angle for security verification.
Graph Neural Networks (GNNs) have recently become increasingly popular due to their ability to learn complex systems of relations or interactions arising in a broad spectrum of problems ranging from biology and particle physics to social networks and recommendation systems. Despite the plethora of different models for deep learning on graphs, few approaches have been proposed thus far for dealing with graphs that present some sort of dynamic nature (e.g. evolving features or connectivity over time). In this paper, we present Temporal Graph Networks (TGNs), a generic, efficient framework for deep learning on dynamic graphs represented as sequences of timed events. Thanks to a novel combination of memory modules and graph-based operators, TGNs are able to significantly outperform previous approaches being at the same time more computationally efficient. We furthermore show that several previous models for learning on dynamic graphs can be cast as specific instances of our framework. We perform a detailed ablation study of different components of our framework and devise the best configuration that achieves state-of-the-art performance on several transductive and inductive prediction tasks for dynamic graphs.
Recommender System (RS) is a hot area where artificial intelligence (AI) techniques can be effectively applied to improve performance. Since the well-known Netflix Challenge, collaborative filtering (CF) has become the most popular and effective recommendation method. Despite their success in CF, various AI techniques still have to face the data sparsity and cold start problems. Previous works tried to solve these two problems by utilizing auxiliary information, such as social connections among users and meta-data of items. However, they process different types of information separately, leading to information loss. In this work, we propose to utilize Heterogeneous Information Network (HIN), which is a natural and general representation of different types of data, to enhance CF-based recommending methods. HIN-based recommender systems face two problems: how to represent high-level semantics for recommendation and how to fuse the heterogeneous information to recommend. To address these problems, we propose to applying meta-graph to HIN-based RS and solve the information fusion problem with a "matrix factorization (MF) + factorization machine (FM)" framework. For the "MF" part, we obtain user-item similarity matrices from each meta-graph and adopt low-rank matrix approximation to get latent features for both users and items. For the "FM" part, we propose to apply FM with Group lasso (FMG) on the obtained features to simultaneously predict missing ratings and select useful meta-graphs. Experimental results on two large real-world datasets, i.e., Amazon and Yelp, show that our proposed approach is better than that of the state-of-the-art FM and other HIN-based recommending methods.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191