ACM Transactions on Intelligent Systems and Technology最新文献

A strategy refers to the rules that the agent chooses the available actions to achieve goals. Adopting reasonable strategies is challenging but crucial for an intelligent agent with limited resources working in hazardous, unstructured, and dynamic environments to improve the system’s utility, decrease the overall cost, and increase mission success probability. This paper proposes a novel hierarchical strategy decomposition approach based on Bayesian chaining to separate an intricate policy into several simple sub-policies and organize their relationships as Bayesian strategy networks (BSN). We integrate this approach into the state-of-the-art DRL method – soft actor-critic (SAC), and build the corresponding Bayesian soft actor-critic (BSAC) model by organizing several sub-policies as a joint policy. Our method achieves the state-of-the-art performance on the standard continuous control benchmarks in the OpenAI Gym environment. The results demonstrate that the promising potential of the BSAC method significantly improves training efficiency. Furthermore, we extend the topic to the Multi-Agent systems (MAS), discussing the potential research fields and directions.

Many deep learning works on financial time-series forecasting focus on predicting future prices/returns of individual assets with numerical price-related information for trading, and hence propose models designed for univariate, single task and/or unimodal settings. Forecasting for investment and risk management involves multiple tasks in multivariate settings: forecasts of expected returns and risks of assets in portfolios, and correlations between these assets. As different sources/types of time-series influence future returns, risks and correlations of assets in different ways, it is also important to capture time-series from different modalities. Hence, this paper addresses financial time-series forecasting for investment and risk management in a multivariate, multitask and multimodal setting. Financial time-series forecasting is however challenging due to the low signal-to-noise ratios typical in financial time-series, and as intra-series and inter-series relationships of assets evolve across time. To address these challenges, our proposed Temporal Implicit Multimodal Network (TIME) model learns implicit inter-series relationship networks between assets from multimodal financial time-series at multiple time-steps adaptively. TIME then uses dynamic network and temporal encoding modules to jointly capture such evolving relationships, multimodal financial time-series and temporal representations. Our experiments show that TIME outperforms other state-of-the-art models on multiple forecasting tasks and investment and risk management applications.

Recommender systems are one of the most successful applications of using AI for providing personalized e-services to customers. However, data sparsity is presenting enormous challenges that are hindering the further development of advanced recommender systems. Although cross-domain recommendation partly overcomes data sparsity by transferring knowledge from a source domain with relatively dense data to augment data in the target domain, the current methods do not handle heterogeneous data very well. For example, using today’s cross-domain transfer learning schemes with data comprising clicks, ratings, user reviews, item meta data, and knowledge graphs will likely result in a poorly-performing model. User preferences will not be comprehensively profiled, and accurate recommendations will not be generated. To solve these three challenges – i.e., handling heterogeneous data, avoiding negative transfer, and dealing with data sparsity – we designed a new end-to-end deep adversarial multi-channel transfer network for cross-domain recommendation named AMT-CDR. Heterogeneous data is handled by constructing a cross-domain graph based on real-world knowledge graphs – we used Freebase and YAGO. Negative transfer is prevented through an adversarial learning strategy that maintains consistency across the different data channels. And data sparsity is addressed with an end-to-end neural network that considers data across multiple channels and generates accurate recommendations by leveraging knowledge from both the source and target domains. Extensive experiments on three dual-target cross-domain recommendation tasks demonstrate the superiority of AMT-CDR compared to eight state-of-the-art methods. All source code is available at https://github.com/bjtu-lucas-nlp/AMT-CDR.

Single-cell technologies are revolutionizing the entire field of biology. The large volumes of data generated by single-cell technologies are high-dimensional, sparse, heterogeneous, and have complicated dependency structures, making analyses using conventional machine learning approaches challenging and impractical. In tackling these challenges, deep learning often demonstrates superior performance compared to traditional machine learning methods. In this work, we give a comprehensive survey on deep learning in single-cell analysis. We first introduce background on single-cell technologies and their development, as well as fundamental concepts of deep learning including the most popular deep architectures. We present an overview of the single-cell analytic pipeline pursued in research applications while noting divergences due to data sources or specific applications. We then review seven popular tasks spanning through different stages of the single-cell analysis pipeline, including multimodal integration, imputation, clustering, spatial domain identification, cell-type deconvolution, cell segmentation, and cell-type annotation. Under each task, we describe the most recent developments in classical and deep learning methods and discuss their advantages and disadvantages. Deep learning tools and benchmark datasets are also summarized for each task. Finally, we discuss the future directions and the most recent challenges. This survey will serve as a reference for biologists and computer scientists, encouraging collaborations.

Vehicle routing problem with time window (VRPTW) is of great importance for a wide spectrum of services and real-life applications, such as online take-out and car-hailing platforms. A promising method should generate high-qualified solutions within limited inference time, and there are three major challenges: a) directly optimizing the goal with several practical constraints; b) efficiently handling individual time window limits; and c) modeling the cooperation among the vehicle fleet. In this paper, we present an end-to-end reinforcement learning framework to solve VRPTW. First, we propose an agent model that encodes constraints into features as the input, and conducts harsh policy on the output when generating deterministic results. Second, we design a time penalty augmented reward to model the time window limits during gradient propagation. Third, we design a task handler to enable the cooperation among different vehicles. We perform extensive experiments on two real-world datasets and one public benchmark dataset. Results demonstrate that our solution improves the performance by up to (11.7% ) compared to other RL baselines, and could generate solutions for instances within seconds while existing heuristic baselines take for minutes as well as maintaining the quality of solutions. Moreover, our solution is thoroughly analysed with meaningful implications due to the real-time response ability.

Large language models (LLMs) are gaining increasing popularity in both academia and industry, owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and daily use, their evaluation becomes increasingly critical, not only at the task level, but also at the society level for better understanding of their potential risks. Over the past years, significant efforts have been made to examine LLMs from various perspectives. This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate. Firstly, we provide an overview from the perspective of evaluation tasks, encompassing general natural language processing tasks, reasoning, medical usage, ethics, education, natural and social sciences, agent applications, and other areas. Secondly, we answer the ‘where’ and ‘how’ questions by diving into the evaluation methods and benchmarks, which serve as crucial components in assessing the performance of LLMs. Then, we summarize the success and failure cases of LLMs in different tasks. Finally, we shed light on several future challenges that lie ahead in LLMs evaluation. Our aim is to offer invaluable insights to researchers in the realm of LLMs evaluation, thereby aiding the development of more proficient LLMs. Our key point is that evaluation should be treated as an essential discipline to better assist the development of LLMs. We consistently maintain the related open-source materials at: https://github.com/MLGroupJLU/LLM-eval-survey.

The significant increase in credit card transactions can be attributed to the rapid growth of online shopping and digital payments, particularly during the COVID-19 pandemic. To safeguard cardholders, e-commerce companies, and financial institutions, the implementation of an effective and real-time fraud detection method using modern artificial intelligence techniques is imperative. However, the development of machine-learning-based approaches for fraud detection faces challenges such as inadequate transaction representation, noise labels, and data imbalance. Additionally, practical considerations like dynamic thresholds, concept drift, and verification latency need to be appropriately addressed. In this study, we designed a fraud detection method that accurately extracts a series of spatial and temporal representative features to precisely describe credit card transactions. Furthermore, several auxiliary self-supervised objectives were developed to model cardholders’ behavior sequences. By employing intelligent sampling strategies, potential noise labels were eliminated, thereby reducing the level of data imbalance. The developed method encompasses various innovative functions that cater to practical usage requirements. We applied this method to two real-world datasets, and the results indicated a higher F1 score compared to the most commonly used online fraud detection methods.

Video surveillance systems provide means to detect the presence of potentially malicious drones in the surroundings of critical infrastructures. In particular, these systems collect images and feed them to a deep-learning classifier able to detect the presence of a drone in the input image. However, current classifiers are not efficient in identifying drones that disguise themselves with the image background, e.g., hiding in front of a tree. Furthermore, video-based detection systems heavily rely on the image’s brightness, where darkness imposes significant challenges in detecting drones. Both these phenomena increase the possibilities for attackers to get close to critical infrastructures without being spotted and hence be able to gather sensitive information or cause physical damages, possibly leading to safety threats.

In this paper, we propose RANGO, a drone detection arithmetic able to detect drones in challenging images where the target is difficult to distinguish from the background. RANGO is based on a deep learning architecture that exploits a Preconditioning Operation (PREP) that highlights the target by the difference between the target gradient and the background gradient. The idea is to highlight features that will be useful for classification. After PREP, RANGO uses multiple convolution kernels to make the final decision on the presence of the drone. We test RANGO on a drone image dataset composed of multiple already existing datasets to which we add samples of birds and planes. We then compare RANGO with multiple currently existing approaches to show its superiority. When tested on images with disguising drones, RANGO attains an increase of (6.6% ) mean Average Precision (mAP) compared to YOLOv5 solution. When tested on the conventional dataset, RANGO improves the mAP by approximately (2.2% ), thus confirming its effectiveness also in the general scenario.

Recommendation models are deployed in a variety of commercial applications in order to provide personalized services for users.

However, most of them rely on the users’ original rating records that are often collected by a centralized server for model training, which may cause privacy issues.

Recently, some centralized federated recommendation models are proposed for the protection of users’ privacy, which however requires a server for coordination in the whole process of model training.

As a response, we propose a novel privacy-aware decentralized federated recommendation (DFedRec) model, which is lossless compared with the traditional model in recommendation performance and is thus more accurate than other models in this line.

Specifically, we design a privacy-aware structured client-level graph for the sharing of the model parameters in the process of model training, which is a one-stone-two-bird strategy, i.e., it protects users’ privacy via some randomly sampled fake entries and reduces the communication cost by sharing the model parameters only with the related neighboring users.

With the help of the privacy-aware structured client-level graph, we propose two novel collaborative training mechanisms in the setting without a server, including a batch algorithm DFedRec(b) and a stochastic one DFedRec(s), where the former requires the anonymity mechanism while the latter does not. They are both equivalent to PMF trained in a centralized server and are thus lossless.

We then provide formal analysis of privacy guarantee of our methods and conduct extensive empirical studies on three public datasets with explicit feedback, which show the effectiveness of our DFedRec, i.e., it is privacy aware, communication efficient, and lossless.

With the ever-increasing dataset size and data storage capacity, there is a strong need to build systems that can effectively utilize these vast datasets to extract valuable information. Large datasets often exhibit sparsity and pose cold start problems, necessitating the development of responsible recommender systems. Knowledge graphs have utility in responsibly representing information related to recommendation scenarios. However, many studies overlook explicitly encoding contextual information, which is crucial for reducing the bias of multi-layer propagation. Additionally, existing methods stack multiple layers to encode high-order neighbor information, while disregarding the relational information between items and entities. This oversight hampers their ability to capture the collaborative signal latent in user-item interactions. This is particularly important in health informatics, where knowledge graphs consist of various entities connected to items through different relations. Ignoring the relational information renders them insufficient for modeling user preferences. This work presents an end-to-end recommendation framework named Knowledge Graph Enhanced Contextualized Attention-Based Network (KGCAN). It explicitly encodes both relational and contextual information of entities to preserve the original entity information. Furthermore, a user-specific attention mechanism is employed to capture personalized recommendations. The proposed model is validated on three benchmark datasets through extensive experiments. The experimental results demonstrate that KGCAN outperforms existing KG-based recommendation models. Additionally, a case study from the healthcare domain is discussed, highlighting the importance of attention mechanisms and high-order connectivity in the responsible recommendation system for health informatics.