Search

Cross-domain recommendations (CDRs), which can leverage the relatively abundant information from a richer domain to improve the recommendation performance in a sparser domain, have attracted great attention due to their flexible recommendation strategies. Nevertheless, existing CDR approaches still suffer from severe data sparsity and low semantic sampling efficiency issues, and hardly employ existing reinforcement learning models to improve cross-domain recommendation accuracy. To this end, we propose a new Knowledge-aware and Deep Reinforced Cross-Domain Recommendation framework over Collaborative Knowledge Graph (KRCDR). Specifically, we formalize the cross-domain recommendation task as a Markov Decision Process, and propose a knowledge-aware dual state representation approach to enhance state representations within and across domains for target users by leveraging knowledge graph information. Then, to improve the training performance, we propose a Constrained Self-supervised Actor-Critic network (CSAC) model, in which a constrained neighbor pruning strategy is devised to narrow the exploration space and improve the sampling efficiency, and the CSAC is developed to improve the recommendation policy. Additionally, in our proposed CSAC model, a self-supervised output layer within domains is used as an actor network to generate the recommendation policy, and a Q-learning output layer across domains is used as a critic network to feedback reward signals. Finally, based on the KRCDR approach, we design a new algorithm to assist in generating cross-domain recommendation results. Extensive experiments have been conducted on several real-world datasets, which demonstrate the superiority of our proposed approach compared with state-of-the-art baseline methods.

ABSTRACTSatellite remote sensing, characterized by extensive coverage, frequent revisits, and continuous monitoring, provides essential data support for addressing global challenges. Over the past six decades, thousands of Earth observation satellites and sensors have been deployed worldwide. These valuable Earth observation assets are contributed independently by various nations and organizations employing diverse methodologies. This poses a significant challenge in effectively discovering global Earth observation resources and realizing their full potential. In this paper, we describe the development of GEOSatDB, the most complete semantic database of civil Earth observation satellites developed based on a unified ontology model. A similarity matching method is used to integrate satellite information and a prompt strategy is used to extract unstructured sensor information. The resulting semantic database contains 127,949 semantic statements for 2,340 remote sensing satellites and 1,021 observation sensors. The global Earth observation capabilities of 195 countries worldwide have been analyzed in detail, and a concrete use case along with an associated query demonstration is presented. This database provides significant value in effectively facilitating the semantic understanding and sharing of Earth observation resources.

Why can pre-trained language models (PLMs) learn universal representations and effectively adapt to broad NLP tasks differing a lot superficially? In this work, we empirically find evidence indicating that the adaptations of PLMs to various few-shot tasks can be reparameterized as optimizing only a few free parameters in a unified low-dimensional intrinsic task subspace, which may help us understand why PLMs could easily adapt to various NLP tasks with small-scale data. To find such a subspace and examine its universality, we propose an analysis pipeline called intrinsic prompt tuning (IPT). Specifically, we resort to the recent success of prompt tuning and decompose the soft prompts of multiple NLP tasks into the same low-dimensional nonlinear subspace, then we learn to adapt the PLM to unseen data or tasks by only tuning parameters in this subspace. In the experiments, we study diverse few-shot NLP tasks and surprisingly find that in a 250-dimensional subspace found with 100 tasks, by only tuning 250 free parameters, we can recover 97% and 83% of the full prompt tuning performance for 100 seen tasks (using different training data) and 20 unseen tasks, respectively, showing great generalization ability of the found intrinsic task subspace. Besides being an analysis tool, IPTcould further help us improve the prompt tuning stability.

Pre-trained Language Models (PLMs) which are trained on large text corpus via self-supervised learning method, have yielded promising performance on various tasks in Natural Language Processing (NLP). However, though PLMs with huge parameters can effectively possess rich knowledge learned from massive training text and benefit downstream tasks at the fine-tuning stage, they still have some limitations such as poor reasoning ability due to the lack of external knowledge. Research has been dedicated to incorporating knowledge into PLMs to tackle these issues. In this paper, we present a comprehensive review of Knowledge Enhanced Pre-trained Language Models (KE-PLMs) to provide a clear insight into this thriving field. We introduce appropriate taxonomies respectively for Natural Language Understanding (NLU) and Natural Language Generation (NLG) to highlight these two main tasks of NLP. For NLU, we divide the types of knowledge into four categories: linguistic knowledge, text knowledge, knowledge graph (KG), and rule knowledge. The KE-PLMs for NLG are categorized into KG-based and retrieval-based methods. Finally, we point out some promising future directions of KE-PLMs.

In this paper, a linear and highly polymerized thermoplastic epoxy (TPE) is treated as the matrix to prepare the fully recyclable and high performance carbon fiber reinforced TPE composites (CFRTPE). The TPE is synthesized from bisphenol A epoxy and aniline through a simple polymerization process, whose mechanical and thermal properties (tensile strength: 65 MPa, tensile modulus: 2.3 GPa, initial decomposition temperature: 365.4 °C) are comparable to that of traditional bisphenol A epoxy based thermoset. The CFRTPE composites samples could be prepared by prepreg through molding and vacuum bag method with the mild conditions (180 °Cfor 20 min). The mechanical and thermomechanical properties testing results show that the CFRTPE could achieve comparable and even better mechanical performance compared with the reference specimen and literature reports. Most importantly, both the full-length CFs and TPE of CFRTPE composites could be recycled conveniently and rapidly at room temperature without damaging the morphologies and mechanical properties, which indicates promising potential for practical applications.

Ride-sharing systems should combine environmental protection (through a reduction of fossil fuel usage), socialization, and security. Encouraging people to use ride-sharing systems by satisfying their demands for safety, privacy and convenience is challenging. Most previous works on this topic have focused on finding a fixed path between the driver and the riders either based solely on their locations or using social information. The drivers’ and riders’ lack of options to change or compute the path according to their own preferences and requirements is problematic. With the advancement of mobile social networking technologies, it is necessary to reconsider the principles and desired characteristics of ride-sharing systems. In this paper, we formalized the ride-sharing problem as a multi source-destination path planning problem. An objective function that models different objectives in a unified framework was developed. Moreover, we provide a similarity model, which can reflect the personal preferences of the rides and utilize social media to obtain the current interests of the riders and drivers. The model also allows each driver to generate sub-optimal paths according to his own requirements by suitably adjusting the weights. Two case studies have shown that our system has the potential to find the best possible match and computes the multiple optimal paths against different user-defined objective functions.

Feature selection is an important task in data mining and pattern recognition, especially for high-dimensional data. It aims to select a compact feature subset with the maximal discriminative capability. The discriminability of a feature subset requires that selected features have a high relevance to class labels, whereas the compactness demands a low redundancy within the selected feature subset. This paper defines a new feature redundancy measurement capable of accurately estimating mutual information between features with respect to the target class (MIFS-CR). Based on a relevance measure and this new redundancy measure, a multi-objective evolutionary algorithm with class-dependent redundancy for feature selection (MECY-FS) is presented. The MECY-FS algorithm employs the Pareto optimality to evaluate candidate feature subsets and finds compact feature subsets with both the maximal relevance and the minimal redundancy. Experiments on benchmark datasets are conducted to validate the effectiveness of the new redundancy measure, and the MECY-FS algorithm is verified to be able to generate compact feature subsets with a high predictive capability. [ABSTRACT FROM AUTHOR]