Advancements in neural networks have significantly enhanced the performance of classification models, achieving remarkable accuracy across diverse datasets. However, these models often lack transparency and do not support interactive reasoning with human users, which are essential attributes for applications that require trust and user engagement. To overcome these limitations, we introduce an innovative framework, Neural Causal Graph (NCG), that integrates causal inference with neural networks to enable interpretable and intervenable reasoning. We then propose an intervention training method to model the intervention probability of the prediction, serving as a contextual prompt to facilitate the fine-grained reasoning and human-AI interaction abilities of NCG. Our experiments show that the proposed framework significantly enhances the performance of traditional classification baselines. Furthermore, NCG achieves nearly 95\% top-1 accuracy on the ImageNet dataset by employing a test-time intervention method. This framework not only supports sophisticated post-hoc interpretation but also enables dynamic human-AI interactions, significantly improving the model's transparency and applicability in real-world scenarios.

Given the fact descriptions of legal cases, the legal judgment prediction (LJP) problem aims to determine three judgment tasks of law articles, charges, and the term of penalty. Most existing studies have considered task dependencies while neglecting the prior dependencies of labels among different tasks. Therefore, how to make better use of the information on the relation dependencies among tasks and labels becomes a crucial issue. To this end, we transform the text classification problem into a node classification framework based on graph reasoning and supervised contrastive learning (SCL) techniques, named GraSCL. Specifically, we first design a graph reasoning network to model the potential dependency structures and facilitate relational learning under various graph topologies. Then, we introduce the SCL method for the LJP task to further leverage the label relation on the graph. To accommodate the node classification settings, we extend the traditional SCL method to novel variants for SCL at the node level, which allows the GraSCL framework to be trained efficiently even with small batches. Furthermore, to recognize the importance of hard negative samples in contrastive learning, we introduce a simple yet effective technique called online hard negative mining (OHNM) to enhance our SCL approach. This technique complements our SCL method and enables us to control the number and complexity of negative samples, leading to further improvements in the model's performance. Finally, extensive experiments are conducted on two well-known benchmarks, demonstrating the effectiveness and rationality of our proposed SCL approach as compared to the state-of-the-art competitors.

In Large Language Models (LLMs), text generation that involves knowledge representation is often fraught with the risk of "hallucinations'', where models confidently produce erroneous or fabricated content. These inaccuracies often stem from intrinsic biases in the pre-training stage or from the incorporation of human preference biases during the fine-tuning process. To mitigate these issues, we take inspiration from Goldman's causal theory of knowledge, which asserts that knowledge is not merely about having a true belief but also involves a causal connection between the belief and the truth of the proposition. We instantiate this theory within the context of Knowledge Question Answering (KQA) by constructing a causal graph that delineates the pathways between the candidate knowledge and belief. Through the application of the do-calculus rules from structural causal models, we devise an unbiased estimation framework based on this causal graph, thereby establishing a methodology for knowledge modeling grounded in causal inference. The resulting CORE framework (short for "Causal knOwledge REasoning'') is comprised of four essential components: question answering, causal reasoning, belief scoring, and refinement. Together, they synergistically improve the KQA system by fostering faithful reasoning and introspection. Extensive experiments are conducted on ScienceQA and HotpotQA datasets, which demonstrate the effectiveness and rationality of the CORE framework.

With the development of edge computing and 5G, the demand for resource-limited devices to execute computation-intensive tasks can be effectively alleviated. The research on computation offloading lays an essential foundation for realizing mobile edge computing, and deep reinforcement learning (DRL) has become an emerging technique to address the computation offloading problem. This article utilizes a DRL-based algorithm to design a decentralized computation offloading framework aimed at minimizing the computational cost. We employ a multiuser system model with a single-edge server suitable for industrial scenarios. Then, we propose a dual-critic deep deterministic policy gradient (DC-DDPG) algorithm based on the deep deterministic policy gradient (DDPG) algorithm to tackle computation offloading and resource allocation problems for all users. DC-DDPG adopts two critic nets in both the primary and target nets to fit the action value of two different optimization objectives, which expedites the convergence during the training process and reduces the computational cost of the edge computation system during operation. Compared with other DRL methods, such as deep Q-network and DDPG, numerical results demonstrate that the proposed DC-DDPG algorithm has a faster convergence speed and performs significantly better than other DRL-based algorithms in terms of system computational cost in computing-intensive tasks, which makes it more suitable for industrial intelligent manufacturing scenarios with large data volume.

Machining equipment often faces health monitoring problems during long-term use. As an essential piece of equipment for industrial processing, the state of the cutter tool is directly related to the quality of machined parts. Therefore, tool wear prediction plays an important role in improving the quality of parts and achieving intelligent management of equipment health. In real industrial scenarios., some complex tool milling needs to change the machining parameters, and the changes in these machining parameters will directly affect the tool wear status. In addition, some redundant data collected by sensors can also affect the model's training speed and prediction accuracy. To solve the above problems, a CNN-AIndRNN dual-input model is proposed in this paper. The method empowers the IndRNN attention regulation by introducing the attention mechanism EleAttG to reduce the influence of redundant information on the model. Meanwhile, CNN and AIndRNN are used to extract local and temporal features of the data to avoid information loss. For the problem of multiple working conditions, both sensor signals and machining parameters are used as model inputs in this paper to emphasize the influence of machining parameters on tool wear. Finally, the hyperparameters of the model are optimally selected using the whale optimization algorithm. The proposed model is validated on the NASA milling dataset. The experimental results show that the proposed model has a smaller RMSE and MAE than other baseline models. Machining parameters as a model input can effectively improve the performance of the model, and WOA can also optimize the model for model prediction accuracy.

This work proposes an innovative approach for the gating mechanism of RNN class models. A transfer function is embedded into the original sigmoid to form a new gate function called xigmoid. The purpose is to alleviate the gradient amplification problem when the models are trying to learn features at far end of a long time series. Using the xigmoid function, original LSTM and GRU are converted to xLSTM and xGRU, respectively. The initialization method for the trainable parameters of xigmoid is also derived and discussed as a necessary support to the new method. Verification experiments are conducted for xLSTM and xGRU against several baseline models, showing both faster convergency/training and better accuracy of the proposed xigmoid-based models. The code and datasets are available at https://github.com/privateos/xigmoid

The cross-modal person re-identification (ReID) aims to retrieve one person from one modality to the other single modality, such as text-based and sketch-based ReID tasks. However, for these different modalities of describing a person, combining multiple aspects can obviously make full use of complementary information and improve the identification performance. Therefore, to explore how to comprehensively consider multi-modal information, we advance a novel multi-modal person re-identification task, which utilizes both text and sketch as a descriptive query to retrieve desired images. In fact, the textual description and the visual description are understood together to retrieve the person in the database to be more aligned with real-world scenarios, which is promising but seldom considered. Besides, based on an existing sketch-based ReID dataset, we construct a new dataset, TriReID, to support this challenging task in a semi-automated way. Particularly, we implement an image captioning model under the active learning paradigm to generate sentences suitable for ReID, in which the quality scores of the three levels are customized. Moreover, we propose a novel framework named Descriptive Fusion Model (DFM) to solve the multi-modal ReID issue. Specifically, we first develop a flexible descriptive embedding function to fuse the text and sketch modalities. Further, the fused descriptive semantic feature is jointly optimized under the generative adversarial paradigm to mitigate the cross-modal semantic gap. Extensive experiments on the TriReID dataset demonstrate the effectiveness and rationality of our proposed solution.

The newly emerging language-based video moment retrieval task aims at retrieving a target video moment from an untrimmed video given a natural language as the query. It is more applicable in reality since it is able to accurately localize a specific video moment, as compared to traditional whole video retrieval. In this work, we propose a novel solution to thoroughly investigate the language-based video moment retrieval issue under the adversarial learning. The key of our solution is to formulate the language-based video moment retrieval task as an adversarial learning problem with two tightly connected components. Specifically, a reinforcement learning is employed as a generator to produce a set of possible video moments. Meanwhile, a multi-task learning is utilized as a discriminator, which integrates inter-modal and intra-modal in a unified framework by employing a sequential update strategy. Finally, the generator and the discriminator are mutually reinforced in the adversarial learning, which is able to jointly optimize the performance of both video moment ranking and video moment localization. Extensive experimental results on two challenging benchmarks, i.e., Charades-STA and TACoS datasets, have well demonstrated the effectiveness and rationality of our proposed solution. Meanwhile, on the larger and unbiased datasets, i.e., ActivityNet Captions and ActivityNet-CD, our proposed framework exhibits excellent robustness.

In order to reduce the power consumption to extend the lifetime of battery-powered Mobile Stations (MSs), IEEE 802.16e has introduced a sleep mode to decrease the usage of air interface resources of serving Base Stations (BSs). In the sleep mode, MS enters into a sleep cycle after negotiating with serving BS and stops its traffic service temporarily. In order to reduce the energy consumption as well as decrease the packet delay especially in low traffic scenario, this paper proposes a novel power saving algorithm for sleep mode in which the sleep interval is increased in power function each time until a maximum sleep interval threshold is reached. Theoretical analysis and simulation experiments demonstrate that the proposed algorithm can achieve significantly less mean energy consumption and mean response time compared with exponential and linear increase algorithms. Furthermore, this paper analyzes the performance impact of each parameter in the algorithm and presents how to configure the sleep mode parameters.

Application layer multicast (ALM) has been proposed as a promising alternative to the IP multicast due to the difficulties in deploying network layer multicast in an operational network at the Internet scale. However, most of existing application layer multicast approaches build their overlay networks based on simple measurement with the end-to-end delay without considering the physical network topology and IP multicast deploying in local scale on the Internet. Thus, they might suffer some inefficiency such as long join delay, large maintenance overhead, and little flexibility to extend. In this paper we propose a novel hybrid multicast approach named SHM, which combining ALM and IP multicast to achieve ubiquitous multicast delivery. Simulation results show that SHM is able to provide more efficient multicast with less maintenance overhead, higher flexibility and reliability.