Individual next location prediction plays a crucial role in location-based applications, such as route navigation and service recommendation. Although the existing research based on deep learning effectively captures users' spatiotemporal travel preferences, there are challenges in the interpretability of location prediction, heavily relying on large-scale historical travel data for model training. Drawing inspiration from the powerful reasoning capabilities of large language models (LLMs), this study proposes a novel multiround continuous dialogue mechanism and candidate set enhancement method, leveraging LLMs for next location prediction through step-by-step reasoning. In the first round of dialogue, we introduce activity prediction as an auxiliary task to narrow down the candidate locations. Subsequently, we establish an activity-aware prompt to enable LLM to achieve accurate location prediction and provide corresponding reasoning. Finally, we incorporate a third round of dialogue to prompt LLM to make necessary corrections by integrating the prediction results of deep learning models. To address the issues of LLMs being affected by element ranking within the candidate set, we propose a new candidate set enhancement method based on the entropy-weighted Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS). Our model can understand user travel preferences by fusing location, activity, and time information through natural language. Extensive experiments are conducted on two public datasets of user check-ins, and the results show that our model achieves prediction performance comparable to deep learning models in full-sample prediction and outperforms them in the few-shot settings. Our model provides logical and explainable reasoning, offering insightful guidance for downstream application tasks.

Exploiting the error tolerance in computation, approximate circuits become an emerging computing paradigm to increase the energy efficiency in digital systems, which is crucial in high-performance and low-power systems for the edge Internet-of-Things (EIoT) devices. Inspired by the state-of-the-art high-efficiency NN accelerators, three techniques are proposed for effectively integrating the approximate computing unit into CNN accelerator to achieve a dynamic energy-accuracy trade-off: (1) An approximate multiplier that can be configured to three precision modes is proposed. A weight pre-encoding method is used to save hardware overhead. (2) For hybrid-accuracy layer-wise mapping, the hessian-aware layer-wise accuracy scaling is proposed, which concerns inference accuracy and hardware overhead simultaneously. A progressive re-training approach is proposed to enable an aggressive approximation configuration and higher power reduction. (3) A tensor multiplication unit (TMU) with timing error detection and correction (TEDC) approach is proposed, enabling an aggressive voltage scaling and a 41.5% power reduction is obtained. An energy-efficient CNN accelerator is proposed and shows how deep learning can be brought to EIoT devices by running each layer at its appropriate computational accuracy. Implemented under 28-nm CMOS technology, the CNN accelerator achieves the energy efficiency of 14.4 TOPS/W. The proposed accelerator and method are conducted on the applications of keyword spotting of GSCD, CIFAR10 and CIFAR100, 44.5%~46.7% multiplication energy is saved while reducing the accuracy by less than 0.6%.

Neural network (NN) operation has high requirements for storage resources and parallel computing, which bring huge challenges to the deployment of NNs in Internet-of-Things (IoT) devices. Consequently, this work proposed a low-power NN architecture, comprising an energy-efficient NN processor and a Cortex-M3 host processor to achieve state-of-the-art (SOTA) end-to-end inference at the edge. The innovations of this article are as follows: 1) to minimize the bit width of the weight while keeping the loss of accuracy within a small range, cross-layer error tolerance has been analyzed, and mixed precision quantization has been adopted for cross-layer mapping; 2) dynamic reconfigurable tensor processing unit (DR-TPU) with approximate computing has been proposed, which brings$1.45\times $computing energy reduction within 0.46% accurate loss in ResNet-50; and 3) a customized input feature map (IFM) reuse and over-writeback strategy has been adopted, eliminating the recurrent fetching from the on-chip and off-chip memories. The times of on-chip storage access can be reduced by 25%–60%, and the capacity of on-chip memory can be reduced to half of the original. The processor has been implemented at 28-nm CMOS technology. Combining the above work, the proposed architecture can achieve a 53.1% reduction of power and 17.2-TOPS/W energy efficiency.

Influence maximization (IM) aims to find a subset of$k$nodes that can maximize the final active node set under an information diffusion model. With the development and popularity of social networks, the IM problem plays an essential role in various applications, such as public opinion analysis, viral marketing, and rumor early warning. However, most of the existing IM solutions have not accessed the risk of negative information from nodes in the future. In reality, a company may face economic loss when negative information breaks out of its spokesman. Therefore, a crisis assessment oriented and topic-based IM problem (TIM-CA) is proposed, which is utilized to model the IM problem by considering the crisis assessment (CA) and topics of users. To solve this problem, we propose a maximum influence arborescence model-based algorithm for TIM-CA, namely, MIA-TIM-CA. The proposed algorithm consists of crisis degree calculation, topic relevance calculation, and influence spread evaluation. More importantly, as for crisis degree calculation, it considers self-, topology-, and topic-based crisis degrees for each node. At the influence spread evaluation stage, MIA-TIM-CA proposes two functions to evaluate the node’s importance and node influence spread. Extensive experiments on two real-world social networks demonstrate that our MIA-TIM-CA outperforms all comparison algorithms on influence spread, crisis score, and running time.

As the artificial intelligence and Internet of Things (AIoT) develop rapidly, the deployment of artificial neural networks in edge computing is becoming significant with great challenge. The binarized convolutional neural network (BCNN) is one of the most widely adopted light-weight ANNs in AIoT, which can achieve the balance of system accuracy and hardware resource consumption, compared to others. To achieve high power and area efficiency in BCNN deployment, many approximate computing (AxC) techniques are integrated to make full use of the resilience of BCNN. As the research focused on the integration of AxC in circuit design, the design of AxC itself is not fully considered when applied to specific applications or domains. Based on circuit-architecture-system co-design, this article proposes an efficient BCNN deployment method, including a quality-circuit co-design method for approximate adder generation, a quality-aware intercompensation approach for addition tree, and a computing quality involved retraining approach for BCNN deployment. Experimental results show that the proposed quality model can achieve 86.43% in average accuracy while evaluating nine types of typical approximate adders. The proposed method is conducted on the applications of keyword spotting of GSCD, MNIST, and CIFAR-10, and we can further rise the approximation degree by 50%–75%, while reducing the accuracy by less than 1%.

The adaptive hinging hyperplane (AHH) model is a popular piecewise linear representation with a generalized tree structure and has been successfully applied in dynamic system identification. In this article, we aim to construct the deep AHH (DAHH) model to extend and generalize the networking of AHH model for high-dimensional problems. The network structure of DAHH is determined through a forward growth, in which the activity ratio is introduced to select effective neurons and no connecting weights are involved between the layers. Then, all neurons in the DAHH network can be flexibly connected to the output in a skip-layer format, and only the corresponding weights are the parameters to optimize. With such a network framework, the backpropagation algorithm can be implemented in DAHH to efficiently tackle large-scale problems and the gradient vanishing problem is not encountered in the training of DAHH. In fact, the optimization problem of DAHH can maintain convexity with convex loss in the output layer, which brings natural advantages in optimization. Different from the existing neural networks, DAHH is easier to interpret, where neurons are connected sparsely and analysis of variance (ANOVA) decomposition can be applied, facilitating to revealing the interactions between variables. A theoretical analysis toward universal approximation ability and explicit domain partitions are also derived. Numerical experiments verify the effectiveness of the proposed DAHH.

Most spatial crowdsourcing systems are designed in a static mode with tasks allocated based on the historical interactions data between crowd participants and crowdsourcing tasks. However, these task assignment algorithms usually ignore the long-term feedback on interactive spatial crowdsourcing systems, resulting in performance degradation. Though reinforcement learning naturally fits the problem of maximizing long term crowdsourcing rewards, deep reinforcement learning-based task assignment is still facing the challenge of interactive spatial crowdsourcing. To address these issues, this paper investigates a challenge problem which we study how to intelligently task assignments for interactive spatial crowdsourcing applications. Therefore, we develop an advanced Embedding-based Deterministic Policy Gradient learning framework to maximize long term crowdsourcing rewards for task assignments, called EDPG-Assignment. EDPG-Assignment is based on deep actor critic learning and combines the improvements of two advanced methods, action embedding and neighbor-based deep Deterministic Policy Gradient, and employed this to optimize the task assignment in interactive crowdsourcing. A matrix factorization method to learn spatial crowdsourcing action embedding for neighbor-based discrete actions similarities evaluation in deep actor critic learning-based task assignment from generated crowdsourcing trajectories without any prior knowledge. The EDPG-Assignment algorithm provided a more stable learning process and showed improved results in real-world dataset.