The time-index models are a class of time series forecasting models that map time-index features to forecasts in continuous space. Compared to the historical-value models, the time-index models can avoid the effect of data sampling frequency and are usually more expressive. However, the vanilla deep time-index model is weak in modeling the high-frequency components of time series and often requires the introduction of many parameters to enhance the modeling capability. Moreover, the time-index model learns only a mapping relationship and ignores the sequence relationship between temporal features, leading to a weak extrapolation capability in the forecast horizon. In this paper, inspired by the ability of quantum implicit neural representations to model the high-frequency components of signals with fewer parameters, we propose Quantum Time-Index Models with Reservoir (QuantumTime). Specifically, we introduce variational quantum circuits to address the challenge of representing high-frequency components in time series. Then, we introduce a reservoir that empowers QuantumTime with powerful extrapolation capabilities by exploiting the rich dynamical properties of reservoir computing. Ultimately, experiments conducted on chaotic datasets and various real-world datasets demonstrate that QuantumTime achieves highly competitive results compared to the state-of-the-art deep time-index model while reducing training parameters by at least 95%. Our approach provides a paradigm for utilizing potential quantum advantage in practical tasks.

Enabling machines with the capability to recognize and comprehend metaphors is a crucial step toward achieving artificial intelligence. In linguistic theories, metaphor can be identified through Metaphor Identification Procedure (MIP) or Selectional Preference Violation (SPV), both of which are typically considered as matching tasks in the field of natural language processing. However, the implementation of MIP poses a challenge due to the semantic uncertainty and ambiguity of literal meanings of words. Simultaneously, SPV often struggles to recognize conventional metaphors. Inspired by Quantum Language Model (QLM) for modeling semantic uncertainty and fine-grained feature matching, we propose a quantum-inspired matching network for metaphor detection. Specifically, we use the density matrix to explicitly characterize the literal meanings of the target word for MIP, in order to model the uncertainty and ambiguity of the literal meanings of words. This can make SPV effective even in the face of conventional metaphors. MIP and SPV are then achieved by fine-grained feature matching. The results of the experiment finally demonstrated our approach has strong competitiveness.

Quantum machine learning aims to leverage quantum computing to enhance the computing capabilities and storage efficiency of classical machine learning. Among them, quantum generative models, as a type of unsupervised machine learning, have been proven to learn distributions that are outside of classical machine learning reach. However, there is limited work on how to truly utilize and validate a possible quantum advantage in practical applications, especially in the field of natural language processing. This paper proposes the first Quantum Topic Model (QTM) based on variational quantum circuits to validate and showcase the possible advantages of quantum computing for generative models. QTM not only generates random samples like generative models but also exhibits fitting capabilities similar to neural networks. Finally, we conducted experiments on real datasets, and the results showed that QTM achieved a 46.9% reduction in model parameters while obtaining 10.4% higher topic coherence and 0.2% higher topic diversity with fewer iterations than neural topic models.

Implicit neural representations have emerged as a powerful paradigm to represent signals such as images and sounds. This approach aims to utilize neural networks to parameterize the implicit function of the signal. However, when representing implicit functions, traditional neural networks such as ReLU-based multilayer perceptrons face challenges in accurately modeling high-frequency components of signals. Recent research has begun to explore the use of Fourier Neural Networks (FNNs) to overcome this limitation. In this paper, we propose Quantum Implicit Representation Network (QIREN), a novel quantum generalization of FNNs. Furthermore, through theoretical analysis, we demonstrate that QIREN possesses a quantum advantage over classical FNNs. Lastly, we conducted experiments in signal representation, image superresolution, and image generation tasks to show the superior performance of QIREN compared to state-of-the-art (SOTA) models. Our work not only incorporates quantum advantages into implicit neural representations but also uncovers a promising application direction for Quantum Neural Networks.

In the era of widespread dissemination through social media, the task of rumor detection plays a pivotal role in establishing a trustworthy and reliable information environment.Nonetheless, existing research on rumor detection confronts several challenges: the limited expressive power of text encoding sequences, difficulties in domain knowledge coverage and effective information extraction with knowledge graph-based methods, and insufficient mining of semantic structural information.To address these issues, we propose a Crowd Intelligence and ChatGPT-Assisted Network(CICAN) for rumor classification.Specifically, we present a crowd intelligence-based semantic feature learning module to capture textual content's sequential and hierarchical features.Then, we design a knowledge-based semantic structural mining module that leverages ChatGPT for knowledge enhancement.Finally, we construct an entity-sentence heterogeneous graph and design Entity-Aware Heterogeneous Attention to integrate diverse structural information metapaths effectively.Experimental results demonstrate that CICAN achieves performance improvement in rumor detection tasks, validating the effectiveness and rationality of using large language models as auxiliary tools.

With the complexity of patent transformation scenarios, the roles of users have become more diverse. Therefore, how to discover the roles of different users in the patent transformation scenarios has become a hot issue. In the process of patent transformation, the behaviors of each user are regular, historical behavior has an impact on the current behavior. Because the Hawkes processes can take into account the characteristic of self-exciting among behaviors, we explored the Dirichlet Mixture model of Hawkes Processes based on variational inference to cluster users for user roles discovery. In this model, different Hawkes processes correspond to different user types. Dirichlet distribution is used as the prior distribution of user clusters. The dependence of current behavior on historical behavior is expressed as intensity function. The variational inference is used to learn the model. The model is evaluated by Precision, Recall and F-measure, which shows that our model has good accuracy.

With the growing importance of intellectual property, the amount of patent increases every year. The patents realize their values by the patent conversion. However, many patents do not realize their values since the paths to realize the patent value have not been found. To predict the paths, we explore a Bayesian neural network based model. In the model, the patents are represented by the function-effects, from which some technical features are extracted. We use Bayesian neural network to predict the paths toward the realization of patent valuation. The model is evaluated by the evaluation measurements. The results show our method performs well in the evaluation measurements. Such model can be applied to further patent recommendation and automated trading.

With the intense competition of global intellectual property, the increasing patents promote the potentiality of patent transactions. Patent valuation is the premise of the patent transaction. Automatic patent valuation faces some challenging issues from valuation feature to valuation model. To solve the above issues, we propose a Bayesian graph convolutional neural network based patent valuation model. In the model, the valuation objects are defined, from which to some valuation features are extracted. Valuation scenario is the constructed, on which Bayesian graph convolutional neural network is used to generate patent value. We evaluate our model by comparing the state-of-the-art model on patent data sets. The results show that our model outperforms other models in the evaluation measurements.

Network coding is being viewed to have the potential for significant throughput improvement in network environment. However, these expected benefits are very fragile to malicious attacks, including message block content corruption and node compromise attacks. To solve these problems, both pollution detection and pollution correction based schemes have been proposed. These schemes are only effective in some limited scenarios. In this paper, we propose a new scheme that combines the benefits of the existing error-detection and error-correction (EDEC) schemes. The proposed scheme is similar in structure to the existing error-control based schemes. However, by appropriately modifying the rate of the underlying error-control scheme, we can improve the network throughput and robustness significantly. Our scheme can detect the malicious attacks by computing whether the syndromes are all zeros. By collecting all the non-zero syndromes, the malicious attacks within the error-decoding capacity of the underlying linear network coding can be removed and the original message can be recovered. Our theoretical analysis and simulation results demonstrate that the proposed EDEC scheme can improve the overall network performance dramatically with only a very moderate increase of the computational overhead.