Search

Training large-scale models needs big data. However, the few-shot problem is difficult to resolve due to inadequate training data. It is valuable to use only a few training samples to perform the task, such as using big data for application scenarios due to cost and resource problems. So, to tackle this problem, we present a simple and efficient method, contrastive label generation with knowledge for few-shot learning (CLG). Specifically, we: (1) Propose contrastive label generation to align the label with data input and enhance feature representations; (2) Propose a label knowledge filter to avoid noise during injection of the explicit knowledge into the data and label; (3) Employ label logits mask to simplify the task; (4) Employ multi-task fusion loss to learn different perspectives from the training set. The experiments demonstrate that CLG achieves an accuracy of 59.237%, which is more than about 3% in comparison with the best baseline. It shows that CLG obtains better features and gives the model more information about the input sentences to improve the classification ability. [ABSTRACT FROM AUTHOR]

Complex tasks in the real world involve different modal models, such as visual question answering (VQA). However, traditional multimodal learning requires a large amount of aligned data, such as image text pairs, and constructing a large amount of training data is a challenge for multimodal learning. Therefore, we propose VL-Few, which is a simple and effective method to solve the multimodal few-shot problem. VL-Few (1) proposes the modal alignment, which aligns visual features into language space through a lightweight model network and improves the multimodal understanding ability of the model; (2) adopts few-shot meta learning in the multimodal problem, which constructs a few-shot meta task pool to improve the generalization ability of the model; (3) proposes semantic alignment to enhance the semantic understanding ability of the model for the task, context, and demonstration; (4) proposes task alignment that constructs training data into the target task form and improves the task understanding ability of the model; (5) proposes generation alignment, which adopts the token-level training and multitask fusion loss to improve the generation ability of the model. Our experimental results show the effectiveness of VL-Few for multimodal few-shot problems. [ABSTRACT FROM AUTHOR]

Multimodal learning is a promising area in artificial intelligence (AI) that can make the model understand different kinds of data. Existing works are trying to re-train a new model based on pre-trained models that requires much data, computation power, and time. However, it is difficult to achieve in low-resource or small-sample situations. Therefore, we propose VL-Meta, Vision Language Models for Multimodal Meta Learning. It (1) presents the vision-language mapper and multimodal fusion mapper, which are light model structures, to use the existing pre-trained models to make models understand images to language feature space and save training data, computation power, and time; (2) constructs the meta-task pool that can only use a small amount of data to construct enough training data and improve the generalization of the model to learn the data knowledge and task knowledge; (3) proposes the token-level training that can align inputs with the outputs during training to improve the model performance; and (4) adopts the multi-task fusion loss to learn the different abilities for the models. It achieves a good performance on the Visual Question Answering (VQA) task, which shows the feasibility and effectiveness of the model. This solution can help blind or visually impaired individuals obtain visual information. [ABSTRACT FROM AUTHOR]

With the application of graph neural network (GNN) in the communication physical layer, GNN-based channel decoding algorithms have become a research hotspot. Compared with traditional decoding algorithms, GNN-based channel decoding algorithms have a better performance. GNN has good stability and can handle large-scale problems; GNN has good inheritance and can generalize to different network settings. Compared with deep learning-based channel decoding algorithms, GNN-based channel decoding algorithms avoid a large number of multiplications between learning weights and messages. However, the aggregation edges and nodes for GNN require many parameters, which requires a large amount of memory storage resources. In this work, we propose GNN-based channel decoding algorithms with shared parameters, called shared graph neural network (SGNN). For BCH codes and LDPC codes, the SGNN decoding algorithm only needs a quarter or half of the parameters, while achieving a slightly degraded bit error ratio (BER) performance. [ABSTRACT FROM AUTHOR]

Channel neural decoding is very promising as it outperforms the traditional channel decoding algorithms. Unfortunately, it still faces the disadvantage of high computational complexity and storage complexity compared with the traditional decoding algorithms. In this paper, we propose that low rank decomposition techniques based on tensor train decomposition and tensor ring decomposition can be utilized in neural offset min-sum (NOMS) and neural scale min-sim (NSMS) decoding algorithms. The experiment results show that the proposed two algorithms achieve near state-of-the-art performance with low complexity. [ABSTRACT FROM AUTHOR]

A novel single-carrier-based space-time code construction scheme to exploit the advantages of a frequency-selective fading channel is investigated in this paper. The proposed construction scheme is based on multiplexing independent streams of phase-rotated space-time codes in a time-interleaved fashion. The advantage of such design is that it guarantees full space-time-multipath diversity by using traditional space-time codes or MIMO signaling schemes originally designed for flat fading channels as the constituent codes. Another advantage is that this approach incurs no loss in bandwidth efficiency and it alleviates the problem of high PAPR in OFDM-based space-time codes. By employing random or algebraic rotations, the design is potentially suitable for any number of transmit antennas or multipaths. The simulation results indicate that full space-time-multipath diversity is attained using this new approach, and comparisons with some existing space-time codes designed for frequency-selective channels are made to show its performance advantage. [ABSTRACT FROM AUTHOR]

Reconfigurable intelligent surface (RIS) has been envisioned as one of the promising solutions for enhancing signal transmissions in high-speed communications (HSC). In this paper, we present a time-varying channel model with distance-dependent Rician factors for the RIS-assisted HSC. Our model not only contains Rayleigh components and Doppler shift (DS) terms but also distance-dependent Rician factors, for characterizing time-varying features. In particular, we show that when the vehicle is far from the base station and the RIS, the channel contains only Rayleigh fading. However, when they are close enough, the channel can be considered as a light-of-sight channel. Based on the proposed model, it is proven that using RIS phase shift optimization, the DS of the cascaded links can be aligned with the DS of the direct link; and if the direct link is blocked, the DS can be removed entirely. Furthermore, we derive the closed-form expressions for the ergodic spectral efficiency and the outage probability of the proposed system. Besides, it is observed that the deployment strategy also affects the system performance. Simulation results validate all analyses. [ABSTRACT FROM AUTHOR]

Cognitive Radio (CR) is a practical technique for overcoming spectrum inefficiencies by sensing and utilizing spectrum holes over a wide spectrum. In particular, cooperative spectrum sensing (CSS) determines the state of primary users (PUs) by cooperating with multiple secondary users (SUs) distributed around a Cognitive Radio Network (CRN), further overcoming various noise and fading issues in the radio environment. But it's still challenging to balance energy efficiency and good sensing performances in the existing CSS system, especially when the CRN consists of battery-limited sensors. This article investigates the application of machine learning technologies for cooperative spectrum sensing, especially through solving a multi-dimensional optimization that cannot be readily addressed by traditional approaches. Specifically, we develop a neural network, which involves parameters that are integral to the CSS performance, including a device sleeping rate for each sensor and thresholds used in the energy detection method, and a customized loss function based on the energy consumption of the CSS system and multiple penalty terms reflecting the system requirements. Using this formulation, energy consumption is to be minimized with the guarantee of reaching a certain probability of false alarm and detection in the CSS system. With the proposed method, comparison studies under different hard fusion rules ('OR' and 'AND') demonstrate its effectiveness in improving the CSS system performances, as well as its robustness in the face of changing global requirements. This paper also suggests the combination of the traditional and the proposed scheme to circumvent the respective inherent pitfalls of neural networks and the traditional semi-analytic methods. [ABSTRACT FROM AUTHOR]

In this paper, we consider a multiple reconfigurable intelligent surface (RIS)-assisted system using positioning information (PI) to explore the potential of Doppler effect mitigation and spectral efficiency (SE) enhancement in high-speed communications (HSC) in the presence of hardware impairments (HWI). In particular, we first present a general multi-RIS-assisted system model for HSC with HWI. Then, based on PI, different phase shift optimization strategies are designed and compared for maximizing SE, eliminating Doppler spread, and maintaining a very low delay spread. Moreover, we compare the performance of different numbers of RISs with HWI in terms of SE and delay spread. Finally, we extend our channel model from line-of-sight to the Rician channel to demonstrate the effectiveness and robustness of our proposed scheme. Numerical results reveal that the HWI of RISs increases the delay spread, but has no impact on Doppler shift and spread. Additionally, the multiple RIS system not only suffers a more severe delay spread, but is limited in SE due to the HWI. When the number of RISs increases from 2 to 16, the range of average spectral efficiency and delay spread are from 4 to 4.6 Bit/s/Hz and from 0.7 μs to 2.5 μs, respectively. In contrast to conventional RIS-assisted systems that require channel estimation, the proposed PI-based RIS system offers simplicity without compromising effectiveness and robustness in both SE enhancement and Doppler mitigation. [ABSTRACT FROM AUTHOR]