Deep learning applications in image classification often rely on extensive training data. However, in specific scenarios, obtaining a substantial amount of training data poses significant challenges. Training models with limited data often leads to overfitting and poor generalization performance. To address this issue, we propose a novel method based on map activation quantization for image classification. Our method quantizes spatial activations of latent representations, modeling the relationship between discrete representations and categories. This effectively captures variations in object properties, providing more reliable classification information and mitigating overfitting in scenarios with limited data. Furthermore, our approach seamlessly integrates with any convolutional neural network model without necessitating alterations to existing architectures or methodologies. Empirical evaluations on natural and medical image datasets demonstrate the superiority of our method in image classification under limited data scenarios, establishing a new technological benchmark. We intend to publicly release the relevant source code.

It is challenging to train a generalizable deep learning classifier with limited training images. Existing few-shot learning approaches try to improve classification performance largely by transferring prior knowledge from upstream large-sample tasks to the current small-sample task. Besides upstream image datasets, prior knowledge may also be obtained from signals of other modalities. In this study, we propose a novel learning framework that can utilize prior knowledge from audio signals to help train an image classifier. In the framework, a pre-trained and fixed audio encoder can transform the audio signal of each class label into a class-specific audio prototype. By attracting image representations to the corresponding audio prototypes during training of the image classifier, within-class image representations become more clustered, while image representations become further apart if they are from different classes. To the best of our knowledge, this is the first work that utilizes audio-based prior knowledge to help train an image classifier with limited training images. The proposed learning framework is compatible with existing learning approaches, making it flexible enough to be combined with existing approaches. Extensive empirical evaluations on both natural and medical image datasets demonstrate that the proposed learning framework significantly outperforms existing methods in image classification with limited training images, thus establishing a new state of the art. The source code will be released publicly.

In the ever-evolving realm of wireless communications, the integration of intelligent reflecting surface (IRS) and mobile edge computing for task offloading has ignited extensive curiosity. However, previous research primarily concentrated on task offloading assuming that a user equipment (UE) possesses the system channel information and the computing resources of base stations (BSs). Acknowledging that UEs typically lack access to such system information, and in pursuit of equilibrium in computing loads between UEs and BSs, we present a bilateral online task allocation approach grounded in partial offloading to minimize task completion latency. Specifically, addressing uncertainties in channel information and available computing resources at BSs, we employ the online ridge regression method on the UE side to continuously adjust the task allocation proportion for offloading to BSs. On the IRS side, we formulate the BS selection as a multi-armed bandit problem, proposing an online learning algorithm based on Thompson sampling to determine the set of BSs for edge computing while managing the task allocation among the selected BSs. Simulation results demonstrate the superior performance of our algorithm.

The integrated access and backhaul (IAB) technology enables cost-effective ultradense network deployment, by replacing wired transmission infrastructure. However, IAB is hindered by high energy consumption, fragile wireless links, and coverage holes. With the capability of enhancing coverage and smart configuration of the radio environment at a low cost, intelligent reflecting surfaces (IRSs) can well address the challenges in IAB. This paper focuses on forming heterogeneous communication between base station (BS) and IRS in large-scale IAB networks, achieving percolation-based connectivity of BSs to ensure further cost-effective BSs deployment and successful data packet delivery between users. The dependency between the two node processes and the asymmetry of communication links introduce new challenges for connectivity in IRS-assisted IAB networks. To address these challenges, continuum percolation theory is applied to scrutinize the topological and analytical properties of the IRS-assisted wireless network. Specifically, the analysis demonstrates the uniqueness of the infinite connected component and proves the topological connectivity within the corresponding connectivity region. Additionally, the paper establishes the necessary and sufficient conditions for achieving network connectivity manifested in the critical densities of both BS and IRS nodes. The theoretical analysis is validated through simulations, confirming alignment between derived bounds and Monte Carlo results.

Mobile edge computing offers a new paradigm to provide more convenient computing services for mobile devices. However, the mobility of devices and the limited coverage of edge servers bring considerable challenges to efficient computation offloading. Moreover, tasks with temporal dependency further complicate the offloading problem in the mobile edge network. In this paper, we take into account the mobility of devices and the fine-grained tasks generated by the mobile device to make full use of the computing resources of devices and edge servers. Considering the temporal dependency among tasks, the offloading problem is formulated as a mixed integer programming which achieves the tradeoff between time latency and energy consumption. Simulation results demonstrate that our proposed algorithm can achieve a significant improvement in terms of energy efficiency and latency compared with other bench mark algorithms.