This paper presents an innovative approach to enhancing neural network performance through the development and implementation of an adaptive activation function, termed adaptive activation (AdAct). AdAct amalgamates various well-established and novel activation functions into a single, learnable framework, allowing dynamic adaptation to specific network layers' needs. We explore the effectiveness of ReLU and its variants, including ELU, LReLU, PReLU, RReLU, and more recent functions like Swish and Mish, integrating them into the AdAct function. Employing ConvNet variants across FMNIST, CIFAR10, SVHN and FER datasets, our study empirically assesses each function's contribution and demonstrates AdAct's potential in optimizing neural networks, especially in selecting optimal activation functions for diverse tasks.
In the context of the Industrial Internet of Things (IIoT), multiple access edge computing (MEC) enables the provision of computational resources closer to users. The work presented in this paper explores a common IIoT application scenario wherein autonomous mobile robots (AMR) operate in a MEC-enabled network and depend on multimedia content stored at the MEC servers over the course of operation. The age of information (AoI) metric is used to measure the freshness of the cached content from the perspective of the AMR. At the same time, the energy cost associated with refreshing the cache files is simultaneously considered as well. This paper delves into achieving an optimal trade-off between minimizing the weighted AoI cost and the energy expended for cache refreshing. We address this problem by introducing a cache refreshing-deep deterministic policy gradient (CR-DDPG) algorithm, a model-free deep reinforcement learning method, to optimize both AoI and energy usage. Various simulation studies are conducted to evaluate the proposed CR-DDPG algorithm, and the results demonstrate that CR-DDPG consistently outperforms its baseline counterparts, rendering it a robust approach for cache-refreshing in dynamic IIoT environments.ย
An important role played by industrial Internet of Things (IIoT) networks is supporting the operations of autonomous mobile robots (AMR) by leveraging multiple-access edge caching servers. At the same time, judicious content caching strategies are essential for minimizing content retrieval delays and the costs associated with updating caches. In this study, a novel multi-agent reinforcement learning (MARL)-based cache update strategy termed multi-agent cache update (MACU) is proposed. MACU leverages the multi-agent deep deterministic policy gradient (MADDPG) framework and aims to optimize cache updates to reduce AoI, minimize events where AoI exceeds acceptable levels, and ensure that the costs associated with performing cache updates are kept low. Furthermore, to mitigate the computational complexity of training agents with MACU, the MACU with Global Critic (MACU-GC) variant is introduced, which diverges from traditional MADDPG by employing a singular global critic for enhanced training efficiency. Extensive numerical evaluations showcase the proposed strategiesโ superiority over conventional DRL-based caching methods, achieving significant improvements in AoI costs, caching costs, and AoI violation costs, while effectively reducing content retrieval latency, enhancing hit rates, and optimizing AoI and link load metrics.
In this work, we propose SVDNet, a novel deep learning (DL) architecture that utilizes singular value decomposition (SVD), for transmit power control in a multiuser multiple input multiple output (MU-MIMO) system. We propose a novel method of training SVDNet in a supervised manner for the power control task by using binary cross-entropy loss functions. SVDNet requires fewer computations than traditional power control algorithms such as weighted minimum mean squared error (WMMSE). Our simulation results show that the proposed SVDNet provides over a 50% increase in sum-rate performance as compared to similar supervised DL-based power control schemes while being significantly more computationally efficient than WMMSE.
This paper presents Multinet, an unsupervised deep learning (DL) approach for power allocation in industrial environments and IIoT applications. Multinet extends the previously proposed singular value decomposition network (SVDNet), which utilizes supervised DL to approximate the performance of the WMMSE algorithm. While SVDNet requires labeled data for training, limiting its scalability and generalization performance, in contrast, Multinet employs unsupervised DL to directly optimize the sum rate maximization objective function, eliminating the need for labeled datasets and improving training efficiency. Simulation studies are conducted to evaluate Multinetโs performance in an industrial environment, utilizing parameters derived from measured large-scale fading characteristics of the industrial radio channel at 5200 MHz. The suitability of Multinet for industrial applications is thus assessed and numerical evaluations demonstrate that Multinet outperforms benchmark supervised and unsupervised DL-based power control schemes in terms of sum rate and energy efficiency.
Skin cancer is the fifth most common cancer across the globe. One in every three cancers detected is skin cancer. Every year, between 2 and 3 million non-melanoma skin cancers and 132,000 melanoma skin cancers occur globally each year. The worldwide number of cases of carcinoma is increasing by epidemic proportions. Catching cancer early often allows for more treatment options. It is crucial to know which type of cancer one has because it affects the treatment options and prognosis. Hence, our work is primarily motivated by the importance of early and efficient detection of skin cancer. In this paper, a novel methodology has been proposed by incorporating Neural Architecture Search and Model Quantization technique, where deep convolutional neural network is utilized to build a skin cancer classifier, which can be deployed on low compute devices. To the best of our knowledge this is the first instance of utilizing Neural Architecture Search in combination with Model Quantization for the purpose of Skin Cancer detection. The yielded results show that, from the model proposed through this methodology, the model size can be reduced without causing a significant change in the accuracy. Thus, making this type of a skin classifier a more effective tool for skin cancer detection allows it for the usage on devices with limited memory and computational capacity.
In the data-driven world, emerging technologies like the Internet of Things (IoT) and other crowd-sourced data sources like mobile devices etc. generate a tremendous volume of decentralized data that needs to be analyzed for obtaining useful insights, necessary for reliable decision making. Although the overall data is rich, contributors of such kind of data are reluctant to share their own data due to serious concerns regarding protection of their privacy; while those interested in harvesting the data are constrained by the limited computational resources available with each participant. In this paper, we propose an end-to-end algorithm that puts in coalescence the mechanism of learning collaboratively in a decentralized fashion, using Federated Learning, while preserving differential privacy of each participating client, which are typically conceived as resource-constrained edge devices. We have developed the proposed infrastructure and analyzed its performance from the standpoint of a machine learning task using standard metrics. We observed that the collaborative learning framework actually increases prediction capabilities in comparison to a centrally trained model (by 1-2%), without having to share data amongst the participants, while strong guarantees on privacy (ฮต, ฮด) can be provided with some compromise on performance (about 2-4%). Additionally, quantization of the model for deployment on edge devices do not degrade its capability, whilst enhancing the overall system efficiency.