Deep Neural Networks (DNNs) have shown significant advantages in a wide variety of domains. However, DNNs are becoming computationally intensive and energy hungry at an exponential pace, while at the same time, there is a vast demand for running sophisticated DNN-based services on resource constrained embedded devices. In this paper, we target energy-efficient inference on embedded DNN accelerators. To that end, we propose an automated framework to compress DNNs in a hardware-aware manner by jointly employing pruning and quantization. We explore, for the first time, per-layer fine- and coarse-grained pruning, in the same DNN architecture, in addition to low bit-width mixed-precision quantization for weights and activations. Reinforcement Learning (RL) is used to explore the associated design space and identify the pruning-quantization configuration so that the energy consumption is minimized whilst the prediction accuracy loss is retained at acceptable levels. Using our novel composite RL agent we are able to extract energy-efficient solutions without requiring retraining and/or fine tuning. Our extensive experimental evaluation over widely used DNNs and the CIFAR-10/100 and ImageNet datasets demonstrates that our framework achieves 39% average energy reduction for 1.7% average accuracy loss and outperforms significantly the state-of-the-art approaches.

A new methodology for the RF/mmWave analog design process, automation and acceleration, is presented in this work. The proposed framework was implemented so as to accelerate the design cycle of analog/RF circuits by creating a dataset in a fully automated manner and training a combination of machine learning models for the optimal design parameters’ prediction. machine learning polynomial regression was adopted to accelerate the design process, predicting the optimal design parameters’ values while genetic algorithm optimization was exploited for the dataset creation automation. To evaluate the efficiency of the proposed methodology, the framework was implemented for the design of a common source Low-Noise-Amplifier, using a 65 nm CMOS process node. The proposed methodology successfully tackles the design cycle speed-up, automation, and acceleration, utilizing machine learning prediction for the design parameters and genetic algorithm for the dataset creation automation instead of the classical, simulation-based, standard design methodology. The provided experimental results have shown the effectiveness of the proposed hybrid approach, creating very precise RF matching networks for LNA designs and achieving >$99.9\%$ wave transmission efficiency while reaching >$99\%$ accuracy on the parameters’ prediction task.

Honey bees play a vital role in ecosystem stability, and the need to monitor colony health has driven the development of IoT-based systems in beekeeping, with recent studies exploring both empirical and machine learning approaches to detect and analyze key hive conditions. In this study, we present an IoT-based system that leverages sensors to record and analyze the acoustic signals produced within a beehive. The captured audio data is transmitted to the cloud, where it is converted into mel-spectrogram representations for analysis. We explore multiple data pre-processing strategies and machine learning (ML) models, assessing their effectiveness in classifying queenless states. To evaluate model generalization, we apply transfer learning (TL) techniques across datasets collected from different hives. Additionally, we implement the feature extraction process and deploy the pre-trained ML model on a deep edge IoT device (Arduino Zero). We examine both memory consumption and execution time. The results indicate that the selected feature extraction method and ML model, which were identified through extensive experimentation, are sufficiently lightweight to operate within the device’s memory constraints. Furthermore, the execution time confirms the feasibility of real-time queenless state detection in edge-based applications.

The rise of edge computing, characterized by swarms of edge devices, marks a significant shift in cloudedge computing landscapes, moving data processing closer to the source of data generation. However, this paradigm introduces complexities in orchestration, as traditional centralized methods become inadequate for effectively managing distributed, dynamic edge environments. In this article, we introduce FEARLESS, a distributed orchestration framework tailored for swarms of edge devices. FEARLESS employs a vertical federated reinforcement learning approach to efficiently orchestrate function invocation requests in serverless swarms. Experimental results demonstrate that FEARLESS significantly reduces the QoS violations of the scheduled tasks by up to 57%, compared to a centralized “least-CPU-utilization” and a “local-execution” approach, while it also achieves approximately up to 20% average total energy reduction.