Tropical Pruning Strategies for ANN optimization in IoT

This study focuses on optimizing Multi-Layer Perceptron (MLP) neural networks for deployment in resource-constrained edge computing environments like smartphones and Internet of Things (IoT) devices. The proposed method involves using auxiliary dense Morphological Neural Networks (MNN) to generate pruning masks for each hidden layer in the MLP. These MNN have shown high efficiency in the pruning process, leading to a significant reduction in the total number of parameters. The effectiveness of our proposal is explained and validated using the Fashion-MNIST dataset. Furthermore, the performance of the optimized MLP is evaluated on an IoT platform, and compared with other state-of-the-art pruning techniques. The experimental results show the effectiveness of the proposed technique in reducing model complexity, improving power efficiency, and decreasing inference response time, while maintaining minimal impact on overall MLP accuracy. These findings have important implications for the development of efficient and scalable deep learning models tailored for edge computing applications.