SCNN: Swarm Characteristic Neural Network

TL;DR

This paper introduces SCNN, a neural network inspired by swarm behavior, which achieves better efficiency and effectiveness with fewer parameters and less data, addressing the need for models suitable for complex problems.

Contribution

The paper presents a novel swarm-inspired neural network architecture, demonstrating improved performance and data efficiency over traditional models.

Findings

Fewer parameters can outperform larger models.

SCNN requires less data for training.

SCNN is more efficient and effective for complex tasks.

Abstract

Deep learning is a powerful approach with good performance on many different tasks. However, these models often require massive computational resources. It is a worrying trend that we increasingly need models that work well on more complex problems. In this paper, we propose and verify the effectiveness and efficiency of SCNN, an innovative neural network inspired by the swarm concept. In addition to introducing the relevant theories, our detailed experiments suggest that fewer parameters may perform better than models with more parameters. Besides, our experiments show that SCNN needs less data than traditional models. That could be an essential hint for problems where there is not much data.