IamNN: Iterative and Adaptive Mobile Neural Network for Efficient Image Classification

TL;DR

This paper introduces IamNN, a ResNet-based neural network that employs parameter sharing and adaptive computation time to achieve efficient image classification with reduced size and input-dependent computation.

Contribution

It presents a novel ResNet-inspired architecture with parameter sharing and adaptive computation, enabling smaller models with input-adaptive complexity.

Findings

Network size is significantly reduced.

Computational cost adapts to input complexity.

Maintains high classification accuracy.

Abstract

Deep residual networks (ResNets) made a recent breakthrough in deep learning. The core idea of ResNets is to have shortcut connections between layers that allow the network to be much deeper while still being easy to optimize avoiding vanishing gradients. These shortcut connections have interesting side-effects that make ResNets behave differently from other typical network architectures. In this work we use these properties to design a network based on a ResNet but with parameter sharing and with adaptive computation time. The resulting network is much smaller than the original network and can adapt the computational cost to the complexity of the input image.