network-shrinking

Model Rubik's Cube: Twisting Resolution, Depth and Width for TinyNets

Overview of TinyNet TinyNet is a technique for downsizing neural architectures through a series of smaller models derived from EfficientNet-B0 with the FLOPs constraint. This method explores the twisting rules for obtaining deep neural networks with minimum model sizes and computational costs while maintaining high efficiency and excellent performance. EfficientNets EfficientNets is a series of techniques designed for obtaining excellent deep neural architectures. The giant formula for enlar

NetAdapt

NetAdapt is an algorithm designed to adapt a pretrained network to a mobile platform with limited resources. It takes into account direct metrics such as latency and energy consumption to optimize the adaptation process. The algorithm is based on empirical measurements, which means it can be applied to any platform, regardless of the underlying implementation. The Problem with Existing Algorithms Many existing algorithms for simplifying networks focus on indirect metrics like the number of MA