English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Pruning convolutional filters has demonstrated its effectiveness incompressing ConvNets. Prior art in filter pruning requires users to specify atarget model complexity (e.g., model size or FLOP count) for the resultingarchitecture. However, determining a target model complexity can be difficultfor optimizing various embodied AI applications such as autonomous robots,drones, and user-facing applications. First, both the accuracy and the speed ofConvNets can affect the performance of the application. Second, the performanceof the application can be hard to assess without evaluating ConvNets duringinference. As a consequence, finding a sweet-spot between the accuracy andspeed via filter pruning, which needs to be done in a trial-and-error fashion,can be time-consuming. This work takes a first step toward making this processmore efficient by altering the goal of model compression to producing a set ofConvNets with various accuracy and latency trade-offs instead of producing oneConvNet targeting some pre-defined latency constraint. To this end, we proposeto learn a global ranking of the filters across different layers of theConvNet, which is used to obtain a set of ConvNet architectures that havedifferent accuracy/latency trade-offs by pruning the bottom-ranked filters. Ourproposed algorithm, LeGR, is shown to be 2x to 3x faster than prior work whilehaving comparable or better performance when targeting seven pruned ResNet-56with different accuracy/FLOPs profiles on the CIFAR-100 dataset. Additionally,we have evaluated LeGR on ImageNet and Bird-200 with ResNet-50 and MobileNetV2to demonstrate its effectiveness. Code available athttps://github.com/cmu-enyac/LeGR.

Towards Efficient Model Compression via Learned Global Ranking