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

To obtain lower inference latency and less memory footprint of deep neuralnetworks, model quantization has been widely employed in deep model deployment,by converting the floating points to low-precision integers. However, previousmethods (such as quantization aware training and post training quantization)require original data for the fine-tuning or calibration of quantized model,which makes them inapplicable to the cases that original data are not accesseddue to privacy or security. This gives birth to the data-free quantizationmethod with synthetic data generation. While current data-free quantizationmethods still suffer from severe performance degradation when quantizing amodel into lower bit, caused by the low inter-class separability of semanticfeatures. To this end, we propose a new and effective data-free quantizationmethod termed ClusterQ, which utilizes the feature distribution alignment forsynthetic data generation. To obtain high inter-class separability of semanticfeatures, we cluster and align the feature distribution statistics to imitatethe distribution of real data, so that the performance degradation isalleviated. Moreover, we incorporate the diversity enhancement to solveclass-wise mode collapse. We also employ the exponential moving average toupdate the centroid of each cluster for further feature distributionimprovement. Extensive experiments based on different deep models (e.g.,ResNet-18 and MobileNet-V2) over the ImageNet dataset demonstrate that ourproposed ClusterQ model obtains state-of-the-art performance.

Towards Feature Distribution Alignment and Diversity Enhancement for Data-Free Quantization