Presentation

· Contributors · Organizations · Search Program · My Schedule · Happening Now · Maps

This content is available for: Workshop Reg Pass. Upgrade Registration

Fast 2D Bicephalous Convolutional Autoencoder for Compressing 3D Time Projection Chamber Data

SessionThe 9th International Workshop on Data Analysis and Reduction for Big Scientific Data (DRBSD-9)

DescriptionHigh-energy large-scale particle colliders produce data at high speed in the order of 1 terabytes per second in nuclear physics and 1 petabytes per second in high-energy physics. Time projection chamber tracking detector data are usually very sparse, which presents a challenge to conventional learning-free lossy compression algorithms such as SZ, ZFP, and MGARD. The 3D convolutional neural network (CNN)-based approach named Bicephalous Convolutional Autoencoder (BCAE) outperforms traditional methods both in compression rate and in reconstruction accuracy. BCAE can also utilize the computation power of graphical processing units. Here, we introduce an improved 3D CNN that achieves X% better compression ratio and Y% better reconstruction accuracy measured in mean absolute error comparing to BCAE. We also introduce a novel 2D CNN variant by treating the radial direction as the channel dimension, resulting a 3x in compression throughput without losing too much in reconstruction accuracy.

Author/Presenters

Yi Huang

Brookhaven National Laboratory

Yihui Ren

Brookhaven National Laboratory