Skip to content

Supporting Vector Machine Classsfications Using High-Level Synthesis

Notifications You must be signed in to change notification settings

jianyicheng/svm_hls_on_zynq

Repository files navigation

svm_hls_on_zynq

Advanced Digital System Design Coursework Acceleration of Supporting Vector Machine (SVM) Classifcations on Zynq7000 SoC

  1. Single core system (1 ARM core + 1 IP core), total execution of 5 ms for 2000 test data sets. IP core latency = 90.
  2. Dual core system (2 ARM core + 2 IP core), total execution of 2.5 ms for 2000 test data sets. IP core latency = 146.
  3. Cascade IP core, maximum latency 121, minimum latency 62. Average latency of 72 for 2000 test data sets.
  4. Matlab SVM coefficient generation code.

Future improvements:

  1. DMA & FIFO implementations to maximize transmission data rate.
  2. NEON instruction to carry out SIMD processing on ARM cores.
  3. Cascade precision analysis (details see report).

About

Supporting Vector Machine Classsfications Using High-Level Synthesis

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages