Details
Presenter(s)
![Rituparna Choudhury Headshot](https://confcats-catavault.s3.amazonaws.com/CATAVault/ieeecass/master/files/styles/cc_user_photo/s3/user-pictures/20731_1.jpg?h=f35c3bdd&itok=cp6UhLuQ)
Display Name
Rituparna Choudhury
- Affiliation
-
AffiliationIndian Institute of Technology Guwahati
- Country
-
CountryIndia
Abstract
Decision Tree (DT) algorithms perform classification of data according to decision criteria obtained during training which results in high computational complexity and latency. In this paper, a Hybrid DT (HDT) is proposed which reduces training complexity and achieves 8x speed-up as compared to conventional DT. The critical path of proposed HDT hardware enables the FPGA to operate at a maximum frequency of 125 MHz. Simulation results show that the proposed hardware achieves $1000times$ speed-up as compared to software-based realisation and $60times$ speed-up as compared to existing FPGA training accelerator.