search

RyanMan1 / PYNQ-SVM-OpenHW-2020

A repository of my Xilinx Open Hardware 2020 submission including a demo of support vector machines on PYNQ, C++ source code and projects for HLS and tcl scripts to re-generate Vivado IPI overlay block designs.
16 stars 8 forks source link

Classification with an SVM model trained using SVM_light #2

Open Afef00 opened 2 years ago

Afef00 commented 2 years ago

Hello, I wanted to use this for a multiclass SVM (on MNIST dataset) on ZCU104 board . I trained the SVM model using SVM_light where I get a model file and a predictions file. The model file contains the information regarding the training dataset like the size of the data used for training. It also contains information like the number of Support Vectors and their values, number of features, the bias value and the type of kernel used for training. The predictions file contains all the values before applying the sign function of the decision function. These values are used for assigning the class which the test instance belongs to. Is it possible to use your work using the given output files from svm_light? Thanks!

RyanMan1 commented 2 years ago

Hi there,

I would recommend having a look at the demo notebook and seeing if you can format your datasets into this format. Then the bitstream can be programmed and the SVM result calculated. Please let me know if you run into any issues.

Cheers! Ryan

Afef00 commented 2 years ago

Ok I will try that. Thank you for the help! Best regards