Hello Team,
I am running code which uses memristor crossbar array my Network has 95 input 16 neuron 1 hidden layer and 6 output using HAR classifier,
- when i run Version 1.4 to check non-idealities and PPA and accuracy i have got 88%, when i used pure MLP for same network got 90.85% and when i run same using memristor crossbar array got 89%, which is okay, because version V1.4 dont have stuck at faults so i have thought to use V1.5 2DInferenceV1.5-dev.
- so when i used 2DInferenceV1.5-dev versionn start to run my code its having result of Ideal baseline : 84.11405295315683% so now iam not getting which one is correct this V1.4 or V1.5 results. so much confused.
- what might be the error is it because quantisation difference in the both version or something else???
if anybody know the solution or suggestion please provide help my work got stuck because of understanding of this.
Hello Team,
I am running code which uses memristor crossbar array my Network has 95 input 16 neuron 1 hidden layer and 6 output using HAR classifier,
if anybody know the solution or suggestion please provide help my work got stuck because of understanding of this.