Hi,
I am working on FFN network training for a couple of weeks. However, the inference result is still not ideal. Large neurons are detected properly while many tiny neurons are left blank. I am wondering are there any method to refill these small blank areas?
Would appreciate any suggestions and insights.