I want to know what exactly the EON tuner does. Let’s say I trained a MobileNetV2 model by transfer learning for my image classification task, then I go to the EON Tuner and click on “Start EON Tuner”, does it remove the unnecessary neurons from the trained model (aka Model Pruning) or does it do something else to optimize the model? Or does it find a completely different model to train the data from scratch (no transfer learning).
The EON Tuner tries many combinations of DSP (feature extraction) blocks and learning blocks (ML models) to help you find an optimum combination to meet your accuracy, RAM, ROM, and timing requirements. It does not do any model pruning.
It’s one implementation of something known as “AutoML.” The idea is to help you choose feature extraction, model architecture, and hyperparameters when designing an ML project. You can read more about the EON Tuner here.
Hope that helps!