Model Compression Techniques

Techniques to reduce the size of Artificial Neural Networks (ANN) models to make them faster, less resource intensive, and therefore better suite them for resource constraints of small devices.

Note

Model Compression Techniques are an important part of Embedded Artificial Intelligence (EAI).


Advantages:

  • Greatly reduced model size.
  • Faster prediction and inference.
  • Reduced the resources required to train or run models.

Disadvantages:

  • Added complexity in training and deploying the model.
  • Decreased accuracy, even if it's often small.
  • Computation resources needed to perform this task.

Techniques: