What is inference cost optimization in production Keras systems?

Inference cost optimization reduces compute, memory, and latency per prediction.

Methods include quantization, pruning, knowledge distillation, and batch inference. These trade accuracy for efficiency in production systems.