Interpretability of Neural Networks

Feature Visualization

Process of generating synthetic inputs that maximize the activation of specific neurons or entire layers, allowing visual understanding of what each neuron has learned to detect.

← Wstecz