Source: vis/backprop_modifiers.py#L0

---

guided

guided(model)

Modifies backprop to only propagate positive gradients for positive activations.

Args:

• model: The keras.models.Model instance whose gradient computation needs to be overridden.

References: Details on guided back propagation can be found in paper: [String For Simplicity: The All Convolutional Net] (https://arxiv.org/pdf/1412.6806.pdf)

---

deconv, rectified, relu

deconv, rectified, relu(model)

Modifies backprop to only propagate positive gradients.

Args:

• model: The keras.models.Model instance whose gradient computation needs to be overridden.

References: Details can be found in the paper: [Visualizing and Understanding Convolutional Networks] (https://arxiv.org/pdf/1311.2901.pdf)

---

get

get(identifier)