NormGrad: Finding the Pixels that Matter for Training

10/19/2019
by   Sylvestre-Alvise Rebuffi, et al.
0

The different families of saliency methods, either based on contrastive signals, closed-form formulas mixing gradients with activations or on perturbation masks, all focus on which parts of an image are responsible for the model's inference. In this paper, we are rather interested by the locations of an image that contribute to the model's training. First, we propose a principled attribution method that we extract from the summation formula used to compute the gradient of the weights for a 1x1 convolutional layer. The resulting formula is fast to compute and can used throughout the network, allowing us to efficiently produce fined-grained importance maps. We will show how to extend it in order to compute saliency maps at any targeted point within the network. Secondly, to make the attribution really specific to the training of the model, we introduce a meta-learning approach for saliency methods by considering an inner optimisation step within the loss. This way, we do not aim at identifying the parts of an image that contribute to the model's output but rather the locations that are responsible for the good training of the model on this image. Conversely, we also show that a similar meta-learning approach can be used to extract the adversarial locations which can lead to the degradation of the model.

READ FULL TEXT
research
04/06/2020

There and Back Again: Revisiting Backpropagation Saliency Methods

Saliency methods seek to explain the predictions of a model by producing...
research
06/06/2019

Segment Integrated Gradients: Better attributions through regions

Saliency methods can aid understanding of deep neural networks. Recent y...
research
11/21/2020

Backdoor Attacks on the DNN Interpretation System

Interpretability is crucial to understand the inner workings of deep neu...
research
02/03/2020

Robust saliency maps with decoy-enhanced saliency score

Saliency methods help to make deep neural network predictions more inter...
research
10/18/2019

Understanding Deep Networks via Extremal Perturbations and Smooth Masks

The problem of attribution is concerned with identifying the parts of an...
research
10/01/2019

Underwhelming Generalization Improvements From Controlling Feature Attribution

Overfitting is a common issue in machine learning, which can arise when ...

Please sign up or login with your details

Forgot password? Click here to reset