Why Did You Order That CBC?

Grad-CAM: Why did you say that?

We propose a technique for making Convolutional Neural Network (CNN)-based models more transparent by visualizing input regions that are ‘important’ for predictions – producing visual explanations. Our approach, called Gradient-weighted Class Activation Mapping (Grad-CAM), uses class-specific gradient information to localize important regions. These localizations are combined with existing pixe...

Where did you get that hat?

C Dm Where did you get that hat? Where did you get that tile? Isn't it a nobby one, and just the proper style? C I should like to have one just the same as that!" C Where'er I go, they s C 1. Now how I came to get this hat, 'tis very strange and funny Grandfather died and left to me his property and money F C F And when the will it was read out, they told me straight and flat C If I would have ...

"And You Did That Why?" - Using an Abstraction Hierarchy to Design Interaction with Autonomous Spacecraft

We are investigating the design of user interaction with autonomous spacecraft in particular, with spacecraft where contact is made only intermittently. Our initial efforts focus on designs supporting existing spacecraft capabilities, where autonomy is reflexive and based on complicated rules. We report here an initial design concept for human interaction based on an abstraction hierarchy, prod...

Grad-CAM: Why did you say that? Visual Explanations from Deep Networks via Gradient-based Localization

We propose a technique for producing ‘visual explanations’ for decisions from a large class of Convolutional Neural Network (CNN)-based models, making them more transparent. Our approach – Gradient-weighted Class Activation Mapping (Grad-CAM), uses the gradients of any target concept (say logits for ‘dog’ or even a caption), flowing into the final convolutional layer to produce a coarse localiz...

Why did you choose that person over me? Ingroup rejection and attributions to discrimination