-
post--activation-atlas
Using feature inversion to visualize millions of activations from an image classification network, we create an explorable activation atlas of features the network has learned which can reveal how the network typically represents some concepts.
-
SurveyJS
Open-Source JSON Form Builder to Create Dynamic Forms Right in Your App. With SurveyJS form UI libraries, you can build and style forms in a fully-integrated drag & drop form builder, render them in your JS app, and store form submission data in any backend, inc. PHP, ASP.NET Core, and Node.js.
i encourage you to try and think about it more as "features are a kind of representation" than "a representation is a collection of features". when I hear the word "features", that suggests to me that each component of the vector is a feature. that's often not how it will work in deep learning. instead what you get is a set of coordinates, a position in a kind of "semantic space" where we expect two sets of coordinates to be near each other if the things they represent are similar.