AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Manifold Hypothesis
I was wondering to what extent, if any, any, of this invalidates the so called manifold hypothesis. I think when most people speak ofintere and deep learning, their intuition is something along the lines of, wellh the model learns some, you know, some manifold in the latent space. And interpolation means i'm just kind of like traversing a geodesic on that manifold. When you visualize the results on an autoin code or something, yout you can see this kind of continuous geometric morphin,. one image into another image.