Thanks so much for publishing this as a Collab notebook! So, so cool! One question... Why are you explicitly naming certain attribute dimensions (e.g. black hair), and how did you know them a priori? Was this something you determined through inspection of previous trained models?
-
-
-
For the CelebA experiments, we use (a subset of) the attribute labels in the original dataset. However, you could also use the decisions of a separately trained classifier. That's similar to what we do with the music dataset, where we use non-differentiable reward rules.
-
Huh... Is CelebA a labeled/Supervised dataset, or are the attribtes latent variables that were discovered by the VAE? (I apologize that I'm not familiar with the details of the data set) I.e., was "black hair" something that was ever explicitly provided to the ML system?
কথা-বার্তা শেষ
নতুন কথা-বার্তা -
লোড হতে বেশ কিছুক্ষণ সময় নিচ্ছে।
টুইটার তার ক্ষমতার বাইরে চলে গেছে বা কোনো সাময়িক সমস্যার সম্মুখীন হয়েছে আবার চেষ্টা করুন বা আরও তথ্যের জন্য টুইটারের স্থিতি দেখুন।