Details about the clustering process

[sayakpaul]

Could you shed some light on the clustering process? In particular,

• Which latent represents did you use to train KMeans?
• If the encoder is being trained using AdvCL then which encoder was used to compute the latent representations that were used during KMeans training?
• What did you use to build the clustering index?