Re:Hashing for large-scale similarity
by mesuvash on Sunday February 17, 2019 @04:34PM
(#19187071)
Thanks for the pointers.
From my personal experience, Auto-encoders are amazing for dense input (images, audio etc), more specifically, when the input feature space is not large. However, in many real-world problems such as recommendation, ranking etc. the feature space is generally very sparse for eg clicks, purchase of items (say 100M items). In such cases, scaling can be challenging with neural models esp Autoencoder.
From my personal experience, Auto-encoders are amazing for dense input (images, audio etc), more specifically, when the input feature space is not large. However, in many real-world problems such as recommendation, ranking etc. the feature space is generally very sparse for eg clicks, purchase of items (say 100M items). In such cases, scaling can be challenging with neural models esp Autoencoder.