Fork me on GitHub

@danieltanfh95 I am not sure that I understand the question. Neanderthal is a matrix/linear algebra library. Transfer learning is a concept from machine learning. Do you have a hello world style simple example that illustrates what exactly you are trying to do?

Daniel Tan10:09:00

My bad, I didn't know the difference, I was looking for a high level library to do transfer learning on image models like resnet in clojure


First you'd have to create a resnet in Clojure (which should be supported by deep diamond, but I haven't tried to do that yet). Then to do transfer learning you "simply" ("just") train it on one domain, and then additionally train it on the other domain (the transfer part). Or maybe you're referring to an already trained resnet that you'd download from the Internet, probably developed in Pyhon, that you'd like to be able to use in Clojure. That would almost certainly require some technical work, but (in theory) it should be supported. You'd need to construct an equivalent network in Clojure, and then copy the parameters from the source that you have. This almost certainly requires that you compare the format of your source and the actual bytes that the network you've constructed in Deep Diamond expects...

👀 1
Daniel Tan14:09:45

any tutorial i can look at to begin :thinking_face:


Regarding DD, the DLFP book. Regarding transfer learning, depends on what exactly you want to do.

Daniel Tan03:09:23

hoping to have some tutorial on just doing transfer learning on something like resnet