NSS’2023
Talk, NSS'23, Canterbury-UK
This week the International Conference on Network and System Security (NSS’23) happens and we have our work on distributed deep learning in the program. It is a joint work of me, Xiangyu Su and Keisuke Tanaka. Su is the main author and has introduced the main idea.
Su noticed that it is possible to harvest of a computational power from, say, of a proof of work protocol, in order to train models for machine learning. In order to do that we had to introduce new notions like “distributed proof-of-deep-learning”.
Basically, instead of “correct hash value” as in proof-of-work, the system trains models up until a certain suitable accuracy level in the learning curve. The paper is already available here and here. Please take a look.
