konstmish

·há 5 anos·discuss

Hey, I’m Konstantin, the author of the paper. It’s a big surprise to see an active discussion here, it was very interesting for me to read the comments. I agree with the people saying that Newton with line search often works as it is, the point of my work is to obtain a theory that can be used later to obtain extensions to harder settings or modifications with more aggressive iterations. Only the time will tell if the method itself is useful in practice. I feel like most questions have been already answered here by the community, but I’m happy to answer any remaining questions.

I also presented this work online and the recording is available on youtube, I hope you’d enjoy it: https://youtu.be/0lXvUMfafBY