This video is really informative, includes definition, intuition, and experiments done by the creator himself, while still being short. Thank you for the good work!
Does the paper really state that you can prune without training to convergence? I thought the big deal was that they showed the existence of a smaller network that can be trained if we use the same initialization. I don't think they were able to find this network without training to convergence.
For the initial conditions that work, have anybody look at how much wiggle room you have. Is there an epsilon-neighborhood of the initial state you can safely start from, and how small is epsilon?
This video is really informative, includes definition, intuition, and experiments done by the creator himself, while still being short. Thank you for the good work!
This seems like a pretty influential concept.
Does the paper really state that you can prune without training to convergence?
I thought the big deal was that they showed the existence of a smaller network that can be trained if we use the same initialization. I don't think they were able to find this network without training to convergence.
Great job! Do you think we can make a lottery prediction by using the LTH method? Thanks.
Really nice figures!
For the initial conditions that work, have anybody look at how much wiggle room you have. Is there an epsilon-neighborhood of the initial state you can safely start from, and how small is epsilon?
Lol, its completely wrong what you say. Go study first before you make a video mate
can you please tell me where is wrong, cos i m interested in this concept. thx