Man, took me awhile to get through it. The ideas are really interesting i just think its too much of a slog. You have (understandably) made new words for various concepts (like inout, triggers, etc.). This makes it hard for the reader.
Particularly, this makes the intro really hard. It kind of reads like gobbiligoop. That is until you read the entire thing, then you look back and it all makes sense. If you want more people to read it (which you should, once again, really interesting perspective to interpretability) I would rewrite the intro, and perhaps the abstract to account for this.
Man, took me awhile to get through it. The ideas are really interesting i just think its too much of a slog. You have (understandably) made new words for various concepts (like inout, triggers, etc.). This makes it hard for the reader.
Particularly, this makes the intro really hard. It kind of reads like gobbiligoop. That is until you read the entire thing, then you look back and it all makes sense. If you want more people to read it (which you should, once again, really interesting perspective to interpretability) I would rewrite the intro, and perhaps the abstract to account for this.
· Reply
Make
jake-denning
your Representive in the
What’s in the box?! – Towards interpretability by distinguishing niches of value within neural networks.
topic?
Share
Moderate