151–160 of 225 posts
Adam Neuman got away with billions with WeWork bankrupt Apple and Google help China's surveillance of dissents Google Chrome monitors straight viewers Solar panels can not slow climate change, only nuclear power can SpaceX to to build new Silicon Valley on Mars
It's basically a buzzword detector. "this is just a tool for detecting buzzwords" => Bad: 0.9991 - Good: 0.0011 "this is merely a device for detecting artificially sophisticated words" => Bad: 0.0019 - Good: 0.9980
That is how it works here, you collect upvotes if you use fancy words. That's why everybody uses the word "orthogonal" here all the time. Have you ever seen that word anywhere else?
Unfortunately far too often both in tech and in CS academia (and I'm talking about informal conversations ie. this is an 'orthogonal idea')
Congratulations on getting your hands dirty and doing everything yourself like computing gradients manually, badly shuffling (non Fisher-Yates), badly js transpose (double swapping), it is a great way to learn.
Congratulations on completing a full pipeline, that's the hard part then it just swapping pieces for better pieces.
I advise non-technical readers not to attach much value to the results of this neural network as it is probably inferior to the even simpler naive Bayes.
The model of the neural network is simplistic :
Concat(Word Vectors)-Dense(120,act=sigmoid)-Dense(60,act=sigmoid)-Dense(2,act=sigmoid)
The Concat operation mean it is especially sensible to dropping or adding a word as it will offset the remaining words and give a totally different vector.
Using word vectors mean it doesn't forgive any spelling mistake as a spelling mistake will usually correspond to vector.
Using a feed forward neural network means formulaic titles with a single word substitution from a good positive example from the training set will often work.
It is trained by gradient descent using a squared error loss, on ~1000 examples one example at a time without cross-validation using a custom written neural network library. (Almost all these bad choices can be solved by using a framework).
It seems to have successfully over-fit as it return Good ~1.0 for positive examples from the training set.
A few 1.0s: Adam Neuman got away with billions with WeWork bankrupt Apple and Google help China's surveillance of dissents Google Chrome monitors straight viewers Solar panels can not slow climate change, only nuclear power can SpaceX to to build new Silicon Valley on Mars
My try: "deep neural network ate my tuna sandich" Bad: 0.0000 - Good: 1.0000
Perfect score!!!!
Bad: 0.0012 Good: 0.9986
"someone ate my sandwich" Bad: 0.9999 - Good: 0.0001
Hacker news can't spell, or maybe associates sandwhich and someone eating it to some repository.
"I plan to rewrite Linux in Rust - Linus Torvalds" Bad: 0.9988 - Good: 0.0013 Pretty sure HN would break if this actually happened.
"I plan to rewrite Linux in Go - Linus Torvalds" Bad: 0.9999 - Good: 0.0001 Rust is 13x better over Go in this benchmark.
Bad: 0.1897 - Good: 0.7988
Looks like we have a winner.