Holla Forums neural network (PNN)

Link: makeadverbsgreatagain.us/mlp/

You know what? We should out-do this son of a bitch.
Goal: To create a Neural Network that identifies friends and foes
Stage 1: Initial data collection
- Collect list of ~3K antifa Twitter accounts
- collect list of ~10K non-antifa Twitter accounts
Stage 2: Dump that into a antifa classification network
- Possibly use makeadverbsgreatagain.us/twitter-ml-2017-04-29.zip
- Feed in the data to a very powerful computer for training
- The code will scrape twitter accounts to see what they say
Stage 3: Finding of false positives and false negatives
- Gather a new set of antifa vs non-antifa accounts
- Input it into the machine to see what will happen
- Get the false positives/negatives to retrain the network
Stage 4: Antifa keyword sentiment analysis
- Using a sentiment classification network like the ones below
- github.com/danielegrattarola/twitter-sentiment-cnn
- github.com/xiaohan2012/twitter-sent-dnn
- See what keywords are positive and negative to them
Stage 5: Non-antifa group sentiment clustering
- Check the non-antifa list to see their sentiment to the keywords
- Use clustering algorithms to see which people have similar sentiments
- Interpret the clusters and add applicable labels to it
Stage 6: New group data classification and correction
- Manually check the result cluster for false positives/negatives
- Retrain the network for the clustering false positives/negatives
- Simply apply the result of 4~6, and apply the techniques in 1~3
- More indexes for discovering friend and foe

Article reference: medium.com/@conspirator0/identifying-political-bot-troll-social-media-activity-using-machine-learning-20dcd56e961a

...

Same thing applies, since one is the subset of the other

No actually it doesn't. Wasting time on antifa is just a D&C tactic you giant giga-kike. It would be the exact same thing if you had started this thread advocating to create one for the alt-kike too.

you start. don't bump this thread unless you have some code.

So you want to hunt the hardest thing while neglecting the basic ones?
First of all, antifa and alt-kike are easier to hunt, there is a bigger. We can't just hunt jews right on the spot, because the number of users, and by extension data is very sparse compared to the former two.
Second of all, jews are very sneaky, they would hide their identity as hard as they can, and keyword analysis would be impossible unless we can sieve out false-positives and false-negatives.
Either get a list of 3000 jewish twitter accounts (and that is a minimum), or we don't make this happen in the first place.

thats preposterous. do you really think its hard to find a few thousand jews on twitter?

One jew leads a thousand reds, so it is harder to do by the numbers.

Dude, start with the coding. These eDgY posts are seconds lost forever for the cause.

...

which side is telling the truth? or are you both equally retarded?

Shalom Herschel, posting for the cause early today i see.

Is this RIDICULOUSLY unreasonable? I'd appreciate any clues why that is so.

This sounds interesting but unfortunately it seems that it'd have little to no use and it'd cost quite a bit to train the network. On the other hand, i can see a lot of people using the "Alt-Right" detector. I think that training a network for more specific things that the people who are interested in this (and a portion of a larger group of normies) deems "wrong" would make it MUCH more interesting and worth something. Making it detect echoy people double standards on who is fit for nationalism (think US and Israel), being able to show the specific part of articles/posts that makes the network believe that is the case, would be great by itself i think.

I can code but i haven't bothered to learn ML yet so i'm clueless how to implement such a system but i could help with the front end for making the detector usable to the masses. I'm pretty sure that if someone or some few people here who knows how to lead a project took the initiative, a bunch of other people would be willing to help! And hopefully this wouldn't end up like most Holla Forums and Holla Forums projects.

An advice that i find very important for anyone writing any code related to this idea: Write it in an accessible language so the most people may help you. I'm pretty sure most people on Holla Forums and Holla Forums mostly knows imperative ceiling-high level languages so keep that in mind. Try not writing anything in C or in Haskell for example, unless you're certain that you can get it done by yourself or with your buddies.

HOLY SHIT OP THIS SOUNDS RAD
I'll make the logo!!!

Write your own code, faggot. It's not hard. You can even do it in FORTH.

Anyone worth their salt can program in basically any language once they know syntax. With source available, syntax is obvious.

I would reckon

For various reasons OP probably can't do it by himself and he has created this thread because of that.

It can also be done in Assembly.

We're on "Holla Forums", not a srs bsns mailing list where everybody has a degree in CS.

>I would reckon

Heh. It's like you want to have bugs in your codes.

And no, literally nobody in Holla Forums knows how to code. Most don't even know what forth is.

But FORTH is more fun.

This is why you don't belong here. You don't need a degree in CS to program. You don't need anything except a brain and the knowledge of the process of problem-solving. If you can think in logical steps, you can program.

FORTH doesn't have bugs, programmers bug FORTH.

I didn't realize it was you. Carry on being a rustcuck.

I'm not a Rustcuck. Can you read?
This is why Holla Forums posters should be banned

Shut the fuck up rustcuck