Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Josh Gruenstein
    @joshuagruenstein
    and is everything scaled between 1 and 0
    Henry Wildermuth
    @FlyingGraysons
    Dude look over the code
    Josh Gruenstein
    @joshuagruenstein
    std::vector<double> digits(uchar j) {
        std::vector<double> v;
        for (size_t i = 0; i < 10; i++) {
            if (j == i) {
                v.push_back(1);
            } else {
                v.push_back(0);
            }
        }
        return v;
    }
    Henry Wildermuth
    @FlyingGraysons
    I'll put up my testing code as well
    Josh Gruenstein
    @joshuagruenstein
    you're comparing a uchar with a size_t
    wtf is going on there
    there's zero way that could work
    Henry Wildermuth
    @FlyingGraysons
    ill make the other one a uchar too
    it does work i tested it
    Josh Gruenstein
    @joshuagruenstein
    wut how
    look at the ascii table
    wait
    what's being passed in
    is it an actual ascii character
    Henry Wildermuth
    @FlyingGraysons
    its using uchar for number not character
    Josh Gruenstein
    @joshuagruenstein
    henry wtf are you doing
    why
    Henry Wildermuth
    @FlyingGraysons
    Its not me its the mnist function i copied
    Josh Gruenstein
    @joshuagruenstein
    lolk
    ok
    so here's an idea
    try commenting out the backprop.train line
    and see if you get the same output
    Henry Wildermuth
    @FlyingGraysons
    Same output from what
    Josh Gruenstein
    @joshuagruenstein
    better yet
    print the weights before and after backprop.train
    Henry Wildermuth
    @FlyingGraysons
    ok
    doing it with 1000 samples
    Josh Gruenstein
    @joshuagruenstein
    you mean training?
    Henry Wildermuth
    @FlyingGraysons
    ya
    Josh Gruenstein
    @joshuagruenstein
    lol whatever floats your boat
    Henry Wildermuth
    @FlyingGraysons
    why not
    Ok literally theres more output than can fit on my screen josh
    Josh Gruenstein
    @joshuagruenstein
    lol yeah no shit you have a shittone of neurons
    Henry Wildermuth
    @FlyingGraysons
    *tonne
    Josh Gruenstein
    @joshuagruenstein
    but are the weights the same
    Henry Wildermuth
    @FlyingGraysons
    You should try messing round with it on your computer
    Its still training btw
    Josh Gruenstein
    @joshuagruenstein
    hmmm
    kinda doing other stuff
    having this convo in the background
    Henry Wildermuth
    @FlyingGraysons
    :(
    They are different
    @joshuagruenstein
    Josh Gruenstein
    @joshuagruenstein
    goddamn
    Henry Wildermuth
    @FlyingGraysons
    But it seems like the tests have almost the exact same results every time
    Josh Gruenstein
    @joshuagruenstein
    huh
    Henry Wildermuth
    @FlyingGraysons
    Just checked
    and it made the exact same weights as last time
    Josh Gruenstein
    @joshuagruenstein
    well doesn't that make sense