Permalink
Commits on Mar 5, 2013
  1. Run all the way through by passing in a file and have it be processed…

    … against the existing model.
    
    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Mar 5, 2013
Commits on Mar 1, 2013
  1. non working version of the perceptron...adding in api to check indivi…

    …dual emails outside of the train and test sets.
    
    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Mar 1, 2013
Commits on Feb 28, 2013
  1. Working, weight adjusting version of the perceptron in Go. The code i…

    …s still shit...needs a big rewrite.
    
    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Feb 28, 2013
  2. Initial Go rewrite. Note - weights are not updating properly.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Feb 28, 2013
  3. Making way for Go version.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Feb 28, 2013
Commits on Feb 10, 2013
  1. Added some printout to capture some stats from testing.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Feb 10, 2013
Commits on Feb 8, 2013
  1. Keep iterating on an already defined model.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Feb 8, 2013
Commits on Feb 7, 2013
  1. Updated perceptron to operate on vectors and do no input munging.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Feb 7, 2013
  2. Preproc all data and split into train and test.

    We don't want the perceptron setting up vectors of data and matching
    inputs to the larger set of inputs during its runtime. That should all
    be at a preprocessing stage so that the perceptron can just run.
    
    This changeset includes the above changes. Also, it will take 10% of the
    limit value and apply that to the test set. This split is so that you
    aren't testing with trained data...you wont know if you're NN is right
    if you don't.
    
    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Feb 7, 2013
Commits on Feb 1, 2013
  1. Perceptron code with slight modifications to preprocessor.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Feb 1, 2013
Commits on Jan 31, 2013
  1. Updates to add offset and limit for loading data into 'data sets'...m…

    …akes it easier to split training and test data out.
    
    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Jan 31, 2013
Commits on Jan 28, 2013
  1. Pickle word counts per doc rather than the text of the document. The …

    …text is irrelevant whereas we need to pass in the real data to the inputs of the neural net.
    
    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Jan 28, 2013
Commits on Jan 27, 2013
  1. Preparation for the perceptron.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Jan 27, 2013
Commits on Dec 11, 2011
  1. Split out to sep file.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Dec 11, 2011
  2. Remove .o files and change function name.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Dec 11, 2011
  3. Super basic. Find last char idx. K&R.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Dec 11, 2011
Commits on Dec 5, 2011
  1. Some erlangy stuff from ErlangDC.

    Signed-off-by: Nick Campbell <nick@perfectsensedigital.com>
    committed Dec 5, 2011
Commits on Oct 29, 2010
  1. Learning about genetic algorithms. This tutorial is from http://www.a…

    …i-junkie.com/ga/intro/gat3.html
    
    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Oct 29, 2010
Commits on Aug 22, 2010
  1. Learning a bit about canvas.

    Signed-off-by: Nick Campbell <nicholas.j.campbell@gmail.com>
    committed Aug 22, 2010
Commits on Jul 17, 2009
Commits on Jul 16, 2009
  1. Quicksorts.

    committed Jul 16, 2009
  2. Wrong spot.

    committed Jul 16, 2009
  3. Hello World

    committed Jul 16, 2009
  4. Initial commit.

    committed Jul 16, 2009