mnistlr2.ssc

class xopts extends Learner.Options with SFilesDS.Opts with GLM.Opts with ADAGrad.Opts;
val mnopts = new xopts

val mdir = "/data/MNIST8M/parts/"
mnopts.fnames = List(FilesDS.simpleEnum(mdir+"cats%02d.smat.lz4", 1, 0),  // File name templates, %02d is replaced by a number
                     FilesDS.simpleEnum(mdir+"part%02d.smat.lz4", 1, 0));
mnopts.nstart = 0;                 // Starting file number
mnopts.nend = 70;                  // Ending file number
mnopts.order = 0;                  // sample order, 0=linear, 1=random
mnopts.lookahead = 2;              // number of prefetch threads
mnopts.featType = 1;               // feature type, 0=binary, 1=linear
mnopts.fcounts = icol(10,784);     // how many rows to pull from each input matrix 
mnopts.eltsPerSample = 300         // how many rows to allocate (non-zeros per sample)


mnopts.batchSize = 1000
mnopts.npasses = 2
mnopts.lrate = 0.001

mnopts.links = 2*iones(10,1);
mnopts.targets = mkdiag(ones(10,1)) \ zeros(10, 784);
mnopts.rmask = zeros(1,10) \ ones(1, 784);

val ds = {
  implicit val ec = threadPool(4);   // make sure there are enough threads (more than the lookahead count)
  new SFilesDS(mnopts);              // the datasource
}

val nn = new Learner(                // make a learner instance
    ds,                              // datasource
    new GLM(mnopts),                 // the model (a GLM model)
    null,                            // list of mixins or regularizers
    new ADAGrad(mnopts),             // the optimization class to use
    mnopts)                          // pass the options to the learner as well

nn.train

mnopts.nstart=71
mnopts.nend=80
mnopts.npasses=1

nn.repredict