HammingNN™ Classifier

Welcome to the HammingNN neural network based nearest neighbor classifier. Written in python, this version of the classifier can be run from your web browser. Please make sure that your browser is set to accept cookies.
You can upload the file(s) you wish to process from your computer to the classifier. Training, test, or classify files should be in Orange format, either the older or the newer tab-delimited style (see Orange file details) while production files should be in python pickle format as generated by HammingNN (using the "Generate production file" button below).
nn.py version: v1_13 hamming.py version: v7_0

Local files to upload:Files on the server:
Training file:
Test file:
Production file:
Classify file:
Verify the training file (-v)
Display explanation of Orange file formats (-o)
Use weighted overall sums for testing, rather than fired counts (-s)
Use one bit per class for continuous attributes (default is one bit per slice) (-x)
Use one bit per class for discrete attributes (default is one bit per unique value) (-z)
Treat continuous attributes as discrete attributes (-d)
Use graded bits for continuous attributes (-g)
Use weighting by prevalence (-w)
Print out additional statistics (-y)

Number of slices for continuous attributes (-n)
Upper limit for finding optimum number of slices (-u)
Depth value (-j)
Variable margin (-k)
Number of attributes to use in classification (-i)
Number of cross-validation folds (default is leave-one-out) (-f)
Number of trials, using random selection of cases for folds (-r)

Use this list of attributes to process (-a)
Use this list of slice values (one value for each attribute) (-b)