Zero2Neuro Tests

The following tests exercise specific Zero2Neuro features, covering different data formats, data configurations, and training options.

  • Categorical Variables: Testing the translation of categorical variables into integer representations
    • Problem type: XOR
    • Data file type: csv and pickle
    • Case 1: Boolean strings (T/F) translated to 1 and 0
    • Case 2: Integer-valued categorical columns (31/32 translated to 1 and 0)
    • Cases 3 & 4: Pickle versions of Cases 1 and 2
  • Tabular Test: Testing different header row and data column configurations using the XOR problem
    • Problem type: XOR
    • Input data: binary feature vector
    • Model type: fully-connected (2 hidden layers: 20 and 10 units)
    • Data file type: csv
    • Multiple data configurations:
      • Header row is non-standard (explicitly specified)
      • Case 1: Define a range of columns to use
      • Case 2: Define a list of columns to use
  • Tabular XLSX Test: Testing different header row and data column configurations using xlsx format
    • Problem type: XOR
    • Input data: binary feature vector
    • Model type: fully-connected (2 hidden layers: 20 and 10 units)
    • Data file type: xlsx (multiple sheets)
    • Multiple data configurations:
      • Case 1: single data table (maps to training set)
      • Case 2: two sheets in the same file are data tables 0 and 1, which are mapped to training and validation data sets, respectively
  • Data Conversion Test: Each data table is converted from a numpy array to a tf-dataset
    • Problem type: XOR
    • Input data: binary feature vector
    • Model type: fully-connected (2 hidden layers: 20 and 10 units)
    • Data file type: xlsx
  • Tabular Weights Test: Testing per-sample weights in tabular data using the XOR problem
    • Problem type: XOR
    • Input data: binary feature vector
    • Model type: fully-connected (2 hidden layers: 20 and 10 units)
    • Data file type: csv
    • Demonstrates the effect of different sample weights on model training
    • Look for: trained model will perform properly for the first three examples, but the output for the last example will typically be incorrect
  • Pickle Test: Testing the pickle data format with a parity problem
    • Data file type: pickle
    • Data representation: numpy
    • Multiple data configurations across 6 variants
    • Fixed train/validation/testing data set splits

Table of contents


This site uses Just the Docs, a documentation theme for Jekyll.