The algorithm of TBPFG is implemented in modules tbpf (numpy implementation) and tbpf_tf (tensorflow implementation up to pf degree 5).
Scripts in the root are for testing of algorithms performance.
- tbpf_numpy.py - Numpy implementation of TBPFG
- tbpf_tf_cpu.py - Tensorflow implementation of TBPFG, TF is forced to use CPU
- tbpf_tf_gpu.py - Tensorflow implementation of TBPFG, TF will use GPU if it is available
- pf_numpy.py - Simple recursive algorithm of PF generation
- pf_scikit.py - Algorithm from scikit-learn library sklearn.preprocessing.PolynomialFeatures
All testing scripts are CMD-line scripts that write the results into the console and saves results into CSV file. The naming convention for results file is res_{script name}_d{degree}_it{number of iterations}.csv.
Input parameters of testing scripts are:
- -h, --help show help message and exit
- -d DEGREE, --degree DEGREE Degree of polynomial features
- -i ITERATIONS, --iterations ITERATIONS Number of iterations over one number of inputs
- --start START Number of inputs start
- --stop STOP Number of inputs stop
- --step STEP Number of inputs step