At the first step of the project a non-functional demo could be written in the documentation, based on the paper, it could be something like the demo below:
# 1. Read the input file and sepparate input from output columns
df = pandas.read_csv('inputDataFile.csv')
all_inputs = df.iloc[:, 1:]
outputs = df.iloc[:, 0]
# 2. Once the data has been managed, then create a layer style architecture pipeline
pipeline = niaaml.pipeline(
data_feed = {
inputs: all_inputs
outputs: outputs
},
feature_selection = {
algorithms: [niaaml.optim.algorithms.de(parameters) ,niaaml.optim.algorithms.pso(parameters)],
transforms: [niaaml.transforms.rescaling]
},
classifier_selection = {
algorithms: [
niaaml.classifiers.mlp(
depths=range(1,10),
training_algorithms=[niaaml.optim.algorithms.de(parameters)]
),
niaaml.classifiers.adaboost()]
}
)
# 3. Run the pipeline and get the resulting model, with the possibility
# of using GPU processing capabilities
devices = niaaml.get_compatible_devices()
optimization = pipeline.optimize(device=devices[0])
# 4. After that, obtain the corresponding model summary
print(optimization.best_pipeline.summary())
At the first step of the project a non-functional demo could be written in the documentation, based on the paper, it could be something like the demo below: