7.2. Definitions¶

Note

The nextnanoevo Python package is under development. Its release is planned for 2024.

The optimization procedure is defined by vector of input variables, metric to be optimized (also called fitness for some algorithm), and function that converts input variabels to metric.

To be able to define these parameters of optimization procedure in nextnanoevo, 2 key instances should be created: NextnanoIO and NextnanoMetricExtractor.

nextnanoevo.IO¶

IO stands for nextnano input-output. This class defines:

input_file_path: input file to run

input_variable_names: variables to variate in the input file

target_output: output datafiles that will be used

To create IO instance, use the following syntax

from nextnanoevo.IO import IO

input_file_path = r'\path\to\input_file.in'

variables = ['variable_name1', 'variable_name2']
# example: variables = ['WellWidth', 'BarrierWidth']

output_files = [('relative', 'path', 'datafile1'),
                ('relative', 'path', 'datafile2')]

# the path of output files is relative to output directory
# example: output_files = [('Strain', 'strain_simulation.dat'),
#                          ('bias_00000', 'Quantum', 'energy_spectrum_quantum_region_Gamma_00000.dat')]

nn_io = IO(input_file_path, variables, output_files)

Metric¶

This class defines:

input_length: number of datafiles that will be processed (should match with NextnanoIO)

output_length: dimensionality of the metric vector

extraction_function: how to extract the final metric vector from the datafiles (the function output should match with output_length)

Extraction function is defined by user. The extraction function will take as input list of nextnanopy datafiles. Visit this tutorial to get to know about nextnanopy datafiles Tutorial 2 - Plotting data files.

To create Metric instance, use the following syntax

from nextnanoevo.MetricExtractor import Metric
import numpy as np


def dummy_extraction_function(dfiles):
    """
    This is an example of simple extraction function that takes first value of first variable from 2 datafiles.
    Reutrn the vector contatining sum and product of these values.
    """
    dfile1 = dfiles[0]
    dfile2 = dfiles[1]

    val1 = dfile1.variables[0].value[0]
    val2 = dfile1.variables[0].value[0]

    return np.array([val1+val2, val1*val2])


nn_metric = Metric(input_length=2, output_length=2, extraction_function=dummy_extraction_function)

# input_length = 2, because the metric processes 2 datafiles
# output_length = 2, because it returns vector of length 2.

In some cases, the metric function can be dependent on the input variables. The workaround to achieve this is to use the output file that contains the input paramters (variables_input.txt in case of nextnano++). Lets have a look at the dummy example with metric dependent on one value from output file and one input variable.

from nextnanoevo.IO import IO
from nextnanoevo.MetricExtractor import Metric

# first, we define the IO instance with one normal output file and one output file with input variables
input_file_path = r'\path\to\input_file.in'

variables = ['variable_name1']
output_files = [('relative', 'path', 'datafile'),
                ('variables_input.txt', )]

nn_io = IO(input_file_path, variables, output_files)

def dummy_extraction_function_with_input(dfiles):
    """
    Example of extraction function that takes first value of datafile and input varibale and returns the ratio
    """
    dfile = dfiles[0]
    input_variables_dfile = dfiles[1]

    value = dfile.variables[0].value[0]
    input_value = input_variables_dfile.variables[0].value



    return np.array([value/input_value])

nn_metric = Metric(input_length=2, output_length=1, extraction_function=dummy_extraction_function_with_input)

# input_length = 2, because the metric processes 2 datafiles
# output_length = 1, because it returns vector of length 1.