cnn.py

#!/usr/bin/env python
# ------------------------------------------------------------------------------------------------------%
# Created by "Thieu Nguyen" at 20:48, 25/03/2020                                                        %
#                                                                                                       %
#       Email:      nguyenthieu2102@gmail.com                                                           %
#       Homepage:   https://www.researchgate.net/profile/Thieu_Nguyen6                                  %
#       Github:     https://github.com/thieunguyen5991                                                  %
# -------------------------------------------------------------------------------------------------------%

from os.path import splitext, basename, realpath
from sklearn.model_selection import ParameterGrid
from models.main.traditional_rnn import Cnn1
from utils.IOUtil import _load_dataset__
from utils.Settings import *
from utils.Settings import cnn1_final as param_grid

if SPF_RUN_TIMES == 1:
    all_model_file_name = SPF_LOG_FILENAME
else:  # If runs with more than 1, like stability test --> name of the models ==> such as: rnn1hl.csv
    all_model_file_name = str(splitext(basename(realpath(__file__)))[0])


def train_model(item):
    root_base_paras = {
        "data_original": dataset,
        "train_split": SPF_TRAIN_SPLIT,  # should use the same in all test
        "data_window": data_window,  # same
        "scaling": SPF_SCALING,  # minmax or std
        "feature_size": SPF_FEATURE_SIZE,  # same, usually : 1
        "network_type": SPF_3D_NETWORK,  # RNN-based: 3D, others: 2D
        "n_runs": SPF_RUN_TIMES,  # 1 or others
        "log_filename": all_model_file_name,
        "path_save_result": SPF_PATH_SAVE_BASE + SPF_DATA_FILENAME[loop] + "/",
        "draw": SPF_DRAW,
        "log": SPF_LOG
    }
    paras_name = "hs_{}-ep_{}-bs_{}-lr_{}-ac_{}-op_{}-lo_{}-dr_{}-fs_{}-ks_{}-ps_{}".format(item["hidden_sizes"], item["epoch"], item["batch_size"],
                    item["learning_rate"], item["activations"], item["optimizer"], item["loss"], item["dropouts"],
                    item["filters_size"], item["kernel_size"], item["pool_size"])
    root_rnn_paras = {
        "hidden_sizes": item["hidden_sizes"], "epoch": item["epoch"], "batch_size": item["batch_size"],
        "learning_rate": item["learning_rate"], "activations": item["activations"], "optimizer": item["optimizer"],
        "loss": item["loss"], "dropouts": item["dropouts"], "paras_name": paras_name
    }
    cnn_paras = {
        "filters_size": item["filters_size"], "kernel_size": item["kernel_size"], "pool_size": item["pool_size"]
    }
    md = Cnn1(root_base_paras=root_base_paras, root_rnn_paras=root_rnn_paras, cnn_paras=cnn_paras)
    md._running__()


for _ in range(SPF_RUN_TIMES):
    for loop in range(len(SPF_DATA_FILENAME)):
        filename = SPF_LOAD_DATA_FROM + SPF_DATA_FILENAME[loop]
        dataset = _load_dataset__(filename, cols=SPF_DATA_COLS[loop])
        feature_size = len(SPF_DATA_COLS[loop])
        data_window = SPF_DATA_WINDOWS[loop]
        # Create combination of params.
        for item in list(ParameterGrid(param_grid)):
            train_model(item)