- This repository contains the official implementation of our paper titled "Deep Feature Selection using Adaptive $\beta$-Hill Climbing aided Whale Optimization Algorithm for Lung and Colon Cancer Detection" published in Biomedical Signal Processing and Control, Elsevier.
- AdBet-WOA is a hybrid meta-heuristic optimization algorithm using the Adaptive
$\beta$ -Hill Climbing local search property as the exploitation agent in the popular Whale Optimization Algorithm. - The overall flowchart of AdBet-WOA and its Pseudo-algorithm are given as follows :
- The end-to-end implementation of our proposed algorithm on the LC25000 dataset is shown as follows :
directory> pip install -r requirements.txt
directory\AltWOA> python main.py -h
usage: main.py [-h] [--number_classes NUMBER_CLASSES] [--train_path TRAIN_PATH] [--validation_path VALIDATION_PATH] [--test_path TEST_PATH] [--epoch EPOCH]
[--lrate LRATE] [--batch_size_training BATCH_SIZE_TRAINING] [--batch_size_validation BATCH_SIZE_VALIDATION] [--load_model_wts LOAD_MODEL_WTS] [-- population_size POPULATION_SIZE] [--number_mutations NUMBER_MUTATIONS] [--epoch_for_algo EPOCH_FOR_ALGO]
optional arguments:
-h, --help show this help message and exit
--number_classes NUMBER_CLASSES Class distribution of data
--train_path TRAIN_PATH
Path to train data
--validation_path VALIDATION_PATH
Path to validation data
--test_path TEST_PATH
Path to test data
--epoch EPOCH Number of iterations
--lrate LRATE Learning rate
--batch_size_training BATCH_SIZE_TRAINING
Send batch size for training
--batch_size_validation BATCH_SIZE_VALIDATION
Send batch size for validation and testing
--load_model_wts LOAD_MODEL_WTS
True: if model weights are required to be loaded
--population_size POPULATION_SIZE
Set population size
--population_for_mating POPULATION_FOR_MATING
Set population for mating
--number_mutations NUMBER_MUTATIONS
Number of mutations
--epoch_for_algo EPOCH_FOR_ALGO
Set the epoch count for algorithm
directory/main.py -num_class {} -train {} -validation {} -test {} -epoch {} -lr {} -batch_tr {} -batch_val {} -load {} -pop_size {} -pop_mating {} -num_mut {} -epoch_algo {}
directory
|
+--> utils
| |
| +--> data.py # The python file which performs the dataloader initialization for train, val and test
| +--> feature_space.py # The python file used to store the generated features in csv files
| +--> trainer.py # The python file for training the dataset with our complete framework
|
+--> algorithm
| |
| +--> adbetwoa.py # The python file which contains our entire proposed algorithm
| +--> algo_utils.py # The python file containing other necessary peripheral functions related to algorithm
| +--> csv_read.py # The python file for reading the csv files storing generated features
| +--> transf_functions.py # The python file with other peripheral methods
|
+--> dos_class
| |
| +--> resnet_encoder.py # The python file which generated the features for two-class classification
|
+--> tres_class
| |
| +--> effnet_encoder.py # The python file which generated the features for three-class classification
|
+--> cinco_class
| |
| +--> effnet_encoder.py # The python file which generated the features for five-class classification
|
+--> model_weights # Folder to save the wiights of the best performing model
|
|
|
+--> main.py # The main function which reads the arguments dependiong on user input and performs the whole
end-to-end operation in a chronological order