plotting.py

from unicodedata import name
import numpy as np
import matplotlib.pyplot as plt
import torch.optim as optim
from torch.utils.data import TensorDataset
import torch
import random

torch.manual_seed(0)
random.seed(0)
np.random.seed(0)

from main import (
    LOSS_FUNCTION,
    IrisClassifier,
    test,
    EPOCHS,
    HIDDEN_LAYER,
)


def plot_change_in_lr(lr_to_try, testing_data, training_data, output_path = "lr_hyperparam_behaviour.png"):
    accuracy_tests = np.zeros_like(lr_to_try)
    
    for idx, lr in enumerate(lr_to_try):
        model = IrisClassifier(HIDDEN_LAYER)
        opt = optim.Adam(model.parameters(), lr=lr)
        fit(model, EPOCHS, training_data, testing_data, opt, LOSS_FUNCTION, False)
        accuracy_tests[idx] = test(model,testing_data)
    
    plt.plot(lr_to_try, accuracy_tests,'o')
    plt.ylabel("accuracy (percentage)")
    plt.xlabel("learning rate")
    plt.title("Change in classification accuracy depending on lr (other params fixed)")
    plt.tight_layout()
    plt.savefig(output_path)
    
    accuracy_tests = torch.tensor(accuracy_tests)
    _, index = torch.max(accuracy_tests, 0)
    
    print(f'The best lr is : {lr_to_try[index]} which gives a testing accuracy of : {accuracy_tests[index]}%')




def example():
    fig = plt.figure(figsize=(16, 6))
    rows = 2
    columns = 4
    grid = plt.GridSpec(rows, columns, wspace = .25, hspace = .25)
    for i in range(rows*columns):
        exec (f"plt.subplot(grid{[i]})")
        plt.annotate('subplot 24_grid[' + str(i) + ']', xy = (0.5, 0.5), va = 'center', ha = 'center')
    plt.show()



def main_plt():
    # X_train, X_test,y_train,y_test, y_names = load_data()
    # train_dataset = TensorDataset(X_train, y_train)
    # test_dataset = TensorDataset(X_test, y_test)
    # train_dataloader, test_dataloader = tensordatasets_to_loaders(train_dataset, test_dataset, BATCH_SIZE)
        
    # plot_change_in_lr([0.01,0.001,0.1,0.005,0.05], test_dataloader, train_dataloader)#, "lr_zoom_in.png"))
    # plot_change_in_lr([0.001], test_dataloader, train_dataloader)#, "lr_zoom_in.png"))
    example()

if __name__ == '__main__':
    main_plt()