main.py

import numpy as np
from tkba import TKBA
from visualization import plot_tkba
import matplotlib.pyplot as plt
from scipy.io import loadmat


def main():
    # Parameters
    m_iter = 1     # Number of iterations
    nr = 0.1       # Noise Rate [0,1]

    print("Loading Dataset...")
    mat_data = loadmat('2D_ClusteringDATASET.mat')
    # Extract the data
    print("Available variables in the .mat file:", list(mat_data.keys()))
    
    if 'data' in mat_data:
        data = mat_data['data']
    elif 'DATA' in mat_data:
        data = mat_data['DATA']
    else:
        print("Dataset variable not found")
        return
    
    print(f"Original dataset shape: {data.shape}")
    
    # Extract only the first two columns if there are more than 2
    if data.shape[1] >= 2:
        data = data[:, :2]  # Take first two columns as in the MATLAB code
    else:
        print("Error: Data must have at least 2 dimensions for clustering")
        return
    
    print(f"Using dataset with {data.shape[0]} samples")
    
    # Add some noise if requested
    if nr > 0:
        noise_data = np.random.rand(int(data.shape[0] * nr), data.shape[1])
        # Replace first portion of data with noise
        n_noise = noise_data.shape[0]
        data[:n_noise, :] = noise_data
    
    # Normalize data to [0, 1] range as in original MATLAB code
    data_min = data.min(axis=0)
    data_max = data.max(axis=0)
    data = (data - data_min) / (data_max - data_min + 1e-8)  # Add small epsilon to avoid division by zero
    
    # Parameters of TKBA
    cim_sig = 0.05    # Kernel Bandwidth for CIM
    kbr_sig = 1.0     # Kernel Bandwidth for KBR
    max_cim = 0.2     # Vigilance Parameter by CIM [0~1]
    lambda_param = 400  # Interval for Node deletion and topology construction
    
    for nitr in range(m_iter):
        print(f'Iterations: {nitr+1}/{m_iter}')
        
        # Randomize data order
        ran = np.random.permutation(data.shape[0])
        data_random = data[ran, :]
        
        # Create and fit the TKBA model
        tkba_net = TKBA(
            cim_sig=cim_sig,
            kbr_sig=kbr_sig,
            max_cim=max_cim,
            lambda_param=lambda_param
        )
        
        tkba_net.fit(data_random)
        
        # Print results
        print(f"Number of clusters found: {tkba_net.num_clusters}")
        print(f"Final edge matrix shape: {tkba_net.edge.shape}")
        print(f"Weight matrix shape: {tkba_net.weight.shape}")
        
        # Plot the results
        plot_tkba(data_random, tkba_net)
        
        plt.show()


if __name__ == "__main__":
    main()