API tools faq
paste
pcstyle

Untitled

pcstyle
Nov 6th, 2025
374
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
Python 4.78 KB | None | 0 0
raw download clone embed print report
  1. import numpy as np
  2. import matplotlib.pyplot as plt
  3. from scipy.spatial.distance import cdist
  4. from sklearn.datasets import load_iris
  5. from sklearn.preprocessing import MinMaxScaler
  6. from sklearn.decomposition import PCA
  7.  
  8. def are_neighbors(grid2d, i, j):
  9.     """
  10.    odleglosc Chebyshev = 1 (8-neighborhood)
  11.    """
  12.     di = abs(grid2d[i, 0] - grid2d[j, 0])
  13.     dj = abs(grid2d[i, 1] - grid2d[j, 1])
  14.     return max(di, dj) == 1  # true jesli sasiaduja
  15.  
  16. def movmean(data, window):
  17.     """
  18.    srednia ruchoma
  19.    """
  20.     result = np.zeros_like(data)
  21.     half = window // 2
  22.     for i in range(len(data)):
  23.         start = max(0, i - half)
  24.         end = min(len(data), i + half + 1)
  25.         result[i] = np.mean(data[start:end])
  26.     return result
  27.  
  28. def som_iris_ex4():
  29.     rng = np.random.default_rng(7)
  30.     R, C = 12, 12  # wieksza siatka
  31.     M = R * C
  32.     iters = 4000  # wiecej iteracji dla wiekszych siatek
  33.     eta0 = 0.12
  34.     sigma0 = max(R, C) / 2  # poczatkowe sasiedztwo (globalne porzadkowanie)
  35.     sigma_final_list = [1, 2, 3]  # koncowe rozmiary sasiedztwa do testowania
  36.    
  37.     iris = load_iris()
  38.     X = iris.data
  39.     scaler = MinMaxScaler()
  40.     X = scaler.fit_transform(X)  # 150x4, kazda cecha do [0,1]
  41.     S = X.shape[0]
  42.    
  43.     print(f"loaded iris: {S} samples, {X.shape[1]} features")
  44.     print(f"grid: {R}x{C} = {M} neurons")
  45.     print(f"testing sigma_final: {sigma_final_list}")
  46.    
  47.     gx, gy = np.meshgrid(np.arange(1, C+1), np.arange(1, R+1))
  48.     grid2d = np.column_stack([gx.ravel(), gy.ravel()])
  49.    
  50.     fig1 = plt.figure(figsize=(10, 6), facecolor='white')
  51.     ax1 = fig1.add_subplot(111)
  52.     ax1.set_xlabel('Iteration')
  53.     ax1.set_ylabel('Topographic Error')
  54.     ax1.set_title(f'R={R}, C={C} — Effect of final σ in {{1,2,3}}')
  55.     ax1.grid(True)
  56.     ax1.box(True)
  57.    
  58.     results = []
  59.    
  60.     for k, sigma_final in enumerate(sigma_final_list):
  61.         print(f'\n=== Testing sigma_final = {sigma_final} ===')
  62.        
  63.         W = rng.random((M, X.shape[1]))  # w [0,1], ten sam skala co X
  64.        
  65.         topoErr = np.zeros(iters)
  66.        
  67.         tau = iters  # jedna stala czasowa przez caly przebieg
  68.        
  69.         for t in range(1, iters + 1):
  70.             sigma_t = sigma_final + (sigma0 - sigma_final) * np.exp(-t / tau)
  71.             eta_t = eta0 * np.exp(-t / tau)
  72.            
  73.             x = X[rng.integers(S), :]
  74.            
  75.             dists = np.sum((W - x) ** 2, axis=1)
  76.             idx = np.argsort(dists)
  77.             bmu1 = idx[0]
  78.             bmu2 = idx[1]
  79.            
  80.             dgrid2 = np.sum((grid2d - grid2d[bmu1, :]) ** 2, axis=1)
  81.             h = np.exp(-dgrid2 / (2 * sigma_t ** 2))
  82.             h = h / np.max(h)
  83.            
  84.             W = W + eta_t * (h[:, np.newaxis] * (x - W))
  85.            
  86.             topoErr[t-1] = 0 if are_neighbors(grid2d, bmu1, bmu2) else 1
  87.            
  88.             if t % 1000 == 0:
  89.                 print(f"  iteration {t}/{iters}, sigma={sigma_t:.3f}, eta={eta_t:.5f}")
  90.        
  91.         win = 50
  92.         topoErr_s = movmean(topoErr, win)
  93.        
  94.         results.append({
  95.             'sigma_final': sigma_final,
  96.             'W': W.copy(),
  97.             'topoErr': topoErr.copy(),
  98.             'topoErr_s': topoErr_s.copy()
  99.         })
  100.        
  101.         # rysuj krzywą TE
  102.         ax1.plot(range(1, iters + 1), topoErr_s,
  103.                 label=f'σ_final={sigma_final}', linewidth=1.5)
  104.    
  105.     ax1.legend(loc='northeast')
  106.     plt.tight_layout()
  107.     plt.show()
  108.    
  109.     fig2 = plt.figure(figsize=(15, 5), facecolor='white')
  110.    
  111.     for k, result in enumerate(results):
  112.         W = result['W']
  113.        
  114.         pca = PCA(n_components=2)
  115.         W2 = pca.fit_transform(W)
  116.        
  117.         ax = fig2.add_subplot(1, len(sigma_final_list), k + 1)
  118.         ax.set_aspect('equal')
  119.         ax.set_title(f'σ_final={result["sigma_final"]}')
  120.         ax.box(True)
  121.        
  122.         WX = W2[:, 0].reshape((R, C))
  123.         WY = W2[:, 1].reshape((R, C))
  124.        
  125.         # linie poziome
  126.         for r in range(R):
  127.             ax.plot(WX[r, :], WY[r, :], '-', linewidth=1.0, color='blue', alpha=0.6)
  128.        
  129.         # linie pionowe
  130.         for c in range(C):
  131.             ax.plot(WX[:, c], WY[:, c], '-', linewidth=1.0, color='blue', alpha=0.6)
  132.        
  133.         # male znaczniki dla neuronow
  134.         ax.plot(W2[:, 0], W2[:, 1], 'ko', markersize=2, markerfacecolor='k')
  135.    
  136.     plt.suptitle('Final Lattices Projected by PCA', fontsize=14)
  137.     plt.tight_layout()
  138.     plt.show()
  139.    
  140.     print('\n== Summary (lower TE is better) ==')
  141.     for result in results:
  142.         te_mean = np.mean(result['topoErr'][-500:])  # srednia z ostatnich 500
  143.         print(f'sigma_final={result["sigma_final"]} -> mean TE (last 500 iters): {te_mean:.4f}')
  144.  
  145. if __name__ == '__main__':
  146.     som_iris_ex4()
  147.  
  148.  
Advertisement
Add Comment
Please, Sign In to add comment
Public Pastes
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!