请给出一个SOM算法的示例，基于python

非常感谢您的提问。以下是一个基于Python的SOM算法示例:

import numpy as np

class SOM:
    def __init__(self, input_dims, output_dims, learning_rate=0.1, radius=10, seed=None):
        if seed:
            np.random.seed(seed)
        self.input_dims = input_dims
        self.output_dims = output_dims
        self.learning_rate = learning_rate
        self.radius = radius

        self.weights = np.random.normal(0, 1, size=(output_dims[0], output_dims[1], input_dims))

    def get_bmu(self, x):
        distances = np.sqrt(np.sum((self.weights - x) ** 2, axis=2))
        return np.unravel_index(np.argmin(distances), distances.shape)

    def update_weights(self, x, bmu):
        for i in range(self.output_dims[0]):
            for j in range(self.output_dims[1]):
                d = np.sqrt((i - bmu[0]) ** 2 + (j - bmu[1]) ** 2)
                if d <= self.radius:
                    h = np.exp(-(d ** 2) / (2 * self.radius ** 2))
                    self.weights[i, j] += self.learning_rate * h * (x - self.weights[i, j])

    def train(self, X, epochs):
        for epoch in range(epochs):
            np.random.shuffle(X)
            for x in X:
                bmu = self.get_bmu(x)
                self.update_weights(x, bmu)
            self.radius *= 0.5
            self.learning_rate *= 0.5

    def predict(self, X):
        y = np.zeros(len(X), dtype=np.uint8)
        for i, x in enumerate(X):
            bmu = self.get_bmu(x)
            y[i] = bmu[0] * self.output_dims[1] + bmu[1]
        return y

input_dims = 4
output_dims = (10, 10)
X = np.random.normal(0, 1, size=(1000, input_dims))
som = SOM(input_dims, output_dims)
som.train(X, 100)
y = som.predict(X)
print(y)

希望这个示例能够对您有所帮助。如果您还有其它问题，请随时向我提出。