感知器實現鳶尾花分類

import tensorflow as tf
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn import preprocessing


def init_weights(shape):
    return tf.Variable(tf.random_normal(shape, stddev=0.01))


# 建立分割線
def plotLine(slope, bias):
    x = np.arange(-3, 3, 0.5)
    y = x * slope + bias
    plt.plot(x, y)


if __name__ == "__main__":
    # 導入數據
    df = pd.read_csv('https://archive.ics.uci.edu/ml/machine-learning-databases/iris/iris.data', header=None)
    features = df.iloc[1:len(df.index), [0, 2]].values
    labels = df.iloc[1:len(df.index), 4].values

    # 調節數據
    scaler = preprocessing.StandardScaler().fit(features)
    features_standard = scaler.transform(features)

    # 選取了兩種花的兩類特徵
    for index, label in enumerate(labels):
        if label == "Iris-setosa":
            plt.scatter(features[index, 0], features[index, 1], color='red', marker='o', label='setosa')
        else:
            plt.scatter(features[index, 0], features[index, 1], color='blue', marker='x', label='versicolor')
    plt.xlabel('petal len')
    plt.ylabel('sepal len')
    plt.show()

    # 轉換標籤
    labels = np.where(labels == "Iris-setosa", 1, -1)

    # 使用sklearn類庫快速分割數據集
    features_train, features_test, labels_train, labels_test = \
        train_test_split(features_standard, labels, test_size=0.33)

    X = tf.placeholder(tf.float32)
    Y = tf.placeholder(tf.float32)

    w = init_weights([2, 1])
    b = tf.Variable(tf.zeros([1, 1]))

    predict_Y = tf.sign(tf.matmul(X, w) + b)

    loss = tf.reduce_mean(tf.square(predict_Y - labels_train))

    optimizer = tf.train.GradientDescentOptimizer(0.01)
    train_step = optimizer.minimize(loss)

    init = tf.global_variables_initializer()
    sess = tf.Session()
    sess.run(init)

    # start train
    for i in range(1000):
        sess.run(train_step, feed_dict={X: features_train, Y: labels_train})

    w1 = sess.run(w).flatten()[0]
    w2 = sess.run(w).flatten()[1]
    b = sess.run(b).flatten()

    for index, label in enumerate(labels_test):
        if label == 1:
            plt.scatter(features_test[index, 0], features_test[index, 1], color='red', marker='o', label='setosa')
        else:
            plt.scatter(features_test[index, 0], features_test[index, 1], color='blue', marker='x', label='versicolor')

    plt.xlabel('petal len')
    plt.ylabel('sepal len')
    plotLine(-w1 / w2, -b / w2)
    plt.show()