如何将数组转换为矩阵？

Question

我有一个具有以下形状的数据集：(2400, 2) (2400,) (1600, 2) (1600,) 我的任务是通过二元逻辑回归执行非线性可分离分类。 但我在可视化部分收到以下错误：

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-19-2754b9327868> in <module>()
      4 
      5 # Plot different regions and color them
----> 6 output = output.reshape(x_vals.shape)
      7 plt.imshow(output, interpolation='nearest', 
      8            extent=(x_min, x_max, y_min, y_max),

ValueError: cannot reshape array of size 2880000 into shape (1200,1200)

如何将数组重塑为矩阵？ 以下是我的实现供参考：

num_features = 2
learning_rate = 0.0001
training_steps = 4000
batch_size = 32
display_step = 50

x_train, y_train = map(list, zip(*[(x,y) for x,y in zip(x_train, y_train) if y==0 or y==1]))
x_test, y_test = map(list, zip(*[(x,y) for x,y in zip(x_test, y_test) if y==0 or y==1]))

x_train, x_test = np.array(x_train, np.float32), np.array(x_test, np.float32)
y_train, y_test = np.array(y_train, np.int64), np.array(y_test, np.int64)

x_train, x_test = x_train.reshape([-1, num_features]), x_test.reshape([-1, num_features])

x_train, x_test = x_train/255., x_test/255.

train_data = tf.data.Dataset.from_tensor_slices((x_train, y_train))
train_data = train_data.repeat().shuffle(5000).batch(batch_size).prefetch(1)

b = tf.Variable(tf.ones((num_features, 2)) * 0.000001, name = "weight")
b0 = tf.Variable(0., name = "bias")

def logistic_regression(x, b, b0):
  return 1. / (1. + tf.exp(-tf.matmul(x, b) - b0))

def loglikelihood(p, y_true):
  return tf.reduce_sum(tf.one_hot(y_true, 2) * tf.math.log(p), axis=-1)

def accuracy(y_pred, y_true):
  correct_prediction = tf.equal(tf.argmax(y_pred, axis=-1), y_true)
  return tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

optimizer = tf.optimizers.Adam()

for step, (batch_x, batch_y) in enumerate(train_data.take(training_steps), 1):
  with tf.GradientTape() as g:
    g.watch([b, b0])
    p = logistic_regression(batch_x, b, b0)
    ll = loglikelihood(p, batch_y)
    ll_sum = tf.reduce_mean(ll)
  grad_b, grad_b0 = g.gradient(ll_sum, [b, b0])
  optimizer.apply_gradients(zip([grad_b, grad_b0], [b, b0]))

  if step % display_step == 0:
    p = logistic_regression(batch_x, b, b0)
    acc = accuracy(p, batch_y)
    p = logistic_regression(x_test, b, b0)
    val_acc = accuracy(p, y_test)
    print("step: %i, acc: %f, val_acc %f" % (step, acc, val_acc))

def predict(x_test):
    return tf.round(logistic_regression(x_test, b, b0))

import numpy as np

x_min, y_min = -12, -12
x_max, y_max = 12, 12
x_vals, y_vals = np.meshgrid(np.arange(x_min, x_max, 0.02), np.arange(y_min, y_max, 0.02))

xy_grid = pd.DataFrame(zip(x_vals.ravel(), y_vals.ravel()), dtype=np.float32)

# Predict output labels for all the points on the grid
output = predict(xy_grid.to_numpy()).numpy()

import matplotlib.pyplot as plt

fig, ax = plt.subplots(1, 1)

# Plot different regions and color them
output = output.reshape(x_vals.shape)
plt.imshow(output, interpolation='nearest', 
           extent=(x_min, x_max, y_min, y_max),
           cmap=plt.cm.Paired,
           aspect='auto',
           origin='lower')

pd.DataFrame(np.concatenate([x_train, 
                             np.expand_dims(y_train, axis=-1)], axis=1)).plot.scatter(0, 1, c=2, colormap='viridis', ax=ax)

预期的结果应该是这样的： expected image

但我得到以下图像： resulting image

Answer 1

错误是由于 x_vals.shape 太小造成的。因此 numpy 无法重塑数据，因为您会丢失一些数据。

您需要将 x_vals 的形状更改为您希望看到的任何形状输出。

例如，如果您想要 1200x2400 的图像，您可以这样做：

x_min, y_min = -12, -12
x_max, y_max = 12, 12
x_vals, y_vals = np.meshgrid(np.arange(x_min, x_max, 0.01), np.arange(y_min, y_max, 0.02))

print(x_vals.shape) # shows x_vals.shape = (1200, 2400) which is 28880000 elements

output = np.arange(2880000)

output = output.reshape(x_vals.shape) # this will now work