python-learning

In the context of TensorFlow and most other frameworks, a tensor with shape [C] is a 1-dimensional tensor with C elements. It is not a 0-dimensional tensor, but rather a 1D tensor.

Let’s clarify the different types of tensor shapes:

1. 0-Dimensional Tensor (Scalar):

• A 0D tensor is a single value.
• Example: A single number like 3, which has the shape [] (empty tuple).
• Shape: []
• Example code:

  scalar_tensor = tf.constant(3)  # Shape is [] (0D tensor)

2. 1-Dimensional Tensor (Vector):

• A 1D tensor is a vector of elements. For example, a tensor with C=4 elements.
• Shape: [C]
• Example: [0., 1., 0., 0.] with shape (4,).
• Example code:

  vector_tensor = tf.constant([0., 1., 0., 0.])  # Shape is (4,) (1D tensor)

3. 2-Dimensional Tensor (Matrix):

• A 2D tensor is a matrix (a grid of values).
• Shape: [rows, columns]
• Example: A matrix with 2 rows and 3 columns: [[1, 2, 3], [4, 5, 6]] with shape (2, 3).
• Example code:

  matrix_tensor = tf.constant([[1, 2, 3], [4, 5, 6]])  # Shape is (2, 3) (2D tensor)

4. 3-Dimensional Tensor (Tensor of Tensors):

• A 3D tensor is essentially a 3D array.
• Shape: [depth, rows, columns]
• Example: A 3D tensor of shape (2, 3, 3) represents 2 matrices, each of size (3, 3).

5. A-batch of Tensors

• A 4D tensor is essentially a 4D array.
• Shape: [batch, height, width, channels]
• Example: np.ones((1, 4, 4, 3)) creates a 4D array with the following dimensions:
  • 1: This is often referred to as the batch size. It indicates that there's one "sample" in this array.
  • 4: This represents the height of the 2D spatial dimensions.
  • 4: This represents the width of the 2D spatial dimensions.
  • 3: This represents the number of channels (e.g., RGB for an image).
  • So, this creates a tensor where every entry is a 1.

---