lucid.newaxis

The newaxis is used to add a new dimension to a tensor or array, effectively increasing its dimensionality by one.

This is commonly used in machine learning and scientific computing for reshaping data and ensuring compatibility with broadcasting rules.

Usage

The newaxis can be used in indexing to create a new axis in a tensor’s shape.

This is especially useful for aligning tensors with different shapes for element-wise operations or reshaping data.

Example 1: Adding a new axis to a 1D tensor

In this example, newaxis is used to add a new axis at the beginning of a 1D tensor.

>>> import lucid
>>> a = Tensor([1, 2, 3])
>>> print(a.shape)  # Output: (3,)
>>> a = a[None, :]  # Adds a new axis at the beginning
>>> print(a.shape)  # Output: (1, 3)

The shape of the tensor a is transformed from (3,) to (1, 3) by adding a new axis.

Example 2: Adding a new axis to a 2D tensor

Here, we add a new axis in the middle of a 2D tensor.

>>> b = Tensor([[1, 2], [3, 4]])
>>> print(b.shape)  # Output: (2, 2)
>>> b = b[:, None, :]  # Adds a new axis in the second position
>>> print(b.shape)  # Output: (2, 1, 2)

In this case, the shape of the tensor b changes from (2, 2) to (2, 1, 2).

Example 3: Adding a new axis to align tensors for broadcasting

Sometimes, newaxis is used to align tensors for broadcasting in operations like addition or multiplication.

>>> c = Tensor([1, 2, 3])
>>> d = Tensor([[4], [5], [6]])
>>> print((c + d).shape)  # Output: (3, 3)

By adding a new axis to c, it aligns with the shape of d, allowing the element-wise addition to work.

Conclusion

The newaxis is a simple yet powerful tool for reshaping tensors by adding new axes.

It is especially helpful for aligning tensors to make them compatible for broadcasting and for preparing data in machine learning tasks.