Arithmetic Operations

TensorFlow provides numerous arithmetic operations for tensor manipulations. These operations and many others in TensorFlow support broadcasting, making it easier to perform element-wise operations on tensors of different shapes.

Addition

For tensors addition we can use tf.add(), .assign_add() methods and a plus + sign. Also we can use broadcasting with the plus sign + or with the tf.add() method.

Broadcasting makes it possible to carry out element-wise operations on tensors of different, but compatible, shapes by virtually expanding the smaller tensor to match the shape of the larger tensor.

Note

For the inplace method, the fundamental element must be a mutable Variable type rather than a constant.

Subtraction

We have analogues of all methods for subtraction as for addition:

• tf.add() changes into tf.subtract();
• Plus sign + changes into minus sign -;
• .assign_add() changes into .assign_sub().

Multiplication (Element-wise)

For multiplication, there isn't an inplace method since matrix multiplication inherently results in a new object. However, other operations have their counterparts:

• tf.add() corresponds to tf.multiply();
• The plus sign + corresponds to the asterisk sign *.

Division

Similar to multiplication, but with tf.divide() and / sign.

Broadcasting

Broadcasting is the term used to describe how tensors with different shapes are automatically and implicitly treated during arithmetic operations so that they appear as if they have the same shape. It allows operations to be performed on tensors of different sizes without explicitly resizing them first.