How To Subtraction of 2 Tensors In TensorFlow?

We will learn how to do suntraction in TensorFlow using tf.subtract() function.

tf.subtract() : Do Element wise subtraction with x & y
It can subtract list, tuple, scaler, TensorFlow variable/constant/placeholder/SparceMatrix with each other and with scaler and with list/tuple.

Note: – operator can be used to subtract 2 tensors

Syntax: tf.subtract(x, y, name=None)

Args
xTensor. Must be one of the following types: bfloat16halffloat32float64uint8int8uint16int16int32int64complex64complex128uint32uint64.
yTensor. Must have the same type as x.
nameA name for the operation (optional).
Returns
Tensor. Has the same type as x.

Subtraction of 2 Variables with TensorFlow

import tensorflow as tf
import numpy as np

Scaler subtract

a = 10
b = 20

tf.subtract(a,b)

subtract 2 List

l1 = [1,2,3,4]
l2 = [5,6,7,8]

tf.subtract(l2, l1)

subtract 2 Tuple

t1 = (1,2,3,4)
t2 = (5,6,7,8)

tf.subtract(t1,t2)

subtract List/Tuple with Scaler

l1 = [1,2,3,4]
a =10

tf.subtract(l1, a)

subtract of 2 list/Tuple with Diiferent Size

If the size of 2 variables is different then tf.subtract() function will through error. If any one of the variables contains only a single item then it will do the element-wise operation.

#this code will through error
l1 = [1,2,3,4]
l2 = [5,6]

tf.subtract(l1, l2)

subtract 2 Numpy Array

npa1 = np.array([1,2,3,4,5])
npa2 = np.array([1,2,3,4,5])

tf.subtract(npa1, npa2)

subtract Numpy array with Scaler

npa1 = np.array([1,2,3,4,5])
a = 10

tf.subtract(npa1, a)

subtract Numpy array with single element np array

npa1 = np.array([1,2,3,4,5])
npa2 = np.array([[1]])

tf.subtract(npa1, npa2)

subtract 2 TF Constant Tensor

tf_ct1 = tf.constant([1,2,3,4])
tf_ct2 = tf.constant([1,2,3,4])

tf.subtract(tf_ct1, tf_ct2)

subtract TF Constant Tensor with Scaler

tf_ct1 = tf.constant([1,2,3,4])
a = 10
tf.subtract(tf_ct1, a)

subtract 2 TF Variable

tf_vr1 = tf.Variable([1,2,3,4])
tf_vr2 = tf.Variable([1,2,3,4])

tf.subtract(tf_vr1, tf_vr2)

subtract TF Variable with Scaler

tf_vr1 = tf.Variable([1,2,3,4])
a = 10
tf.subtract(a, tf_vr1)

subtract 2 Numpy array with different Shape

x = np.ones(6).reshape(1, 2, 1, 3) # (2,2,3,3)
print("x: ", x)
y = np.ones(6).reshape(2, 1, 3, 1)
print("y: ", y)
 
tf.subtract(x, y)

Leave a Reply