Python | método tensorflow.math.argmin()

TensorFlow es una biblioteca de Python de código abierto diseñada por Google para desarrollar modelos de aprendizaje automático y redes neuronales de aprendizaje profundo. argmin() es un método presente en el módulo matemático de tensorflow. Este método se utiliza para encontrar el valor mínimo en los ejes.

Syntax:
tensorflow.math.argmin(
    input,axes,output_type,name
)

Arguments:
1. input: It is a tensor. Allowed dtypes for this tensor are float32,
          float64, int32, uint8, int16, int8, complex64, int64, qint8, quint8,
          qint32, bfloat16, uint16, complex128, half, uint32, uint64. 
2. axes: It is also a vector. It describes the axes to reduce the tensor.
         Allowed dtype are int32 and int64. Also [-rank(input),rank(input)) is the range allowed.
         axes=0 is used for vector.
3. output_type: It defines the dtype in which returned result should be.
                Allowed values are int32, int64 and the default value is int64.
4. name: It is an optional argument which defines name for the operation.
 
Return:
A tensor of output_type which contains the indices of the minimum value along the axes. 

Ejemplo 1:

# importing the library
import tensorflow as tf
 
# initializing the constant tensor
a = tf.constant([5,10,5.6,7.9,1,50]) # 1 is the minimum value at index 4
 
# getting the minimum value index tensor
b = tf.math.argmin(input = a)
 
# printing the tensor
print('tensor: ',b)
 
# Evaluating the value of tensor
c = tf.keras.backend.eval(b)
 
#printing the value
print('value: ',c)

Producción:

tensor:  tf.Tensor(4, shape=(), dtype=int64)
value: 4

Ejemplo 2: 

Este ejemplo usa un tensor de forma (3,3).

# importing the library
import tensorflow as tf
 
# initializing the constant tensor
a = tf.constant(value = [9,8,7,3,5,4,6,2,1],shape = (3,3))
 
# printing the initialized tensor
print(a)
 
# getting the minimum value indices tensor
b = tf.math.argmin(input = a)
 
# printing the tensor
print('Indices Tensor: ',b)
 
# Evaluating the tensor value
c = tf.keras.backend.eval(b)
 
# printing the value
print('Indices: ',c)

Producción:

tf.Tensor(
[[9 8 7]
 [3 5 4]
 [6 2 1]], shape=(3, 3), dtype=int32)
 Indices tensor: tf.Tensor([1 2 2], shape=(3,), dtype=int64)
 Indices: [1 2 2]