In a tensor, the index starts with 0.
Syntax:
Parameter:
1. tensor_input is a tensor that has numeric elements.
It can be 1or 2 dimensional.
2. If the tensor is two-dimensional, then it is possible to specify the axis to get an index of minimum values in rows or columns.
If axis=0, the index of minimum values is returned across column-wise, and if axis=1, the index of minimum values is returned across row-wise.
If the axis is not specified, then it will return the index of minimum values column-wise.
Return
Return a Tensor with the minimum value indices.
Example 1:
Let’s create a one-dimensional tensor in js that has integer values and return a minimum value index.
<!-- CDN Link that delivers the Tensorflow.js framework -->
<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs"></script>
<body>
<center><h2>Tensorflow.js - tf.argMin() </h2></center>
<script>
let values = tf.tensor1d([34,56,78,90]);
//actual tensor
document.write("Actual Tensor: ",values);
document.write("<br>");
document.write("<br>");
//apply argMin() on the above tensor
document.write("Minimum value Position:- "+tf.argMin(values));
</script>
</body>
</html>
Output:
34 is the minimum among all the elements, and it is present in the 1st position. An index is 0.
So 0 is returned.
Example 2:
Let’s create a tensor that has 2 dimensions with 4 rows and 2 columns that has integer values and return minimum value indices across columns.
<!-- CDN Link that delivers the Tensorflow.js framework -->
<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs"></script>
<body>
<center><h2>Tensorflow.js - tf.argMin() </h2></center>
<script>
let values = tf.tensor2d([10,13,15,6,67,5,10,2],[4,2]);
//actual tensor
document.write("Actual Tensor: ",values);
document.write("<br>");
document.write("<br>");
//apply argMin() on the above tensor
document.write("Minimum value Positions across columns:- "+tf.argMin(values,0));
</script>
</body>
</html>
Output:
Working:
Minimum value among (10,15,67,10) is 10 and (13,6,5,2) is 2.
Index positions of 10 and 2 are 0 and 3.
Example 3:
Let’s create a tensor that has 2 dimensions in js with 4 rows and 2 columns that has integer values and return minimum value indices across rows.
<!-- CDN Link that delivers the Tensorflow.js framework -->
<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs"></script>
<body>
<center><h2>Tensorflow.js - tf.argMin() </h2></center>
<script>
let values = tf.tensor2d([10,13,15,6,67,5,10,2],[4,2]);
//actual tensor
document.write("Actual Tensor: ",values);
document.write("<br>");
document.write("<br>");
//apply argMin() on the above tensor
document.write("Minimum value Positions across rows:- "+tf.argMin(values,1));
</script>
</body>
</html>
Output:
Working:
Minimum values among [10,13] is 10, [15, 6] is 6, [67, 5] is 5 and [10, 2 ] is 2.
Index positions of the above minimum values are – 0,1,1,1.
Example 4:
Let’s create a tensor that has 2 dimensions in js with 4 rows and 2 columns that have integer values and return the indices of the minimum values.
<!-- CDN Link that delivers the Tensorflow.js framework -->
<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs"></script>
<body>
<center><h2>Tensorflow.js - tf.argMin() </h2></center>
<script>
let values = tf.tensor2d([1,2,3,4,5,6,7,8],[4,2]);
//actual tensor
document.write("Actual Tensor: ",values);
document.write("<br>");
document.write("<br>");
//apply argMin() on the above tensor
document.write("Minimum value index:- "+tf.argMin(values));
</script>
</body>
</html>
Output:
Working:
Minimum value in the [1,3,5,7] column is 1, and its index is 0.
Minimum value in the [2,4,6,8] column is 2, and its index is 0.
Conclusion
In this Tensorflow.js tutorial, we have seen how to return the index of minimum elements present in a tensor using the tf.argMin() method. In a 2D tensor, if axis=0, the index of minimum values is returned column-wise, and if axis=1, the index of minimum values is returned across row-wise. By default, it will return the index of minimum values column-wise.
