how to set rmse cost function in tensorflow

Question

I have cost function in tensorflow.

activation = tf.add(tf.mul(X, W), b)
cost = (tf.pow(Y-y_model, 2)) # use sqr error for cost function

I am trying out this example. How can I change it to rmse cost function?

Answer 1

tf.sqrt(tf.reduce_mean(tf.square(tf.subtract(targets, outputs))))

And slightly simplified (TensorFlow overloads the most important operators):

tf.sqrt(tf.reduce_mean((targets - outputs)**2))

Answer 2

The formula for root mean square error is:

The way to implement it in TF is tf.sqrt(tf.reduce_mean(tf.squared_difference(Y1, Y2))).

---

The important thing to remember is that there is no need to minimize RMSE loss with the optimizer. With the same result you can minimize just tf.reduce_mean(tf.squared_difference(Y1, Y2)) or even tf.reduce_sum(tf.squared_difference(Y1, Y2)) but because they have a smaller graph of operations, they will be optimized faster.

But you can use this function if you just want to tract the value of RMSE.

Answer 3

Now we have tf.losses.mean_squared_error

Therefore,

RMSE = tf.sqrt(tf.losses.mean_squared_error(label, prediction))