Create custom loss function in tensorflow with a loop

Question

I would like to create a custom loss function for a tensorflow model using y_true and y_pred but I got the following error: ValueError: Cannot infer num from shape (None, 1) this is my custom metric:

def custom_metric(y_true,y_pred):

    y_true = float(y_true)
    y_pred = float(y_pred)
    y_true = tf.unstack(y_true)
    y_pred = tf.unstack(y_pred)

    sqr_pred_error = K.square(y_true - y_pred)
    sqr_y_true = K.square(y_true)
    r = []
    for i in y_true:
        if sqr_pred_error[i] < sqr_y_true[i] or sqr_pred_error[i] == sqr_y_true[i]:
            result = 1
            print("result: 1")
        else:
            result = 0
            print("result: 0")
        r.append(result)
    r = tf.stack(r)

    return  K.sum(r)/K.shape(r)

Answer 1

You probably don't need a loop in there. Looks like you just need a bunch of 0s and 1s.

• 1 - If sqr_pred_error <= sqr_y_true
• 0 - else

Then you can do the following.

def custom_metric(y_true,y_pred):

    y_true = tf.cast(y_true, 'float32')
    y_pred = tf.cast(y_pred, 'float32')
    
    sqr_pred_error = K.square(y_true - y_pred)
    sqr_y_true = K.square(y_true)

    res = tf.where(sqr_pred_error<=sqr_y_true, tf.ones_like(y_true), tf.zeros_like(y_true))
    return  K.mean(res)