re-executions of activation_maximization generate different results for same graph/weights

[omerbrandis]

hello,

I've made the following minor change to the visualize dense layer example:

for CurrentAttempt in range(0,4):
        print('attempt number {}'.format(CurrentAttempt))
        activation = activation_maximization(loss,callbacks=[Print(interval=50)])
        image = activation[0].astype(np.uint8)

        subplot_args = { 'nrows': 1, 'ncols': 1, 'figsize': (3, 3),
                         'subplot_kw': {'xticks': [], 'yticks': []} }
        f, ax = plt.subplots(**subplot_args)
        ax.imshow(image)
        ax.set_title('Ouzel', fontsize=14)
        plt.tight_layout()
        plt.show()

the generated images for ouzel class are different for each execution.
( i can upload the images if that will help ).

I'm not sure if this is an error of some kind or expected behavior .
can someone please shed some light on this ?
Omer.

[keisen]

Hi, @omerbrandis .

the generated images for ouzel class are different for each execution.

It's exactly expected behavior. There are two reasons for that.
First, in default, the initial values of the image is randomly generated.
(If you don't want that, you can gave arbitrary values as initial values to tf-keras-vis using seed_input option.)

tf-keras-vis/tf_keras_vis/activation_maximization.py

Line 15 in b420516

seed_input=None,

Second, in default, the image is randomly modified each step using InputModifiers.

tf-keras-vis/tf_keras_vis/activation_maximization.py

Line 17 in b420516

input_modifiers=[Jitter(2), Rotate(2)],

For that reason, the image during activation maximization is changing as follow.