Returning the Hessian of the empirical risk

[tranlm]

Hi,

I came across your implementation of the Influence Function and am interested in separately grabbing the Hessian that is needed for the calculations. Is there a way for me to grab it without having to break apart the code?

thank you very much.

[zironycho]

Hi tranlm,

Did you mean Hessian vector product like here? If you are using python tensorflow, you have to import python module and just call as below.

from tensorflow.python.ops.gradients_impl import _hessian_vector_product

_hessian_vector_product(ys, xs, v)

In our code in here for Hessian vector product.

Thank you.

[tranlm]

Thanks!

I noticed while looking into this that the original authors have also released their code implementation of the paper. Have you noticed any implementation differences?

[teradepth]

Our implementation is basically not much different. The original one only considered output layer as trainable parameters to calculate influence and it only applied to their experimental networks, but ours use all layers and is more general applicable to any networks, and more optimized and easy to use.