Inclusion of Houdini based loss for better attacks

[iamgroot42]

Having read about Houdini and its success in various domains to construct adversarial samples, it would be great to have it in cleverhans.

[goodfeli]

Yes, any PR is very welcome.

[npapernot]

This would be nice to address once #423 is merged

[feedforward]

@npapernot I working on this, will be able to submit PR in a week.

[npapernot]

Thanks @feedforward. It might make sense to hold on before #423 is merged so that we can use the new loss class directly.

[npapernot]

@feedforward now that #423 is merged, this would be a good time to make the PR. Thanks!

[feedforward]

@npapernot I am working on it, will submit the PR in couple of days, thanks for the info.

[vikramnitin9]

@npapernot, @feedforward - I'd like to work on this, if there's no PR already in the works. Any details I should know before starting off?

[npapernot]

Taking into account the instructions you can find in the main README, you should feel free to create a PR that addresses any issue open in this tracker. Thanks for your help

@feedforward do you have any WIP for this issue or should @vikramnitin9 start from scratch?

[npapernot]

@vikramnitin9 any update on this? Thanks!

[vikramnitin9]

Hi @npapernot, I've started a WIP PR (#967).

Thanks for the reminder and apologies for the delay. I've been busy with a lot of other stuff in the last two months, hence couldn't look into this earlier.