Christopher Brix

To my knowledge, the target sentence is generated word by word. Usually, the next word is calculated based on the input sequence and the decoder state. I'd like to pass...

I'm at a loss here. Currently, the only place where an `ExtendWithBeamLayer` layer is created, is from within the `get_layer` method of the `_SubnetworkRecCell`. But since this layer is only...

For Debona: ``` TOOL_NAME=Debona REPO=https://github.com/ChristopherBrix/Debona COMMIT=792575d18bb5f83cb8699dda6b9097dc41438e3d SCRIPTS_DIR=Debona ``` https://github.com/ChristopherBrix/Debona Thank you for organizing this competition!

Neurify can analyze images as inputs. To do so, they need to be transformed to text files (containing their respective pixel black/white values). For MNIST, this generates those files: https://github.com/tcwangshiqi-columbia/Neurify/tree/master/general/images

I'm not sure I understand - Neurify always needs a trained model and a concrete input for this model. You can substitute the image with basically anything else, you just...

Ok, I see what's confusing here. In addition to the concrete input, Neurify has a `INF` value (I think by default it's 10), that's used to define the maximal change...

I'm not the author of Neurify, but I've been working on bugfixes and improvements for the last few months. So far, I always focused on MNIST, so I'm not sure...

I don't see a reason why this shouldn't work, but you might have to change more. Neurify identifies the gold target label `g`, and tries to find a different label...

I've never used Neurify in this context (I'm not the inventor, but started working on an improved version). Maybe @tcwangshiqi-columbia could clarify whether Neurify supports safety properties out of the...