Christopher Olah

@ncammarata - Great point. I'd only been thinking about the container. I'll need to ponder that a bit.

@ncammarata This is a really interesting use case. Totally feasible and would be nice to have.

Hi Cristina, Is it possible you're using an out of date version of lucid? Chris On Thu, Feb 13, 2020 at 11:17 AM Cristina Segalin wrote: > Trying to run...

Hello! Try doing this instead: `param_f = lambda: param.color.to_valid_rgb(param.spatial.naive((1,28,28,1)))`

Lucid needs an input parameterization that matches the model. The default one is 3 channels. If you have a model with a different number of channels, you need to change...

Ah, you are using `make_vis_T()` instead of `render_vis()`. Yep, you need to create a session. :) Try doing this: ```py logit_images = [] print("Logit ", end="") for logit_idx in range(10):...

This is absolutely possible! There's a bunch of ways you could do it, depending on how fine grained you need the control to be. The more principled fine-grained ones are,...

There's a detailed paper covering it here: https://distill.pub/2018/differentiable-parameterizations/ (see section 3) Unfortunately, generalizing it to other media would be a significant research problem.

Some random interesting things: * The influence of training pre-processing on caricatures (by @ludwigschubert):  * Comparison of three different InceptionV1 implementations (probably due to pre-processing):  * Example of...

I've been thinking about "attribution caricatures" a lot more. See [examples](https://storage.googleapis.com/clarity-public/demos/attribution-caritcatures/index.html) and [notebook](https://colab.research.google.com/github/tensorflow/lucid/blob/master/notebooks/misc/attribution_caricatures.ipynb). Attribution caricatures can be made to the output classes, as we saw earlier:  But they can...