OpenAI Microscope - Visualizing Neural Networks

 So OpenAi has a new tool they have made available called Microscope.  It lets you visualize every significant layer and neuron in 8 different deep learning neural networks used for processing images.

The are AlexNet, AlexNet (Places), Inception V1, InceptionV1 (Places), VGG19, InceptionV3, Inception V4, and ResNet v2 50.


The OpenAI Microscope is based on two concepts, a location in a model and a technique. Metaphorically, the location is where you point the microscope, the technique is what lens you affix to it.


Our models are composed of a graph of “nodes” (the neural network layers), which are connected to each other through “edges.” Each op contains hundreds of “units”, which are roughly analogous to neurons. Most of the techniques we use are useful only at a specific resolution. For instance, feature visualization can only be pointed at a “unit”, not its parent “node”.

Check it out for ResNet v2 50.
This tool is amazing!

Currently they offer DeepDream and Caricature visualization techniques. DeepDream is an artificial, optimized image that maximizes the activations of all units in an op.  Caricature is an artificial, optimized image that maximizes the activations in response to a real, given image. You can select from a list of different images.

Once you dive into a specific layer, you get additional visualization options.

Once you pick a specific neuron, you can dive in even deeper.


Other visualization options may become available in microscope in the future.


They have also released the Lucid Library, which contains a ton of research code useful for creating visualizations.  Looks like it is all coded in tensorflow.  You can access the Lucid Library here.

The library contains a number of very useful Colab notebooks you can use to dive into trying out different aspects of the Lucid Library.


Hopefully some bits of the lucid library can get ported over to fastai api in PyTorch.

Why do we care so much about feature visualization?
Here's a great online paper by some folks at Google on 'The Building Blocks of Interpretability' that does a deep dive into trying to use feature visualization to help understand what is going on in neural nets..

Comments

Post a Comment

Popular posts from this blog

CycleGAN: a GAN architecture for learning unpaired image to image transformations

Image
 Like the Doublemint Twins touting the joys of Doublemint gum, 2 GANs are surely better than 1 GAN.  Especially if we package them together inside of one meta GAN module.  And this is exactly what the CycleGAN architecture does. Have you ever harbored dark secrets of turning a horse into a zebra?  The CycleGAN was developed to do just that. Learn how to turn a horse into a zebra.  And more. Now right away you can notice a difference between the image to image transformation GAN architectures we've been discussing over the last few posts.  Those last few posts described systems that learn from a database of matched input-output image pairs. And if your goal is to turn an edge representation into a nicely filled in continuous tone image, it's easy to build your database of matched input-output image pairs that your GAN system can then learn off of.  Take a continuous tone image (which will be the output of the database pair entry), then run it through an edge detector algorithm. 
Read more

Pix2Pix: a GAN architecture for image to image transformation

Image
 I thought following up yesterday's TraVelGAN post with a Pix2Pix GAN post would be useful to compare what is going on in the 2 architectures.  Two different approaches to the same problem. I stole this Pix2Pix Overview slide below from an excellent deeplearning.ai GAN course (note that they borrowed it from the original paper) because it gives you a good feel for what is going on inside of the Pix2Pix architecture.    Note how the Generator part is very much like an auto-encoder architecture, but rebuilt using the U-Net architecture features (based on skip-connections) that fastai has been discussing in their courses for several years before it became more widely known to the deep learning community at large  (and which originally came from an obscure medical image segmentation paper) . So the Generator in this Pix2Pix GAN is really pretty sophisticated, consisting of a whole image to image auto-encoder network with U-Net skip connections to generate better image quality at highe
Read more

Smart Fabrics

Image
"What makes smart fabrics revolutionary is that they have the ability to do many things that traditional fabrics cannot, including communicate, transform, conduct energy and even grow."  This is a quote from Rebeccah Pailes-Friedman, who is a professor at Pratt with a research focus on developing wearable technology and smart textiles. Enhancing textiles can be viewed through different lens.  Enhancements can be aesthetic, or performance boosting, or durability boosting. 
Read more