Claudio Silva

Claudio Silva

Institute Professor, New York University

Lab 11: Deep Learning

Lab 11: Deep Learning

Slides

The slides I showed this week can be found here.

Miscellaneous

  • I’m writing feedback for each of your project proposals - intending to post this weekend
  • Homework 4 (OPTIONAL - extra credit) has been posted to Brightspace
    • Amount of extra credit TBD
    • Due April 19th

How to Write a Good Research Paper

  • The ability to clearly communicate your ideas to a specific audience is a necessary skill for any profession - not just academia!
  • We discussed the purpose of each section of a research paper, how to craft a good argument, and the expectations for your own research paper about your final project. See slides linked above for details

Visualizing Deep Learning

  • In lecture this week, you looked at TensorFlow Playground and CNN Explainer, two visual tools for exploring the inner workings of neural networks
  • We explored NeuroCartography, a tool for scalable automatic visual summarization of concepts in deep neural networks
    • The tool can be found here
    • The paper can be found here
  • There are many other tools like this, for many different types of models:
    • ActiVis, for visual exploration of industry-scale deep neural network models
    • Dodrio, for visual exploration of transformer models
    • GAN Lab, for visualizing and understanding the inner workings of general adversarial networks
    • Visual Auditor, for identifying and understanding biases in machine learning models
    • Bluff, for visualizing and understanding adversarial attacks on neural networks
    • Diffusion Explainer, for visualizing how stable diffusion turns text prompts into images

Further Reading