Verified Deep Learning

A book by Aws Albarghouthi

Deep learning has transformed the way we think of software and what it can do. But deep neural networks are fragile and their behaviors are often surprising. In many settings, we need to provide formal guarantees on the safety, security, correctness, or robustness of neural networks. This in-progress book covers foundational ideas from formal verification and their application to reasoning about deep learning.

You can get all available chapters as one pdf or access individual chapters below. Keep in mind that the book is constantly evolving.

About this book

I Neural networks and correctness

• A new beginning
• Semantics of neural networks
• Correctness properties

II Constraint-based verification

• Decidable theories of first-order logic
• Encoding neural networks logically
• Decision procedures
• Specialized decision procedures

III Abstraction-based verification

• Abstraction for neural networks
• Relational abstract domains
• Training with abstract values

IV Verification and reinforcement learning

• Neural networks as policies
• Verifying RL policies
• Efficient neural policy verification
• Enforcing properties in RL

---

For comments, contact the author.
Please use the following to cite this book.

  
  @book{albarghouthi-book,
    title = {Verified Deep Learning},
    author = {Aws Albarghouthi},
    publisher = {verifieddeeplearning.com},
    note = {\url{http://verifieddeeplearning.com}},
    year = {2020}
  }
  

Last updated