Link to paper
The full paper is available here.
You can also find the paper on PapersWithCode here.
Abstract
- Biological and artificial neural networks can make reliable distinctions between similar inputs.
- The immune system is similar in that it can make reliable distinctions and is partly learned.
- Self and nonself peptides have nearly identical distributions but are strongly inhomogeneous.
- The immune system targets the spaces in between the samples, which would be overfitting in conventional learning problems.
- The structure of the underlying distribution in sequence space predicts responses to cancer neoantigens.