Week 3

12/10/2023

Work done

• Literature review work
• Greedy epsilon algorithm implemented for choosing actions
• Scoring (SP score) and action functions implemented

Topics for meeting

• How is deep learning RL loss calculated and back-propagated (still not working properly)
• Huber loss ?

Meeting Summary

• Discussion on catalytic sites and how only ~20% of a sequence needs to stay the same
• Scoring networks - do they take dependence into account
• Looked into substitution model of evolution to generate better sequences

Things to do

• Finish RL agent
• Look into models of DNA evolution for better sequence generation
• Finish MSA literature review section