Project Contacts

For further information about this project, please contact:

• Dr Nicholas Chancellor (nick.chancellor@newcastle.ac.uk)
• Dr Yonatan Yehezkeally (yonatan.yehezkeally@newcastle.ac.uk)

Project Description

This MSc Advanced Computer Science dissertation project investigates the feasibility of using neural autoencoders to learn constraint-satisfying run-length limited (RLL) codes for composite DNA data storage. The project evaluates pure neural and hybrid neural-symbolic architectures to minimize sequence violation rates and bit error rates.

Software or Compute Methods

Python, PyTorch, Anaconda, and custom Transformer/LSTM neural network architectures. Compute resources (specifically GPU acceleration) are required to train complex sequence-to-sequence models and evaluate stabilization methods across multiple randomized seeds.