- Using deep reinforcement learning to solve the multiple sequence alignment problem
- Joachim Vanneste
- 2460800v
- Kevin Bryson
- 2 hours Research into MSA, progressive alignment and optimisation techniques
- 0.5 hours Set-up git repository
- 3 hours Research into iterative alignment and probabilistic methods
- 1 hour Plan for project
- 2 hours Plan for meeting
- 1 hour Meeting
- 4 hours Reseach for literature review
- 5 hours Reseach for literature review
- 5 hours Start writing literature review
- 2 hours Start of alignment agent
- 5 hours DQN testing with connect 4
- 3 hours CNN to extract features
- 3 hours DQN finished and training loop started
- 1 hour Training loop finished
- 2 hours Meeting preparation
- 1 hour Meeting
- 3 hours Literature review
- 1 hour Meeting write-up
- 8 hours Work on MSA RL agent - gradient updates
- 3 hours Finish RL agent - slighly changed DQN model
- 3 hours Start RL agent training
- 5 hours Look into saving weights for transfer learning
- 2 hours First metrics from trained agent on 5x5s
- 1 hour RL agent hyperparameter optimisation
- 3 hours Look into MARL, cooperative AI
- 3 hours More research into MARL
- 2 hours Work on literature review
- 5 hours Work on literature review MSA section
- 3 hours Literature review
- 1 hour Meeting write-up
- 3 hours Collect more data
- 3 hours Test with larger sequences
- 5 hours Work on literature review DQN section
- 2 hours Work on literature review
- 2 hours Meeting
- 1 hour Meeting write-up
- 5 hours Look into creating a new layer
- 2 hours Start work on new layer - created pairwise kernel
- 3 hours Work on literature review
- 2 hours Reseach into real data sets to use
- 2 hours New network created with new layer
- 1 hour Think about whether to use FC layers
- 4 hours Work on interm report
- 2 hours Work on literature review
- 2 hours Meeting
- 1 hour Meeting write-up
- 2 hours Work on interm report
- 2 hours Reseach into RL for MSA for report
- 6 hours Work on interm report
- 2 hours Collect histograms
- 2 hours Running old model
- 4 hours Work on interm report
- 2 hours Work on literature review
- 2 hours Meeting
- 1 hour Meeting write-up
- 2 hours Complexity analysis of my method
- 2 hours Complexity analysis of several MSA methods to compare
- 3 hours Planning MARL solution
- 4 hours Implementing start of MARL - black and white agents created
- 2 hours Implementing of MARL solution
- 1 hour Complexity analysis resutls plotted
- 2 hour Meeting and write up