About the project
The project combines custom implementations with reference methods and explicitly defined experimental conditions. Small control tasks help expose implementation faults and limitations in the evaluation early. Repeated runs with different random initialisations and statistical analysis form part of the intended comparison. A successful technical trial is not presented as robust evidence of performance.
Focus areas
- Custom implementation of learning methods
- Comparison with simple baselines
- Documented experimental conditions and analyses
Our role and project context
Independent research and development project