How michaelhush/m-loop works
M-LOOP automates online optimization of experiment or controller parameters. A user launches a configured run; the controller coordinates a learner and an experiment interface, returns measured costs to guide later parameter proposals, and can launch visualizations after optimization. The graph preserves the principal runtime stages and the documented visualization capability. Evidence is limited to the README, source index, and excerpts provided; unsampled subsystem details are not inferred.
Main entry points: interfaces.py Interface runner · launchers.py Run launcher · visualizations.py Result visualizations · controllers.py Controller
Components 12
Run entry
- Run launcher
mloop/launchers.py
Optimization loop
- Controller
mloop/controllers.py - Optimization learners
mloop/learners.py
Experiment interfaces
- Interface runner
mloop/interfaces.py - File interface
mloop/interfaces.py - Shell interface
mloop/interfaces.py
Models and results
- Neural network models
mloop/neuralnet.py - Test landscapes
mloop/testing.py - Utilities and archives
mloop/utilities.py - Result visualizations
mloop/visualizations.py
Other components
- User
- Experiment system
Connections 15
- User → Run launcher: launches
README.rst - Run launcher → Interface runner: creates
mloop/launchers.py - Run launcher → Controller: creates
mloop/launchers.py - Run launcher → Utilities and archives: loads config
mloop/launchers.py - Run launcher → Result visualizations: launches extras
mloop/launchers.py - Controller → Optimization learners: starts; exchanges proposals and costs
mloop/controllers.py - Controller → Interface runner: starts; exchanges parameters and costs
mloop/controllers.py - Interface runner → Experiment system: evaluates parameters
mloop/interfaces.py - File interface → Interface runner: extends
mloop/interfaces.py - Shell interface → Interface runner: extends
mloop/interfaces.py - Optimization learners → Neural network models: uses models
mloop/learners.py - Interface runner → Test landscapes: uses test landscape
mloop/interfaces.py - Result visualizations → Optimization learners: reads learner results
mloop/visualizations.py - Result visualizations → Controller: reads controller archive
mloop/visualizations.py - Result visualizations → Utilities and archives: loads archive
mloop/visualizations.py