Workflow

Think Before You Code

• What is the objective?
• Does the code have a logical basis for achieving the objective?
• Break the action down into smaller steps -- is each step feasible?
• Besides the benefits, what are the downsides?

Prefer Fewer Mistakes Over Speed

• When your understanding is unclear, ask before reaching a conclusion
• Estimate the effort required and check whether the schedule is feasible
• Simply estimating effort without simulating whether it can be achieved with current resources will cause stalling at some point
• Refactoring absolutely requires test scripts. Perform incremental refactoring and testing cycles
• Modifying code without understanding it leads to bugs
• Automate experiment management with tools like mlflow to reduce manual operations, which in turn reduces human error

For Smooth Remote Project Progress

• Create data generation scripts and figure generation scripts separately, and make figure generation runnable with a single command (You will forget the steps if you use Excel)
• Document your program explanations thoroughly. Use README and wiki for collaboration and your own future reference
• Distinguish between correlation and causation. Events driven by the same factor or intervening variables may be correlated (spurious correlation), but that does not mean they have a cause-and-effect relationship (causation).