Pathway / Computing
Computing from procedure to learning systems
Make procedures explicit, reason about data and cost, then cross the boundary from software systems into evaluated machine learning.
- Route
- 5 connected concepts
- Access
- Open · account optional
- Evidence
- Local states, never a score
- Release state
- 3 field trials · 12 sketches overall
Next useful action
Procedures and state
Continue with the first unseen idea on this route.
- 01
Procedures and state
UnseenRead programs as precise transformations of state, beginning with Python.
- Record the relevant state before and after each assignment in a short Python procedure.
- Use a loop invariant to predict the result of an accumulation procedure.
- Distinguish rebinding a local name from mutating an object shared with the caller.
- 02
Data and invariants
UnseenChoose representations by the promises your program must preserve.
- Objectives and precise evidence coverage are pending schema-v2 conversion.
- 03
Algorithms and cost
UnseenCompare procedures by how their required work grows, not by one stopwatch reading.
- Objectives and precise evidence coverage are pending schema-v2 conversion.
- 04
Ownership and systems
UnseenUse Rust ownership to make resource lifetime and aliasing part of the program's structure.
- Objectives and precise evidence coverage are pending schema-v2 conversion.
- 05
Treat a trained model as one component in an evidence-producing system.
- Objectives and precise evidence coverage are pending schema-v2 conversion.