Module 1 · Foundations
Module 1 worked feedback
Use this after trying the practice. Compare your reasoning as well as the final value.
Boundary behavior
At 85, review is requested. At exactly 80, review is not requested because 80 > 80 is false. At 81, 81 > 80 is true. If you expected review at 80, you may have interpreted the requirement as “at least 80”; that would require >=, a different rule. Test the requirement rather than silently changing the operator.
Example explanation: “This program follows a threshold chosen by a person. It never fits parameters from observations, so its successful output does not demonstrate ML training.” AI is the broader field; ML learns from data; deep learning uses multilayer neural networks; generative AI produces content. Output fluency alone does not establish factual correctness.
Example problem brief
An IT operator needs to review storage before it becomes unavailable. The input is fictional percentage usage, and the output is a review message. An explicit threshold is an understandable starting point for this demonstration. A missing or invalid measurement could give misleading evidence, so the function rejects values outside 0–100 and nonfinite values. A real operational policy would require further context and testing; this teaching rule does not establish a safe production threshold.
A workload predictor could help when anticipating future demand matters and suitable historical data exist. Its evaluation must compare predictions with held-out outcomes. A generated response could help draft language, but cannot replace the measurement or supply trustworthy facts without checks.
Saved work and interfaces
The command's --output argument names the text file to create. Reusing the path replaces that generated file. A notebook stores cells and optionally outputs; its kernel holds the live variables. Restarting the kernel and running all cells exposes reliance on an old variable. Passing a Python-cell check does not establish that a particular notebook interface is accessible.
Example discussion response
A simple storage alert has an explicit, inspectable condition. An LLM adds expense and uncertainty without a clear benefit for this comparison. For drafting a support reply, language generation may help if approved facts are provided and a person checks the draft. I would compare errors, operational effort, and the stakeholder's actual need before replacing an existing method.
If an assistant called the threshold a “trained model,” reject that claim: there is no training step. Your verification is stronger when it cites the function and the two boundary runs.
Return to Module 1, or continue to Module 2.
Construction check and misconception feedback
Question: Why is changing 80 to 90 not training?
No parameter was estimated from examples; a person edited the decision rule.
Judge your answer with the study workflow criteria. If you cannot reproduce the central operation, return to the worked step and try changed inputs before reading the answer again.
Continue the learning route
For feedback on your own functions, use construction solutions. The results here describe the separate complete reference.