D682: Artificial Intelligence Optimization for Computer Scientists

D682 is a later-stage artificial intelligence course in the Bachelor of Science in Computer Science at Western Governors University.

WGU describes it as guiding students through the implementation and optimization of AI solutions, using research to decide which approach suits a scenario and hands-on exercises to implement and rigorously test them.

It also covers creating assumptions and interpretations from data for predictive analytics, and adapting solutions to new applications. The prerequisite is Introduction to AI for Computer Scientists (D429). Students find the course demanding because success depends on both coding and clear written justification of what changed and why.

What D682 Covers

Key Concepts Explained

Evaluating Before Optimizing

You cannot say a model improved without a baseline and a metric. Choose measures that fit the problem, such as accuracy, precision and recall, or error for numerical predictions, and record the starting values.

Train, Validate, Test

Keep a held-out test set untouched until the end. Tuning against it leaks information and makes results look better than they are.

Assumptions in Prediction

Forecasts rely on assumptions, such as stable patterns over time or representative data. State them and discuss what would break them.

A Reliable Experiment Workflow

Optimization work is easier to document when you follow a repeatable workflow. The steps below match the competencies.

1. Define the task and metric. State what the model must do and how success is measured.
2. Prepare data. Clean it, split it into training, validation and test sets, and record how.
3. Build a baseline. A simple model gives you a point of comparison.
4. Tune one change at a time. Examples include features, parameters or model type.
5. Evaluate on held-out data. Report final numbers once, on the test set.
6. Interpret and adapt. Explain what the results mean, then describe what you would change for another application.

Record each experiment in a small table with the change, the metric and a comment. This makes your write-up straightforward and helps you answer questions about why you made a decision.

Be careful about claims. Report limits such as small data size, class imbalance or sensitivity to random seeds, and avoid saying a model is "best" unless you compared it fairly.

Typical Work and How to Approach It

Your task instructions state the exact deliverables. Work in this kind of course typically includes the following.

Where Students Get Stuck

• No baseline. Always record the starting performance.
• Changing many things at once. You cannot tell what helped.
• Data leakage. Keep test data separate.
• Thin write-up. Explain the reasoning, not only the numbers.

Study Tips for D682

• Revisit your D429 notes on terminology and algorithms.
• Keep an experiment log with date, change, metric and result.
• Plot results so improvements are visible.
• Run a simple version first, then refine.

Final Checklist Before You Submit

• A baseline and metric are recorded.
• Each change is documented separately with its result.
• Test data was kept apart from tuning.
• Assumptions and limits of any predictions are stated.
• The adaptation section explains what changes for a new use.
• Your code runs from a clean start and is commented.

Re-run your work from scratch once. If you cannot reproduce your numbers, neither can an evaluator.

How We Help with D682

Send the task instructions, rubric and your notes. A computing tutor can explain concepts such as metrics, validation and tuning, review your write-up for clarity, or walk through a custom solution on a similar problem. We do not write or submit your code or results for you.

GradeEssays is independent and not affiliated with WGU. This is a study and reference aid; submit only your own work under WGU's academic integrity policy. We do not log in to course portals or take assessments. Free revisions; full refund if late.