From phd-skills
Designs ML experiments: ablation studies, baseline comparisons, experiment matrices; estimates GPU/API costs; generates config stubs, execution scripts, and analysis plans.
npx claudepluginhub fcakyon/phd-skills --plugin phd-skillsThis skill uses the workspace's default tool permissions.
You are helping a researcher design rigorous experiments. Follow this methodology systematically.
Provides Python code patterns for reproducible experiments: random seeds, environment logging, train/test splits, cross-validation, A/B testing, and power analysis. For ML/statistical designs.
Generates factorial, response surface, and Taguchi experimental designs to optimize multi-factor systems with minimal runs. Useful for screening variables, discovering interactions, A/B/n testing, parameter tuning.
Applies Formal Design of Experiments (DOE) methodology with factorial designs, blocking, randomization, and optimal strategies to maximize information while minimizing resources. Grounds responses in domain-specific reference files.
Share bugs, ideas, or general feedback.
You are helping a researcher design rigorous experiments. Follow this methodology systematically.
Before designing any experiment:
Every ablation study must change exactly ONE variable at a time. For each factor:
Template for each ablation row:
| Run ID | Factor | Value | Fixed Config | Expected Outcome |
|--------|--------|-------|-------------|-----------------|
For multi-factor studies, use a structured matrix:
Always calculate total runs before committing:
Total runs = product of all factor levels
GPU hours = total runs × hours_per_run
For each experiment plan, estimate:
Flag if total cost exceeds reasonable bounds and suggest prioritization.
Generate configuration stubs that match the user's existing config format. Read existing configs first to match:
Create a concrete execution plan:
Before running, define how results will be analyzed:
Before finalizing the experiment plan:
Always produce: