Regression Problems — ML Study Repository

📖 Techout (L6 AIML Interview Prep)

→ REGRESSION_TECHOUT.md — Master index (start here)

File	Sections	Focus
REGRESSION_PART1_FOUNDATIONS.md	§1–4	Big Picture, Simple Linear, Cost Function, Gradient Descent
REGRESSION_PART2_CLASSICAL_ML.md	§5–10	Multiple Regression, Feature Eng, Polynomial, Regularization, Metrics, Diagnostics
REGRESSION_PART3_MODERN_ML.md	§11–14	DeepAR, LLM Era, E2E AdTech Pipeline, L6 Cheat Sheet

All examples use Bing Ads / ad-click prediction as the running AdTech context.
Python stack: scikit-learn · PyTorch · statsmodels · LightGBM · HuggingFace · Chronos

→ CLAUDE_INSTRUCTIONS.md — System prompt for Claude to extend this document
   Enforces: LaTeX math · Mermaid diagrams · Python code · AdTech examples · L6 interview callouts

---

🏃 Running the Examples

# Install Python ML stack
pip install numpy pandas scikit-learn torch statsmodels lightgbm \
            transformers "chronos-forecasting[training]" neuralforecast \
            matplotlib seaborn

# Plot ad click forecasting
python plot-ad-clicks.py

Java Examples (Gradle — original implementation, kept as reference)

Class	What it demonstrates
AdClickLinearRegression.java	Simple regression on ad impressions → clicks
CovidCaseLinearRegression.java	OLS with Apache Commons Math SimpleRegression
CovidCaseLinearRegressionSpark.java	Distributed regression via Apache Spark MLlib
AirPassengersDeepAR.java	Probabilistic time-series forecasting with DJL DeepAR

All techout code examples use Python (scikit-learn, PyTorch, LightGBM, HuggingFace, Chronos) — the industry-standard stack for AIML engineering interviews.

---

Resources

• https://www.coursera.org/learn/ml-foundations
• https://www.coursera.org/learn/ml-regression
• https://learn.microsoft.com/en-us/azure/machine-learning/concept-automated-ml