14.6.1: Review Questions
- Page ID
- 58359
\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)
\( \newcommand{\dsum}{\displaystyle\sum\limits} \)
\( \newcommand{\dint}{\displaystyle\int\limits} \)
\( \newcommand{\dlim}{\displaystyle\lim\limits} \)
\( \newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\)
( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\)
\( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)
\( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\)
\( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)
\( \newcommand{\Span}{\mathrm{span}}\)
\( \newcommand{\id}{\mathrm{id}}\)
\( \newcommand{\Span}{\mathrm{span}}\)
\( \newcommand{\kernel}{\mathrm{null}\,}\)
\( \newcommand{\range}{\mathrm{range}\,}\)
\( \newcommand{\RealPart}{\mathrm{Re}}\)
\( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)
\( \newcommand{\Argument}{\mathrm{Arg}}\)
\( \newcommand{\norm}[1]{\| #1 \|}\)
\( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)
\( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\AA}{\unicode[.8,0]{x212B}}\)
\( \newcommand{\vectorA}[1]{\vec{#1}} % arrow\)
\( \newcommand{\vectorAt}[1]{\vec{\text{#1}}} % arrow\)
\( \newcommand{\vectorB}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\( \newcommand{\vectorC}[1]{\textbf{#1}} \)
\( \newcommand{\vectorD}[1]{\overrightarrow{#1}} \)
\( \newcommand{\vectorDt}[1]{\overrightarrow{\text{#1}}} \)
\( \newcommand{\vectE}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash{\mathbf {#1}}}} \)
\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\(\newcommand{\longvect}{\overrightarrow}\)
\( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)
\(\newcommand{\ket}[1]{\left| #1 \right>}\)
\(\newcommand{\bra}[1]{\left< #1 \right|}\)
\(\newcommand{\braket}[2]{\left< #1 \vphantom{#2} \right| \left. #2 \vphantom{#1} \right>}\)
\(\newcommand{\braopket}[3]{\left< #1 \vphantom{#2}\vphantom{#3} \right| #2 \vphantom{#1}\vphantom{#3} \left| #3 \vphantom{#1}\vphantom{#2} \right>}\)
\(\newcommand{\qmvec}[1]{\mathbf{\vec{#1}}}\)
\(\newcommand{\op}[1]{\hat{\mathbf{#1}}}\)
\(\newcommand{\expect}[1]{\langle #1 \rangle}\)
\(\newcommand{\dfn}[1]{\emph{\textbf{#1}}}\)
Review Questions
Multiple Choice
-
XGBoost is an example of which ML approach?
-
a) Bagging b) Gradient boosting c) Deep learning d) K-means clustering
-
-
Which Explainable AI method uses game-theoretic Shapley values to attribute each feature's contribution?
-
a) LIME b) SHAP c) PCA d) AutoML
-
-
Federated learning was developed primarily to address concerns about:
-
a) Model accuracy b) Training speed c) Data privacy — training models without centralizing raw data d) Visualization quality
-
Short Answer
-
Compare bagging (e.g., Random Forest) and boosting (e.g., XGBoost). How does each combine multiple models? What are their respective strengths?
-
What is AutoML and how does it democratize machine learning? Describe the workflow: what does the user provide, and what does AutoML automate?
-
Explain SHAP and LIME for model interpretability. Why is explainability important in regulated industries like finance and healthcare?
-
Describe the data mesh concept. How does it differ from a traditional centralized data warehouse approach? What are its four principles?
True/False
-
True or False: Deep learning models always outperform simpler models like logistic regression regardless of dataset size.
-
True or False: Apple uses differential privacy to learn aggregate usage trends from millions of devices without revealing individual user behavior.
-
True or False: AutoML eliminates the need for any human oversight in the ML process.
Answer Key:
1-b, 2-b, 3-c, 4. Bagging: trains models in parallel on bootstrapped samples, averages predictions — reduces variance (Random Forest adds random feature selection). Boosting: trains models sequentially, each focusing on errors of the previous — reduces bias. Bagging: more robust to overfitting, parallelizable. Boosting: often higher accuracy but can overfit if not tuned. 5. AutoML automates: data preprocessing, feature engineering, algorithm selection, hyperparameter tuning, model evaluation. User provides: dataset and target column (and optionally time/resource constraints). AutoML outputs: ranked leaderboard of models with performance metrics. Democratizes ML by letting non-experts build models. 6. SHAP: assigns each feature a Shapley value showing its marginal contribution to each prediction — global and local explanations. LIME: creates a simple interpretable model that approximates the complex model around a specific prediction. Important in regulated industries because decisions affecting people (loan approvals, medical diagnoses) must be explainable for compliance, trust, and fairness audits. 7. Data mesh: decentralized architecture where domain teams own their data products. Differs from centralized warehouse where one team controls all data. Four principles: (1) domain-oriented decentralized ownership, (2) data as a product, (3) self-serve data platform, (4) federated computational governance. 8. False, 9. True, 10. False.