Explain key ML/stats concepts
Company: C3 AI
Role: Data Scientist
Category: Machine Learning
Difficulty: Medium
Interview Round: Onsite
You are taking an ML/Stats screening with conceptual multiple-choice questions. Answer the following:
1. **CNN vs. RNN**
- What kinds of input structure does each model assume?
- Give 1–2 examples of tasks where CNNs are typically preferred and tasks where RNNs (or sequence models) are preferred.
2. **Gini impurity (decision trees)**
- Define Gini impurity for a node with class probabilities \(p_1,\dots,p_K\).
- Compute the Gini impurity for a binary node with \(p=0.8\) and \(1-p=0.2\).
3. **Entropy vs. Gini for split criteria**
- Define entropy for a node and compare it to Gini impurity.
- Explain how they differ in sensitivity and whether they usually produce meaningfully different trees in practice.
4. **Multicollinearity (linear/logistic regression)**
- What is multicollinearity and why is it a problem?
- Name at least two ways to detect it and two ways to mitigate it.
5. **Pearson correlation**
- Define Pearson correlation and list key assumptions/limitations.
- Explain why correlation \(\neq\) causation and give one example of confounding.
Quick Answer: This question evaluates a candidate's understanding of core machine learning and statistical concepts, covering neural network input structures and model selection (CNN vs RNN), impurity measures and split criteria (Gini and entropy), multicollinearity diagnostics and mitigation in regression, and Pearson correlation with its assumptions and limits. It is commonly asked in Data Science and Machine Learning interviews to probe model-selection intuition, diagnostic and interpretive skills, and statistical reasoning, and it primarily targets conceptual understanding with some simple quantitative application.