Skip to content

ML Fundamentals & Concepts

Core ML theory for SDE-ML, MLE, and AI Engineer interviews


Overview

This section covers the theoretical foundations of ML that top tech companies evaluate in concept-based interviews. You'll need to explain these concepts clearly and know when to apply them.


Document Structure

DocumentFocus
concepts-overviewHow to approach ML concept questions
linear-regressionOLS, gradient descent, assumptions
logistic-regressionBinary classification, MLE, decision boundary
decision-treesSplitting criteria, pruning, CART
svmMargins, kernels, soft margin
knnDistance metrics, k selection, curse of dimensionality
neural-networksBackprop, activations, architectures
clusteringK-Means, DBSCAN, hierarchical
model-evaluationMetrics, cross-validation, bias-variance

Interview FAQ Section

Comprehensive Q&A guides with visualizations for the most commonly asked ML interview questions.

TopicKey Questions
Interview FAQ IndexComplete FAQ navigation
Bias-Variance TradeoffError decomposition, learning curves
RegularizationL1 vs L2, dropout, early stopping
Gradient DescentSGD, Adam, momentum, saddle points
Cross-ValidationK-fold, data leakage, stratified CV
Ensemble MethodsBagging vs boosting, XGBoost
Hypothesis TestingP-values, A/B testing, Type I/II errors

Interview Question Types

TypeWhat They AskExample
Algorithm MechanicsHow does X work?"Explain how gradient descent works"
Trade-offsWhen would you use X vs Y?"Decision Tree vs Random Forest?"
AssumptionsWhat assumptions does X make?"Linear regression assumptions?"
Failure ModesWhen does X fail?"When does KNN perform poorly?"
HyperparametersHow do you tune X?"How do you choose k in K-Means?"

Study Order

Week 1: Supervised Learning Basics

  1. Linear Regression
  2. Logistic Regression
  3. Model Evaluation

Week 2: Tree-Based & Instance-Based 4. Decision Trees 5. KNN 6. SVM

Week 3: Deep Learning & Clustering 7. Neural Networks 8. Clustering (K-Means, DBSCAN)


Quick Reference: Algorithm Selection

Problem TypeFirst ChoiceWhen to Use
RegressionLinear RegressionLinear relationship, interpretability
Binary ClassificationLogistic RegressionInterpretable probabilities
Multi-classRandom ForestNon-linear, robust
High-dimensionalSVM with RBFn_features > n_samples
ClusteringK-MeansSpherical clusters, known k
Density-based clusteringDBSCANArbitrary shapes, outliers