🔬 Applied Scientist & ML Specialist938 Problems · 127 Topics🌐 Linked to Tech Vault

TalentMe Science Depth: Universal Applied Science & ML Depth Problem Bank

938 end-to-end science depth, system design, transition bridge & BQ problems covering derivations, production trade-offs, pitfalls, and follow-ups.

🎯

Science Depth Mastery Tracker

0% Mastered

Locally synced in your browser. Track, review, and master all 938 scientific & design topics.

Mastered: 0
Reviewing: 0
Unstarted: 119
🚀 Start Next Unpracticed →
🌐
Linked to TalentMe Tech Vault & MindmapsM2: Classical Machine Learning
All 938 questions are directly linked to TalentMe's Tech Vault guides, atomic cards, and interactive mindmaps.
🧭Classical ML Mindmap & Cards→📖Decision Trees, GBDT & XGBoost→🎯MLE Cheatsheet→
Found 24 in-depth questions
M2-049Classical Machine LearningK-Nearest Neighbors & Metric Learning
Medium

Explain Why KNN Strictly Requires Feature Scaling and Contrast Common Distance Metrics

⚡ One-Sentence Answer

KNN relies on geometric distance metrics; unscaled features with large numerical magnitudes dominate distance calculations completely, reducing smaller-scale features to zero influence.

Math & Tradeoffs→
M2-050Classical Machine LearningK-Nearest Neighbors & Metric Learning
Medium

Detail How K-Nearest Neighbors Extends to Continuous Regression and Unsupervised Anomaly Detection

⚡ One-Sentence Answer

In regression, KNN averages the continuous target values of the $K$ closest neighbors; in anomaly detection, instances with unusually large average distance to their $K$ nearest neighbors are flagged as outliers (Local Outlier Factor).

Math & Tradeoffs→
M2-051Classical Machine LearningK-Nearest Neighbors & Metric Learning
Hard

Explain the Curse of Dimensionality and Its Pathological Consequences on Distance-Based Models

⚡ One-Sentence Answer

As dimensionality $d$ increases, the volume of space grows exponentially ($V \propto r^d$); all pairwise distances converge to the same value (distance concentration), and all data points become isolated outliers on the outer shell of the hypersphere.

Math & Tradeoffs→
M2-052Classical Machine LearningClustering Algorithms
Easy

Formulate the K-Means Algorithm, Coordinate Descent Objective, and Convergence Guarantees

⚡ One-Sentence Answer

K-Means minimizes Within-Cluster Sum of Squares (WCSS) via alternating coordinate descent between cluster assignment and centroid updating, guaranteeing convergence to a local minimum in a finite number of iterations.

Math & Tradeoffs→
M2-053Classical Machine LearningClustering Algorithms
Easy

Detail Methodologies for Selecting the Optimal Number of Clusters K in Unsupervised Learning

⚡ One-Sentence Answer

The optimal cluster count $K$ is evaluated via the Elbow Method (inflection point of inertia), Silhouette Analysis (measuring cohesion vs separation in $[-1, 1]$), the Gap Statistic (comparing against uniform null reference), and Bayesian Information Criterion (BIC in GMMs).

Math & Tradeoffs→
M2-054Classical Machine LearningClustering Algorithms
Medium

Compare K-Means, Gaussian Mixture Models (GMM), and DBSCAN Across Assumptions, Geometry, and Complexity

⚡ One-Sentence Answer

K-Means enforces hard spherical partitions; GMM provides probabilistic soft clustering with ellipsoidal covariance; DBSCAN discovers arbitrarily shaped density clusters without pre-specifying $K$ and naturally isolates noise.

Math & Tradeoffs→
M2-055Classical Machine LearningClustering Algorithms
Medium

Derive the Expectation-Maximization (EM) Algorithm and Prove Its Monotonic Likelihood Ascent via Jensen's Inequality

⚡ One-Sentence Answer

The EM algorithm maximizes incomplete log-likelihood with latent variables by iteratively constructing a tight Evidence Lower Bound (ELBO) in the E-step via $q(z) = p(z\mid x, \theta)$ and maximizing that lower bound in the M-step.

Math & Tradeoffs→
M2-056Classical Machine LearningClustering Algorithms
Hard

Detail Unsupervised Clustering Evaluation: Internal Metrics vs. External Ground-Truth Metrics

⚡ One-Sentence Answer

Internal metrics evaluate cluster geometry without ground-truth labels (Silhouette Score, Davies-Bouldin Index, Calinski-Harabasz Index); External metrics evaluate alignment with ground-truth classes (Adjusted Rand Index, Normalized Mutual Information).

Math & Tradeoffs→
M2-057Classical Machine LearningDimensionality Reduction
Easy

Formulate Principal Component Analysis (PCA): Maximum Variance and Minimum Reconstruction Error Perspectives

⚡ One-Sentence Answer

PCA finds an orthogonal linear transformation that projects data onto directions of maximum variance, which is mathematically identical to finding the linear subspace that minimizes mean squared reconstruction error.

Math & Tradeoffs→
M2-058Classical Machine LearningDimensionality Reduction
Easy

Explain the Exact Mathematical Equivalence Between PCA and Singular Value Decomposition (SVD)

⚡ One-Sentence Answer

For a zero-centered data matrix $X = U\Sigma V^T$, the right singular vectors $V$ are the principal component loading directions, the singular values relate to eigenvalues via $\lambda_i = \frac{\sigma_i^2}{N-1}$, and $U\Sigma$ provides the projected principal coordinates.

Math & Tradeoffs→
M2-059Classical Machine LearningDimensionality Reduction
Medium

Enumerate the Limitations of PCA and Scenarios Where Linear PCA Fails

⚡ One-Sentence Answer

PCA assumes linear relationships, is sensitive to unscaled magnitudes and extreme outliers, maximizes total variance without considering class labels (unsupervised), and fails to capture non-linear manifold structures.

Math & Tradeoffs→
M2-060Classical Machine LearningDimensionality Reduction
Medium

Compare t-SNE and UMAP: Manifold Assumptions, Global vs. Local Structure, and Computational Scaling

⚡ One-Sentence Answer

t-SNE models local neighborhood similarities via Gaussian-Student-t divergences ($O(N^2)$); UMAP models Riemannian manifold topology via fuzzy simplicial sets ($O(N\log N)$), running $10\times$ faster while preserving both global and local structure.

Math & Tradeoffs→
M2-061Classical Machine LearningDimensionality Reduction
Hard

How to Choose Target Dimensions in Dimensionality Reduction? Variance Ratio Method and Limitations

⚡ One-Sentence Answer

Select dimensions that explain 90–95% of cumulative variance; however, because this heuristic is downstream-agnostic, downstream validation performance should guide the final choice.

Math & Tradeoffs→
M2-062Classical Machine LearningFeature Engineering
Easy

Common Categorical Feature Encoding Methods and Applicable Scenarios

⚡ One-Sentence Answer

One-hot for low cardinality, target encoding for high cardinality, frequency encoding for prevalence, hashing for unbounded streams, and embeddings for deep architectures.

Math & Tradeoffs→
M2-063Classical Machine LearningFeature Engineering
Easy

Why Bin Numerical Features? Common Binning Strategies and Use Cases

⚡ One-Sentence Answer

Binning captures non-linearities, provides robustness against outliers, and enhances business interpretability via equal-width, equal-frequency, or supervised tree/chi-square approaches.

Math & Tradeoffs→
M2-064Classical Machine LearningFeature Engineering
Medium

Feature Interactions and Polynomial Features: Concepts and Scenarios

⚡ One-Sentence Answer

Linear models cannot capture feature cross-effects without explicit interaction terms ($x_i \cdot x_j$), whereas tree-based models and neural networks can learn them automatically.

Math & Tradeoffs→
M2-065Classical Machine LearningFeature Engineering
Medium

Key Points in Time Feature Engineering and Preventing Temporal Leakage

⚡ One-Sentence Answer

Construct lag features, rolling window statistics, cyclical encodings, and elapsed durations while strictly ensuring right-aligned windows to avoid future data leakage.

Math & Tradeoffs→
M2-066Classical Machine LearningFeature Engineering
Hard

What is the Hashing Trick (Feature Hashing)? Pros and Cons

⚡ One-Sentence Answer

It projects arbitrary strings or categories into a fixed-size index space using hash functions, bounding memory and handling streaming vocabularies at the cost of irreversibility and collisions.

Math & Tradeoffs→
M2-067Classical Machine LearningFeature Selection
Easy

Comparing Filter, Wrapper, and Embedded Feature Selection Methods

⚡ One-Sentence Answer

Filter methods use statistical metrics independently of models; wrapper methods optimize feature subsets iteratively via model performance; embedded methods integrate selection into training.

Math & Tradeoffs→
M2-068Classical Machine LearningFeature Selection
Medium

Why Can Removing 'Useless Features' Sometimes Degrade Model Performance?

⚡ One-Sentence Answer

Weak or individually uninformative features can form strong non-linear interactions, act as suppressors of noise, or provide implicit regularization.

Math & Tradeoffs→
M2-069Classical Machine LearningFeature Selection
Medium

Feature Selection Strategies Under Severe Multicollinearity

⚡ One-Sentence Answer

Diagnose with VIF, correlation heatmaps, or condition indices; resolve via Ridge/ElasticNet regularization, PCA orthogonalization, or hierarchical clustering pruning.

Math & Tradeoffs→
M2-070Classical Machine LearningFeature Selection
Hard

How to Evaluate Feature Selection Effectiveness? Key Considerations and Validation

⚡ One-Sentence Answer

Compare downstream validation metrics, inference latency, and model stability with and without the candidate features, strictly performing selection inside the cross-validation loop.

Math & Tradeoffs→
M2-071Classical Machine LearningClass Imbalance Learning
Easy

Major Approaches for Handling Class Imbalance

⚡ One-Sentence Answer

Address imbalance at the data level (re-sampling), algorithmic level (cost-sensitive loss, Focal Loss), and post-processing level (threshold tuning).

Math & Tradeoffs→
M2-072Classical Machine LearningClass Imbalance Learning
Easy

SMOTE Algorithm: Core Mechanism, Theoretical Risks, and Limitations

⚡ One-Sentence Answer

SMOTE synthesizes minority samples by linear interpolation along the line segments joining nearest neighbors; risks include blurring decision boundaries and amplifying noise in high dimensions.

Math & Tradeoffs→