🔬 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: 938
🚀 Start Next Unpracticed →
🌐
Linked to TalentMe Tech Vault & MindmapsGlobal System
All 938 questions are directly linked to TalentMe's Tech Vault guides, atomic cards, and interactive mindmaps.
🧭Math Foundations Mindmap & Cards→📖Bayesian Inference, Entropy & KL Divergence→🎯DS Track→
Found 24 in-depth questions
M2-080Classical Machine LearningMissing Values & Data Leakage
Hard

Why Global Normalization Before Splitting Causes Data Leakage and How to Fix It

⚡ One-Sentence Answer

Global standardization uses test set mean and variance to scale training data, artificially shrinking generalization error; fix by fitting scalers strictly on training folds.

Math & Tradeoffs→
M2-081Classical Machine LearningModel Calibration
Easy

Defining Model Calibration: Why High Accuracy Does Not Imply Good Calibration

⚡ One-Sentence Answer

A model is calibrated when its predicted probability matches empirical ground truth frequency; a highly accurate model can still be severely overconfident.

Math & Tradeoffs→
M2-082Classical Machine LearningModel Calibration
Medium

Comparing Platt Scaling and Isotonic Regression for Probability Calibration

⚡ One-Sentence Answer

Platt scaling fits a parametric sigmoid to model logits; Isotonic regression fits a non-parametric piecewise-constant isotonic step function.

Math & Tradeoffs→
M2-083Classical Machine LearningModel Calibration
Medium

Expected Calibration Error (ECE): Mathematical Formulation and Practical Limitations

⚡ One-Sentence Answer

ECE partitions predictions into probability bins and computes the weighted average gap between confidence and accuracy; limitations include sensitivity to bin count and edge bias.

Math & Tradeoffs→
M2-084Classical Machine LearningModel Calibration
Medium

Temperature Scaling for Calibrating Deep Neural Networks

⚡ One-Sentence Answer

A single scalar parameter $T > 0$ divides pre-softmax logits, softening or sharpening probabilities without changing the argmax prediction order or classification accuracy.

Math & Tradeoffs→
M2-085Classical Machine LearningModel Calibration
Hard

Scenarios Where Calibration is Critical: Core Use Cases and Concrete Examples

⚡ One-Sentence Answer

Calibration is vital whenever decisions depend on expected value calculations, safety cost thresholds, or multi-model probability fusion.

Math & Tradeoffs→
M2-086Classical Machine LearningEvaluation Metrics & Hyperparameter Tuning
Easy

Common Evaluation Metrics for Classification, Regression, and Ranking

⚡ One-Sentence Answer

Classification uses Accuracy, F1, PR-AUC, and ROC-AUC; regression uses RMSE, MAE, and MAPE; ranking uses NDCG, MRR, and MAP.

Math & Tradeoffs→
M2-087Classical Machine LearningEvaluation Metrics & Hyperparameter Tuning
Medium

Comparing Grid Search, Random Search, and Bayesian Optimization

⚡ One-Sentence Answer

Grid search suffers from exponential dimensional explosion; Random search explores continuous spaces efficiently; Bayesian optimization constructs probabilistic surrogates to balance exploration and exploitation.

Math & Tradeoffs→
M2-088Classical Machine LearningEvaluation Metrics & Hyperparameter Tuning
Medium

Hyperparameter Importance Hierarchy and Efficient Tuning Strategies

⚡ One-Sentence Answer

Prioritize learning rate and regularization first, architecture second, and secondary parameters last; employ coarse-to-fine search and early-pruning strategies.

Math & Tradeoffs→
M2-089Classical Machine LearningEvaluation Metrics & Hyperparameter Tuning
Hard

Handling Randomness and Seed Variance in Hyperparameter Tuning

⚡ One-Sentence Answer

Evaluate configurations across multiple random seeds, report mean and confidence intervals, and apply paired statistical tests to distinguish genuine improvements from seed noise.

Math & Tradeoffs→
M2-090Classical Machine LearningEvaluation Metrics & Hyperparameter Tuning
Hard

Why Offline and Online Evaluations Diverge: Causes and Mitigation Strategies

⚡ One-Sentence Answer

Divergence stems from distribution drift, proxy metric misalignment, feedback loops, and latency; mitigate via off-policy evaluation, correlation tracking, and canary A/B tests.

Math & Tradeoffs→
M2-091Classical Machine LearningFeature Selection
Hard

Evaluating Feature Importance in Neural Networks vs Tree Models

⚡ One-Sentence Answer

Neural networks utilize gradient-based attribution (Integrated Gradients, Saliency) and perturbation methods, whereas tree models rely on impurity gain, split count, and TreeSHAP.

Math & Tradeoffs→
M2-092Classical Machine LearningLinear Regression
Medium

Weighted Least Squares (WLS) and Handling Heteroscedasticity

⚡ One-Sentence Answer

WLS weights each residual inversely by the error variance ($w_i = 1/\sigma_i^2$), transforming heteroscedastic data into homoscedastic space to restore BLUE optimality.

Math & Tradeoffs→
M2-093Classical Machine LearningLinear Regression
Medium

Interaction and Polynomial Terms in Regression: Formulations and Regularization

⚡ One-Sentence Answer

Incorporate non-linear effects and conditional dynamics via product terms; mitigate overfitting using the hierarchical principle, centering, and L1/L2 regularization.

Math & Tradeoffs→
M2-094Classical Machine LearningLogistic Regression & GLM
Medium

Multinomial Logistic Regression (Softmax Regression) and Parameter Identifiability

⚡ One-Sentence Answer

Softmax regression generalizes logistic regression to $K$ classes; overparameterization requires fixing one class vector to zero (reference class) to ensure parameter identifiability.

Math & Tradeoffs→
M2-095Classical Machine LearningLogistic Regression & GLM
Hard

Relationship Between Logistic Regression and the Maximum Entropy Model

⚡ One-Sentence Answer

They are mathematically dual: maximizing likelihood in logistic regression is equivalent to maximizing conditional entropy subject to empirical moment constraints.

Math & Tradeoffs→
M2-096Classical Machine LearningRegularization (L1 / L2)
Medium

Bayesian Interpretation of L1 and L2 Regularization

⚡ One-Sentence Answer

L2 regularization corresponds to a Gaussian prior (shrinkage); L1 regularization corresponds to an independent Laplace prior (sparsity via its non-differentiable peak at zero).

Math & Tradeoffs→
M2-097Classical Machine LearningRegularization (L1 / L2)
Hard

Why Weight Decay Must Be Decoupled in Adam (AdamW)

⚡ One-Sentence Answer

L2 regularization couples weight decay with gradient second moments, diluting regularization for large-gradient weights; AdamW restores true weight decay by decoupling it from gradient scaling.

Math & Tradeoffs→
M2-098Classical Machine LearningRegularization (L1 / L2)
Hard

Spectral Norm Regularization and Lipschitz Continuity Constraints

⚡ One-Sentence Answer

Bounding the spectral norm (maximum singular value) of each layer's weight matrix bounds the network's global Lipschitz constant, stabilizing GAN training and ensuring robustness.

Math & Tradeoffs→
M2-099Classical Machine LearningRegularization (L1 / L2)
Hard

Stochastic Depth (DropPath) vs Standard Dropout: Mechanisms and Differences

⚡ One-Sentence Answer

Dropout randomly zeroes out individual activations, whereas Stochastic Depth drops entire residual blocks during training, shortening effective depth and perfectly preserving BatchNorm statistics.

Math & Tradeoffs→
M2-100Classical Machine LearningBias-Variance Tradeoff & Model Selection
Medium

AIC, BIC, and Adjusted R-squared: Penalized Metrics vs Cross-Validation

⚡ One-Sentence Answer

AIC and BIC penalize model complexity analytically based on likelihood; AIC asymptotically equates to Leave-One-Out CV, while BIC enforces selection consistency.

Math & Tradeoffs→
M2-101Classical Machine LearningBias-Variance Tradeoff & Model Selection
Medium

Learning Curves vs Validation Curves: Formulations and Diagnostic Workflows

⚡ One-Sentence Answer

Learning curves plot error against training set size to diagnose data sufficiency; validation curves plot error against a hyperparameter to diagnose bias-variance trade-offs.

Math & Tradeoffs→
M2-102Classical Machine LearningBias-Variance Tradeoff & Model Selection
Medium

How Model Ensembling Impacts Bias and Variance

⚡ One-Sentence Answer

Bagging reduces variance through parallel averaging; Boosting reduces bias through sequential residual fitting; Stacking optimizes combinations of diverse base hypotheses.

Math & Tradeoffs→
M2-103Classical Machine LearningBias-Variance Tradeoff & Model Selection
Medium

When Does Adding More Data Fail to Improve Model Performance?

⚡ One-Sentence Answer

Adding data fails when performance is bottlenecked by high model bias, irreducible Bayes noise, distribution shift, or systematic label corruption.

Math & Tradeoffs→