第 14 / 37 页
非代码面试题
显示 20 / 738 道匹配题目
答题状态：未尝试未正确已正确
ID题目领域难度题型进度权限
4231Grouped Permutation RemedyIf several sector dummy variables move together and share the same economic information, what diagnostic is often better than permuting one dummy at a time?机器学习中等derivation未尝试面试订阅4232Time-Split Remedy for Leakage RiskA feature may be available only with reporting delay. What evaluation setup is more convincing than random train/test splitting?机器学习中等derivation未尝试面试订阅4233Retrain-After-Drop CheckWhy can 'drop feature X and retrain' tell a different story from permutation importance on the original fitted model?机器学习中等derivation未尝试面试订阅4234Conditional Importance RemedyIf a feature is highly correlated with others, what is the point of conditional importance rather than plain marginal permutation?机器学习中等derivation未尝试面试订阅4235Stability RemedyIf importance rankings swing wildly across folds, what is the right reaction?机器学习中等derivation未尝试面试订阅4236Importance Is Not CausalityWhy is it dangerous to treat feature importance as if it were a causal ranking?机器学习中等essay未尝试面试订阅4237Why Trees Overcredit Splittable FeaturesWhy do impurity-based importances tend to overcredit features with many possible split points?机器学习中等essay未尝试面试订阅4238Why Correlation Makes Rankings FragileWhy do strongly correlated features make importance rankings fragile?机器学习中等essay未尝试面试订阅4239Why You Need Multiple Importance ViewsWhy is it often wise to look at more than one feature-importance diagnostic?机器学习中等essay未尝试面试订阅4261Why Standardization Can Change PCA DramaticallyWhy is standardizing features often important before PCA when the raw variables use very different units?机器学习中等essay未尝试面试订阅4262Why Whitening Can Amplify NoiseWhy can aggressive whitening make downstream models numerically noisier?机器学习中等essay未尝试面试订阅4263Why PCA Helps Before Some ModelsWhy does flipping the sign of a principal-component loading vector not change the PCA solution in substance?机器学习中等essay未尝试面试订阅4264When PCA Can HurtWhen can PCA actually hurt a predictive pipeline?机器学习中等essay未尝试面试订阅4265A Fast Sanity Check for PCA QuestionsWhy might the first principal component be poor for predicting a label even though it explains the most variance?机器学习中等essay未尝试面试订阅4286Why Scaling Matters for k-MeansYou expect non-convex moon-shaped clusters plus some noise points. Would you try k-means or DBSCAN first, and why?机器学习中等essay未尝试面试订阅4287Why k-Means Struggles on Two MoonsWhy can k-means behave poorly when one feature has a much larger scale than the others?机器学习中等essay未尝试面试订阅4288Why Outliers Distort CentroidsWhy can silhouette score prefer a smaller k even when larger k lowers SSE?机器学习中等essay未尝试面试订阅4289Why Hierarchical Clustering Can Be UsefulWhy can clustering raw price levels of stocks be misleading compared with clustering normalized returns or features?机器学习中等essay未尝试面试订阅4290A Fast Sanity Check for Clustering AnswersWhy can clustering results vary a lot across random seeds even on the same dataset?机器学习中等essay未尝试面试订阅4301Label Smoothing Loss 1Mixup combines one-hot labels for class 1 and class 4 in a 4-class problem with lambda = 0.3 on class 1's example. What mixed target vector is produced?机器学习中等数值题未尝试面试订阅