第 10 / 23 页
非代码面试题
显示 20 / 453 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
2415Why a Stable but Biased Model May Be Preferred OperationallyWhy might a desk prefer a slightly biased model that behaves predictably over a lower-bias model whose outputs swing wildly retrain to retrain?机器学习困难essay未尝试面试订阅2430Why the Weighted-Brier Bayes Act Is Still a Weighted Mean 7For binary Y and weighted squared loss alpha Y (1-q) 2 + beta (1-Y) q 2, derive the Bayes probability q as a function of p=P(Y=1|X).机器学习困难derivation未尝试面试订阅2438Why Convexity Makes Averaging Predictions SafeWhy does convexity of a loss function support the intuition that averaging similar predictors often cannot hurt too much?机器学习困难essay未尝试面试订阅2463Reusing the Test Set After DebuggingA model is evaluated on test, a bug is found, the code is fixed, and the same test set is used again to verify the fix and choose among two corrected versions. Why is that second use no longer a clean test?机器学习中等essay未尝试面试订阅2470Rare Category Thresholding After Seeing Test CompositionSuppose you choose the minimum frequency for keeping a category only after inspecting how many rare categories appear in the test set. Why is that already a contaminated design choice?机器学习困难essay未尝试面试订阅2483Why Centering Leaves Slopes Unchanged 13Why does centering x and y leave the fitted slope unchanged in simple OLS with an intercept?机器学习中等derivation未尝试面试订阅2488Why Residual Mean Is Zero With an Intercept 18Why must OLS residuals sum to zero whenever an intercept is included?机器学习困难derivation未尝试面试订阅2525One Newton Step for an Intercept-Only Logistic ModelAn intercept-only logistic model is fit to 7 positives and 3 negatives. Starting from b 0 = 0, what is one Newton step b 1 for minimizing the negative log-likelihood?机器学习困难数值题未尝试面试订阅2552Which Split Becomes Best After a Perturbation 19Split A originally has gain 1.20 and split B has gain 1.05. After one row is corrected, A loses 0.10 gain while B gains 0.08. Which split is now best?机器学习中等数值题未尝试免费2565Validation Pruning With an Alpha Charge 23Replacing a single leaf by a 3-leaf subtree reduces validation loss by 4.5. If the complexity charge is alpha = 1.2 per extra leaf, should you keep the subtree?机器学习困难数值题未尝试面试订阅2610Scale-Update Invariance Between Eta and Gamma 6Why does multiplying every leaf update gamma m by c and dividing the learning rate eta by c leave the final additive score unchanged?机器学习困难derivation未尝试面试订阅2615Why Calibration Can Degrade Before Ranking 19Why can late-stage boosting sometimes keep ranking examples well while making the predicted scores less well calibrated?机器学习困难essay未尝试面试订阅2618Why Many Small Corrections Can Beat One Big Tree 21Why can an additive sequence of small boosting steps outperform a single large tree with similar in-sample flexibility?机器学习中等essay未尝试面试订阅2650Usable Training Days After Purging and EmbargoA 100-day event-study sample uses a contiguous 20-day validation block in the middle. Labels look ahead 5 days, and you impose a 2-day embargo on each side of the validation block. How many days remain usable for training?机器学习困难数值题未尝试面试订阅2659Train-Set Size After Purging One Side of a FoldA dataset has 500 time-ordered observations. One validation block uses observations 301 through 350. If a 10-observation purge is applied immediately before the validation block and nowhere else, how many observations remain eligible for training?机器学习困难数值题未尝试面试订阅2664How Many Distinct Hyperparameter Winners Can Outer CV ProduceA nested CV uses 7 outer folds and selects exactly one hyperparameter setting inside each outer fold. What is the maximum possible number of distinct winning hyperparameter settings across outer folds?机器学习困难derivation未尝试面试订阅2668Why Embargo Matters Even With Backward-Looking FeaturesSuppose features use only past prices, yet labels depend on future returns over an event window. Why can an embargo still be necessary around the validation block?机器学习中等essay未尝试面试订阅2669Why Purging and Embargo Solve Different ProblemsWhy is purging not the same thing as embargoing in time-series validation?机器学习中等essay未尝试面试订阅2678Net Signal-to-Cost RatioA strategy's gross expected edge is 5 bps per trade, but execution-cost uncertainty has a standard deviation of 12 bps per trade. What is the gross-edge-to-cost-noise ratio?机器学习中等数值题未尝试面试订阅2680Why Low R-Squared Can Still Be Valuable Yet Hard to VerifyWhy can a signal with tiny explanatory power still be economically useful, while also being unusually hard to validate convincingly?机器学习困难essay未尝试面试订阅