第 7 / 11 页
非代码面试题
显示 20 / 215 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
2402Second Crossover With a Lower-Bias Flexible ModelA flexible model has excess error 0.02 + 24/n, while a simpler model has excess error 0.14 + 6/n. At what sample size do they tie?机器学习中等derivation未尝试面试订阅2404Data Multiplier Needed to Push Variance Below a Noise Floor FractionA model's variance term is currently 0.30, and irreducible noise is 0.05. If variance scales exactly like 1/n, by what factor must the dataset grow so the variance term falls to 0.05?机器学习中等derivation未尝试面试订阅2405Recover the Irreducible NoiseA model has test MSE 0.92, bias 2 0.15, and variance 0.27. What irreducible noise term is implied?机器学习简单数值题未尝试面试订阅2408Variance of a Three-Model Independent AverageThree independently trained models each have variance 1.8 and negligible bias. What is the variance of their equal-weight average?机器学习简单数值题未尝试面试订阅2409Why More Data Usually Helps a Variance-Dominated Model FirstWhy does collecting more data usually help a high-variance model more than a high-bias model?机器学习困难essay未尝试面试订阅2412Why Model Rankings Can Flip as n GrowsWhy can a simple model beat a flexible one at small n and then lose badly once n is large?机器学习中等essay未尝试面试订阅2413Why Bagging Mainly Targets VarianceWhy is bagging usually described as a variance-reduction tool rather than a bias-reduction tool?机器学习中等essay未尝试面试订阅2414Why Irreducible Noise Caps the Best Achievable Test ErrorWhy can model improvements stall even after both bias and variance seem small?机器学习困难essay未尝试面试订阅2420Why Deployment Preferences Can Differ From Benchmark-MSE PreferencesWhy can the model that minimizes benchmark MSE fail to be the one a production team actually deploys?机器学习困难essay未尝试面试订阅2425Why Asymmetric Loss Changes the Optimal Prediction 21Why does changing the relative penalty on overprediction versus underprediction generally move the Bayes act away from the conditional mean?机器学习困难essay未尝试面试订阅2429Total Huber Loss on a Residual SetUsing Huber loss with delta = 1, compute the total loss on residuals 0.5, -1.2, and 3.0.机器学习中等数值题未尝试面试订阅2433Pinball Loss Subgradient at the Kink 9For pinball loss rho tau(r)=tau r if r>=0 and (tau-1)r if r<0, what is the subgradient set at r=0?机器学习中等derivation未尝试面试订阅2434Why a Strictly Proper Loss Must Prefer Truthful Probabilities 10Why is it desirable for a probabilistic classification loss to be strictly proper rather than merely classification-accurate?机器学习困难derivation未尝试面试订阅2435Why Log-Loss Punishes Overconfidence More Than Brier LossWhy does log-loss react much more harshly than Brier loss when a model assigns near-certainty to the wrong class?机器学习困难essay未尝试面试订阅2439Why Asymmetric Loss Moves the Target Away From the MeanWhy does an asymmetric loss generally make the optimal constant prediction move away from the mean of the target distribution?机器学习困难essay未尝试面试订阅2444Why Quantile Loss Is Useful in Risk Forecasting 17Why is pinball loss natural when the target is a VaR-like forecast rather than a mean forecast?机器学习中等derivation未尝试面试订阅2445Why Tail Forecasts Need Tail-Aligned Losses 25Why is it often a mistake to optimize plain squared loss when the operational task really cares about an extreme tail quantile?机器学习困难essay未尝试面试订阅2452Future Restatements Merged Into Historical FeaturesA researcher joins fundamentals after they were restated months later, then backtests on the original trade dates. Why is this a split-discipline failure even if no test labels were touched?机器学习中等essay未尝试面试订阅2454Feature Screening Before the SplitA team ranks 5,000 candidate features by correlation with the target on the full dataset, keeps the top 30, and only then creates train and test. Why is the later split not enough to rescue the experiment?机器学习中等essay未尝试面试订阅2455Repeated Validation Peeking During ResearchA researcher keeps trying new transformations and only retains the ones that improve the same validation score. Why does the validation set stop being a clean model-selection tool?机器学习困难essay未尝试面试订阅