第 18 / 41 页
非代码面试题
显示 20 / 811 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
4115A Fast Sanity Check for Risk-Parity AnswersWhat is a fast sanity check after you compute a risk-parity or inverse-vol allocation?金融与交易中等essay未尝试面试订阅4121Spread-Impact Decomposition 1A buy slice arrives when the mid is 50. Crossing the spread costs 0.02 dollars per share, temporary impact is 0.03, and permanent impact is estimated at 0.01. What are the expected average fill price and the post-trade mid immediately after the market digests the permanent impact?金融与交易中等数值题未尝试面试订阅4122Spread-Impact Decomposition 2A sell slice arrives when the mid is 32.5. Crossing the spread costs 0.015 dollars per share, temporary impact is 0.025, and permanent impact is estimated at 0.008. What are the expected average fill price and the post-trade mid immediately after the market digests the permanent impact?金融与交易中等数值题未尝试面试订阅4123Spread-Impact Decomposition 3A buy slice arrives when the mid is 80.2. Crossing the spread costs 0.03 dollars per share, temporary impact is 0.04, and permanent impact is estimated at 0.015. What are the expected average fill price and the post-trade mid immediately after the market digests the permanent impact?金融与交易中等数值题未尝试面试订阅4131VWAP Schedule in a Liquidity-U-Shaped DayYou need to buy 6% of daily volume in a liquid index future by the close, and the PM is benchmarked strictly to day-long VWAP rather than arrival. Which schedule is the most natural starting point?金融与交易中等derivation未尝试面试订阅4132Alpha Decay and Front-LoadingA signal is expected to decay within the next 20 minutes, and waiting is more expensive than crossing a bit more spread now. Should the schedule become more or less front-loaded?金融与交易中等derivation未尝试面试订阅4133Low-Signal Passive LiquidationYou need to liquidate a medium-size position over several days with no urgent alpha and strong concern about footprint. Which scheduling style is the better default: high-POV aggression or a lower-participation passive schedule?金融与交易中等derivation未尝试面试订阅4134Closing Auction Use CaseA benchmarked portfolio must minimize tracking error to the official close, and the stock has deep closing-auction liquidity. Which venue or schedule component becomes especially attractive?金融与交易中等derivation未尝试面试订阅4135Participation Cap in Thin NamesIn a thin name, why might a trader impose a strict maximum participation rate even when the order is behind schedule?金融与交易中等derivation未尝试面试订阅4136Why Quoted Spread Understates True CostWhy can quoted spread be a poor summary of true execution cost for a real institutional order?金融与交易中等essay未尝试面试订阅4137Why Passive Orders Can Still Lose MoneyWhy is it wrong to think that passive orders are always cheap just because they do not cross the spread?金融与交易中等essay未尝试面试订阅4138Why Impact Models Need a Urgency TermWhy is an execution problem incomplete if it models market impact but ignores alpha decay or urgency?金融与交易中等essay未尝试面试订阅4139Why Dark Liquidity Is Not Free LiquidityWhy can dark-pool access reduce displayed footprint and still fail to lower true cost much?金融与交易中等essay未尝试面试订阅4156Small Labeled Sample with Plausible StructureFor an observation x, a generative model summarizes the evidence as likelihood ratio p(x|Y=1)/p(x|Y=0) = 5. If the prior probability of class 1 is 0.2, what posterior probability P(Y=1|x) follows, and what 0.5-threshold decision does that imply?机器学习中等derivation未尝试面试订阅4157Lots of Labels but Misspecified Density StoryFor an observation x, a generative model summarizes the evidence as likelihood ratio p(x|Y=1)/p(x|Y=0) = 0.5. If the prior probability of class 1 is 0.4, what posterior probability P(Y=1|x) follows, and what 0.5-threshold decision does that imply?机器学习中等derivation未尝试面试订阅4161Why Naive Bayes Can Work Despite Wrong IndependenceYou have only a few hundred labeled observations, but domain knowledge gives a plausible class-conditional structure and you also have many unlabeled feature vectors. Would you start with a generative or a discriminative model first?机器学习中等essay未尝试面试订阅4162Why Discriminative Models Often Win AsymptoticallyA classifier was trained last quarter, and now only the class prevalence has shifted while the conditional shape of features given class appears stable. Which side, generative or discriminative, is easier to adjust quickly?机器学习中等essay未尝试面试订阅4163Why Generative Models Handle Missingness More NaturallyYou have millions of labeled examples and care only about predictive accuracy on the deployed label, not about simulating x. Which side usually deserves the first try?机器学习中等essay未尝试面试订阅4164Why Generative and Discriminative Can Share a BoundaryA production system frequently has one sensor missing at test time, but your model family can factor the joint feature likelihood cleanly. Which side gets a practical advantage?机器学习中等essay未尝试面试订阅4165A Fast Sanity Check for Gen-vs-Disc QuestionsThe research desk wants not only labels but also synthetic feature draws conditional on each class for stress testing. Which side is the more natural starting point?机器学习中等essay未尝试面试订阅