题目5061 · 机器学习
Why Nested Evaluation Matters
Why is it unfair to compare two tuned models on the same validation folds that were also used to pick their hyperparameters?
打开 →找到 30 个结果
中文题目题目4196 · 机器学习
Kernel Decision Function Evaluation 1
A kernel SVM prediction uses two support vectors. Their signed contributions at a test point are +1.2 and -0.4, and the bias is -0.1. What score and predicted class result?
打开 →题目4198 · 机器学习
Kernel Decision Function Evaluation 3
Using the degree-3 polynomial kernel K(x,z)=(x·z+1)^3, what is K((1,1),(2,-1))?
打开 →题目1758 · 统计
Selection-on-Survivors Bias in Strategy Evaluation
A desk only records post-launch performance for strategies that first clear an internal backtest hurdle. Why does regressing realized performance on backtest score inside the launched set generally fail to recover the unconditional relationship?
打开 →题目2467 · 机器学习
Unsupervised Preprocessing Can Still Distort Evaluation
Why can fitting an unsupervised step like PCA or quantile normalization on all rows still make the final reported test error too optimistic?
打开 →模块1.2.2 · 金融与量化投资 · 股票
估值与交易机制
equities · valuation · relative-valuation · multiples · comps · dcf · wacc · fcff
打开 →模块4.2.3 · 量化全流程 · Alpha 研究
信号评估与合成
signal-evaluation · information-coefficient · rank-ic · spearman-ic · information-ratio · fundamental-law · grinold-kahn · breadth
打开 →模块4.1.2 · 量化全流程 · 量化数据
基本面、另类数据与数据基础设施
fundamental-data · financial-statements · income-statement · balance-sheet · cash-flow-statement · valuation-ratios · p-e-ratio · ev-ebitda
打开 →课程因子表现与中国 A 股 · 因子投资
因子择时与轮动:实务派的默认选择
周三早上 8:50,上海陆家嘴某百亿私募的月度因子轮动委员会;参考宏观面板由 NBS 制造业 PMI、社融存量、SHIBOR、CSI300 IV、10y 国债 1y SHIBOR 斜率与新增基金账户数构成。上周 HML 打印 +4% 让股票组 PM 主张把价值在因子风险中的权重从 20% 提到 30%;宏观主管报出最新 ISM PMI 与收益率曲线斜率两项均...
打开 →题目4387 · 机器学习
Before Adding More Dimensions
Two hyperparameter settings differ in mean CV score by only 0.001, while the estimated standard error is 0.010. What is the first sensible interpretation?
打开 →题目4390 · 机器学习
Before Blaming The Search
The current best setting sits at extreme values on both the learning-rate and regularization grids. What should your next search action be?
打开 →题目4389 · 机器学习
Before Choosing Nested CV
A categorical encoder was fit once on all rows and then reused inside cross-validation. What is the immediate correction?
打开 →题目4361 · 机器学习
Before Comparing Curves
Before you compare ROC and PR curves across models, what dataset property should you check first?
打开 →题目4363 · 机器学习
Before Picking a Threshold
What is the first quantity you should pin down before optimizing a classification threshold?
打开 →题目4362 · 机器学习
Before Recalibrating
A model looks poorly calibrated. What should you check first before concluding the model itself is broken?
打开 →题目4388 · 机器学习
Before Reporting Best Score
A time-series tuning run says a very short lookback window wins, but recent live performance has deteriorated sharply. What should you inspect first before widening the search?
打开 →题目4364 · 机器学习
Before Trusting AUC
AUC improved a little after retraining. What should you ask first before declaring the new model practically better?
打开 →题目4386 · 机器学习
Before Tuning
Training AUC is very high but CV AUC is near chance. Before trying more hyperparameter values, what is the first diagnostic step?
打开 →题目4365 · 机器学习
Calibration Is Not Enough
Why should you hesitate before selecting a perfectly calibrated model that has weak ranking power?
打开 →题目2239 · 数理金融
CDS Product Intuition 19
Why can short-dated CDS have a small RPV01 even when the spread itself is wide?
打开 →题目3737 · 随机过程
Choosing a Measure for a Macroeconomic Scenario Tree
A strategist is building a scenario tree for inflation, GDP, and unemployment to assess portfolio vulnerability. Which measure is conceptually natural?
打开 →题目5051 · 机器学习
Cost-Sensitive Deployment Choice 11
Model A makes 8 false positives and 2 false negatives on a validation set. Model B makes 6 false positives and 5 false negatives. If a false negative costs 10 units and a false positive costs 1 unit(s), which model has lower expected validation cost and what are the two costs?
打开 →题目2233 · 数理金融
Coupon Implied by an Observed Upfront 13
A CDS market spread is 0.09 and RPV01 = 3.5. The desk observes an upfront of 0.14 per unit notional. Under upfront ≈ (s_mkt - coupon)*RPV01, what coupon is implied?
打开 →题目2235 · 数理金融
Coupon Needed for a Target Upfront 15
A CDS has market spread 0.04 and risky annuity RPV01 = 4.8. What standard coupon would make the linearized upfront equal 0.024 per unit notional?
打开 →题目5106 · 金融与交易
Growing Perpetuity 1
The next payment of a growing perpetuity is 5 in one year and then grows at 0.02 forever. If the discount rate is 0.07, what is the present value?
打开 →题目5109 · 金融与交易
Growing Perpetuity 4
The next payment of a growing perpetuity is 4.5 in one year and then grows at 0.025 forever. If the discount rate is 0.08, what is the present value?
打开 →课程信号评估与合成 · Alpha 研究
IC、IR 与主动管理基本定律
周一上午,你在上海的一家 量化 私募。研究主管 在桌边停下来,看了一眼你 上周提交的 12 1 动量 信号的 DSL,问了一句话:「IC 是 多少?」这就是 4.2.3 模块 整个 评估 工序的 起点。你 已经 按 4.2.2 的 规范 把 信号 构造 完毕——alpha 公式 写好了,标准化 流水 跑通了,T+1 滞后 处理过了——下一步 不是 再 优化 ...
打开 →题目2221 · 数理金融
Infer Flat Hazard From a Par CDS Quote 1
Under the flat-intensity approximation for a CDS, the par spread satisfies s ≈ lambda*(1-R). If the par spread is 0.018 and recovery is 0.4, what flat hazard lambda is implied?
打开 →题目4666 · 数理金融
Inverse Spot From Affine Local Vol 1
A local-vol surface is parameterized by sigma_loc(S,t) = 0.2 + 0.12*(S/100 - 1). At what spot level S does the local volatility equal 0.212?
打开 →题目4667 · 数理金融
Inverse Spot From Affine Local Vol 2
A local-vol surface is parameterized by sigma_loc(S,t) = 0.18 + -0.08*(S/80 - 1). At what spot level S does the local volatility equal 0.188?
打开 →