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 →找到 15 个结果
中文题目题目4314 · 机器学习
Before Trusting Early Stopping
Your validation metric is noisy day to day. Before treating the first local peak as the stopping point, what should you calibrate?
打开 →题目4415 · 机器学习
Before Trusting Feature Importance
A return feature looks very important in a trained model. What should you check first before concluding it captures genuine alpha?
打开 →题目2513 · 机器学习
Why Correlated Features Frustrate Pure Lasso 17
Why does pure lasso often behave erratically when several features are highly correlated and similarly predictive?
打开 →题目4462 · 机器学习
Before Optimizing Weights
What should you inspect first before trusting an optimizer's exact signal weights?
打开 →题目2266 · 数理金融
Copula Desk Intuition 11
Why should a modeler distrust a dependence setup that implies a very benign calm regime and an absurdly catastrophic stress regime?
打开 →题目4804 · 数理金融
HJM Scenario Analysis 14
Why is fitting today's curve perfectly not enough to trust an HJM model for exotic rate options?
打开 →题目4680 · 数理金融
Local Vol Diagnostic 15
Before trusting a local-vol calibration, what should you inspect first besides raw fit error?
打开 →题目4381 · 机器学习
More Configurations, More Optimism
Fold scores vary wildly because different time periods behave very differently. What is the first tuning response you should consider before trusting a single mean CV number?
打开 →题目4784 · 数理金融
Short-Rate Model Scenario Analysis 19
Why is calibrating only today's discount curve not enough to trust a short-rate model for option risk?
打开 →题目5431 · 金融与交易
Stale Print Adjustment 1
Current mid is 100. A stale print was 100.4 when the mid at that time was 99.9. You only trust 0.5 of that old deviation now, and a fresh peer signal adds 0.06. What fair value do you get?
打开 →题目5432 · 金融与交易
Stale Print Adjustment 2
Current mid is 50.2. A stale print was 49.8 when the mid at that time was 50. You only trust 0.3 of that old deviation now, and a fresh peer signal adds -0.04. What fair value do you get?
打开 →题目5435 · 金融与交易
Stale Print Adjustment 5
Current mid is 149.8. A stale print was 150.5 when the mid at that time was 149.7. You only trust 0.25 of that old deviation now, and a fresh peer signal adds 0.08. What fair value do you get?
打开 →题目2715 · 机器学习
Why Economic Logic Acts Like a Prior Against Noise
Why does a strategy with a credible economic mechanism deserve more trust than a statistically similar strategy with no coherent story?
打开 →题目5336 · 金融与交易
Portfolio VaR Below Sum Of Desk VaRs
At the daily risk meeting, the CRO notes that the standalone historical VaRs of three desks add up to 42 million, but the portfolio historical VaR comes out to only 31 million after a new basis book was added. What is the most likely explanation, and what is one concrete systems
打开 →