FRM Part II · FRM Exam Part II · Advances in Artificial Intelligence: Implications for Capital Markets Activities
A quantitative trading firm uses a deep reinforcement learning agent to execute large equity orders. Backtests show strong results, but in live trading the agent performs poorly after a regime change to high volatility. Which explanation is MOST consistent with a known limitation of such AI trading models?
The best explanation is distribution shift. The agent learned a policy from historical data that does not resemble the new high-volatility regime, so its behavior does not generalize out of sample, even though backtests looked strong.
- AThe agent was trained on historical data whose distribution differs from the new regime, so its learned policy does not generalizeCorrect
- BReinforcement learning cannot be backtested under any circumstances
- CHigh volatility always improves execution quality for learning agents
- DThe agent's poor performance proves the model was underfitted to noise
Explanation
Models learn patterns from past data; when the data-generating process shifts (distribution shift), the learned policy may fail. Underfitting to noise is not the point, and volatility does not guarantee better execution. Backtesting is possible, just imperfect.
Did you get it right without looking?
One question tells you little. A timed set on Advances in Artificial Intelligence: Implications for Capital Markets Activities shows your real accuracy, how long you take and where you lose marks.
More Advances in Artificial Intelligence: Implications for Capital Markets Activities questions
- A bank compares a transparent logistic regression with a deep neural network for a trade surveillance alert system. The neural network has h…
- A firm adds an L1 (lasso) penalty to a regression-based machine learning model with 200 candidate predictors. Compared with an unpenalized m…
- A trading desk deploys a deep neural network to generate short-term signals. Validators find that the model performs well in backtests but i…
- A trading firm uses a reinforcement-learning algorithm for execution. A risk manager worries that the algorithm could learn behaviour that r…
- A trading firm's AI surveillance model was trained on two calm years of order-book data. During a sudden volatility spike, alert patterns ch…
- A trading desk deploys a deep neural network to generate execution signals. The model performs well in testing, but the risk committee canno…