Machine Learning
8 min read
Temporal Splits and the Graph ML Leakage Problem
How look-ahead bias silently overfits link prediction models, and why the fix is architectural not a hyperparameter.
Read essay ↗No alpha tips. Just math and techniques.
How look-ahead bias silently overfits link prediction models, and why the fix is architectural not a hyperparameter.
Read essay ↗Static cointegration tests give you a hedge ratio for yesterday. Here's how to track it in real time without refitting the model.
Read essay ↗