Easy Bayesian regularization for fitting financial time series and curves
summary
Claim: massively overparameterize curves and then remove degrees of freedom with convex regularizers and you get robust fits. Uses Python notebooks (code on GitHub), weighted least squares with inverse bid-ask spread weighting, derivative-penalties like a second-derivative penalty to enforce smoothness, and a time-update as a Kalman-filter style diffusion with a step-covariance matrix. Demonstrations show tighter confidence bands, better handling of noisy bid/ask data, easy incorporation of inequality constraints via quadratic programming, and sensible behavior across jumps.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:04Overparameterize curves and regularize
- 4:39Penalizing derivatives for smoothness
- 7:10Weighted least squares with spreads
- 10:48Bayesian smoothing for time series
- 16:42Combining curves and time series
- 25:59Real-world constraints and tricks
- 31:18Handling jumps and change detection
- 33:24Q&A
related talks
01h 26mNew Trends in Estimation and Inference
49mSo you think you know about linear regression ...
01h 26mProbabilistic Programming and Bayesian Computing with PyMC
41mReliably forecasting time-series in real-time
47mScale Calibration for High-Dimensional Robust Regression
01h 10mBayesian Optimization and Machine Learning for Accelerating Experiments in the Physical Sciences
38mLearning credit networks with applications in financial industry
01h 24mBayesianism and Survival Analysis