Understanding exchange network dynamics with Python
summary
Builds a mixture latency model encoding queueing, bursts and reaction windows, fits it with pymc3 (variational inference) alongside small supervised pieces from scikit-learn, and implements a discrete‑event simulation in simpy that models switches, queues and a matching engine. Fits match observed distributions reasonably well, uses energy‑distance and random/grid search for calibration, and reports that being 50 ns faster raises success rate toward 88% success.
AI-written summary · about
chapters (14)
- 0:00Exchange network dynamics intro
- 1:11Automated trading systems
- 3:39Network performance data challenges
- 5:07Encoding system knowledge
- 6:30Probabilistic programming with PyMC3
- 9:59PyMC3 model results
- 11:52Discrete event simulation basics
- 14:30Network delays in simulations
- 16:43Simple network simulation
- 19:54Randomness and bursts in simulations
- 21:42Fitting simulation to data
- 23:09PyMC3 vs SimPy
- 25:01Summary
- 26:15Q&A
related talks
43mSimulations in Python: Discrete Event Simulation with SimPy
35mWhat we learned by running a large custom Bayesian...
01h 26mProbabilistic Programming and Bayesian Computing with PyMC
01h 27mProbabilistic Python: An Introduction to Bayesian Modeling with PyMC
30mDisease Modeling with Scipy and PyMC
42mSolving Real-World Business Problems with Bayesian Modeling
33mUntangling the complexity of demand forecasting models: building a Market Simulator
39mProbablistic Programming Data Science with PyMC3