Crate atelier_quant 

atelier-quant

Quantitative modelling toolkit for the atelier-rs trading engine.

This crate provides point-process models, interarrival-time analysis, and probabilistic sampling utilities aimed at modelling the microstructure of limit order book event streams.

Point-process models

Model	Module	Estimation
Univariate Hawkes (exponential kernel)	hawkes	MLE via projected gradient ascent with Armijo line search.
Homogeneous Poisson	poisson	Closed-form MLE: λ̂ = (n−1)/T.

Both models expose simulation (hawkes::HawkesProcess::generate_values, poisson::PoissonProcess::generate_values), goodness-of-fit diagnostics (time-rescaling residuals, compensator), and information criteria (AIC, BIC) for model comparison.

Interarrival analysis

The arrivals module bridges raw exchange data (loaded via atelier-data) to the estimation routines:

1. Extract timestamps from Orderbook or Trade vectors.
2. Compute interarrival deltas at any TimeResolution.
3. Produce descriptive statistics (mean, variance, CV, skewness, excess kurtosis) to guide model selection.

Probability distributions

The probs module provides lightweight sampling from Uniform, Normal, Poisson, and Exponential distributions via the Sampling trait.

Re-exports

pub use probs::*;

pub use arrivals::*;

Modules

arrivals

Interarrival-time extraction, validation, and descriptive statistics.

artifact

Model artifact schema for inter-arrival forecasting.

config

Configuration schemas for inter-arrival model binaries.

errors

Error types for the arrivals module.

forecast

Forecast evaluation and Monte-Carlo ensemble reduction.

hawkes

Self-exciting Hawkes point-process model.

poisson

Homogeneous Poisson point-process model.

probs

Probability distribution sampling and density estimation.