Skip to content

Latest commit

 

History

History
55 lines (52 loc) · 2.32 KB

FEATURES.md

File metadata and controls

55 lines (52 loc) · 2.32 KB
Raw

Existing features

  • regression methods
    • polynomial
    • GLMs: logistic, (quasi-)Poisson, Gamma, exponential
  • optimization methods
    • numerical differentiation, partial derivatives, automatic differentiation (currently with autodiff crate)
    • optimizers
      • Adam, Levenberg-Marquardt, SGD with (Nesterov) momentum
  • numerical integration of functions
    • trapezoid, Romberg, 5-point Gauss-Legendre quadrature
  • basic statistical distributions
    • continuous
      • (Multivariate) Normal, Beta, Gamma, Chi Squared, Student's T, Uniform, Exponential, Pareto, Gumbel
    • discrete
      • Bernoulli, Binomial, Poisson, Discrete Uniform
    • sampling, PDFs/PMFs
    • analytic means and variances
  • mathematical and statistical functions
    • gamma, digamma, beta
    • logistic, logit, (general) boxcox transform, softmax
    • binomial coefficients
    • linear interpolation (with extrapolation)
  • statistical methods
    • (sample) covariance, mean, variance, min, max
  • time series models
    • autoregressive models
    • related functions
      • autocorrelation, autocovariance, differencing
  • validation methods
    • resampling
      • bootstrap, jackknife
  • linear algebra: both BLAS/LAPACK and Rust implementations
    • vector and matrix structs
      • overloaded arithmetic operations for combinations of {matrix, vector, scalar} with automatic broadcasting a la numpy
    • general utilities
      • dot product, (blocked) matrix multiplication, matrix inversion, Toeplitz matrix, Vandermonde matrix, (infinity) norm, linear solve, transpose, design matrix
      • vector-vector, scalar-vector, vector-scalar operations with loop unrolling
    • decompositions and solvers
      • LU, Cholesky

Planned features

  • distributions: CDFs, fitting to data
  • more time series models (SARIMA, exponential smoothing models, trend decomposition)
  • non-linear optimizers (BFGS)
  • ODE integrators (leapfrog, RK4)
  • clustering algorithms (k-means/EM, DBSCAN)
  • more regression models (mixed models, GP, penalized models, splines)
  • prediction trees (CART, random forests, gradient boosted trees)
  • order statistics (quantiles)
  • statistical tests (t-test, ANOVA, Kolmogorov-Smirnov, Anderson-Darling)
  • data preprocessing (outlier detection, standardization, dimensionality reduction (PCA))
  • more linear algebra decompositions (QR, SVD)
  • samplers? rejection, RWM, HMC, NUTS, (dynamic) nested sampling