1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123

// Copyright 2018 Stefan Kroboth
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://apache.org/licenses/LICENSE-2.0> or the MIT license <LICENSE-MIT or
// http://opensource.org/licenses/MIT>, at your option. This file may not be
// copied, modified, or distributed except according to those terms.

//! Optimizaton toolbox
//!
//! TODOs
//!
//! * Stopping criterions which can be stacked, also return the reason why a computation terminated
//! * keep track of cost function values
//! * count the number of cost function / gradient evaluations and return them
//! * redesign how lower and upper bound are dealt with. making them optional should be better.
#![recursion_limit = "1024"]
#![cfg_attr(feature = "clippy", feature(plugin))]
#![cfg_attr(feature = "clippy", plugin(clippy))]
#![warn(missing_docs)]
#[macro_use]
extern crate error_chain;
extern crate ndarray;
extern crate ndarray_linalg;
extern crate num;
extern crate rand;

/// Macros
#[macro_use]
pub mod macros;

use std::default::Default;
use num::{Bounded, ToPrimitive};
use errors::*;
use parameter::ArgminParameter;
use result::ArgminResult;

/// Trait for cost function values
/// TODO: Do this with trait aliases once they work in rust.
pub trait ArgminCostValue: Bounded + ToPrimitive + Copy + Default + PartialOrd {}
impl<T> ArgminCostValue for T
where
    T: Bounded + ToPrimitive + Copy + Default + PartialOrd,
{
}

/// Trait every solve needs to implement (in the future)
pub trait ArgminSolver<'a> {
    /// Parameter vector
    // type A: ArgminParameter<Self::A>;
    type Parameter: ArgminParameter;
    /// Cost value
    type CostValue: ArgminCostValue;
    /// Hessian
    type Hessian;
    /// Initial parameter(s)
    type StartingPoints;
    /// Type of Problem (TODO: Trait!)
    type ProblemDefinition;

    /// Initializes the solver and sets the state to its initial state
    // fn init(&mut self, &'a Problem<'a, Self::A, Self::B, Self::C>, &Self::D) -> Result<()>;
    fn init(&mut self, &'a Self::ProblemDefinition, &Self::StartingPoints) -> Result<()>;

    /// Moves forward by a single iteration
    fn next_iter(&mut self) -> Result<ArgminResult<Self::Parameter, Self::CostValue>>;

    /// Run initialization and iterations at once
    fn run(
        &mut self,
        &'a Self::ProblemDefinition,
        &Self::StartingPoints,
    ) -> Result<ArgminResult<Self::Parameter, Self::CostValue>>;

    /// Handles the stopping criteria
    fn terminate(&self) -> bool;
}

/// Definition of the return type of the solvers
pub mod result;

/// Traits for implementing parameter vectors
pub mod parameter;

/// Problem formulation
pub mod problem;

/// Operator
pub mod operator;

/// A set of test functions like Rosenbrock's function and so on.
pub mod testfunctions;

/// Backtracking line search
pub mod backtracking;

/// Simulated Annealing
pub mod sa;

/// Gradient Descent
pub mod gradientdescent;

/// Nelder Mead method
pub mod neldermead;

/// Newton method
pub mod newton;

/// Landweber algorithm
pub mod landweber;

/// Conjugate Gradient method
pub mod cg;

/// Errors using `error-chain`
mod errors;

#[cfg(test)]
mod tests {
    #[test]
    fn it_works() {
        assert_eq!(2 + 2, 4);
    }
}