API

struct PiecewiseLinearFunction{T<:AbstractFloat}

Fields

• x::Vector{T} where T<:AbstractFloat: x coordinates of the breakpoints

• y::Vector{T} where T<:AbstractFloat: y coordinates of the breakpoints

• left_slope::AbstractFloat: slope of the function to the left of the first breakpoint

• right_slope::AbstractFloat: slope of the function to the right of the last breakpoint

Evaluate the piecewise linear function f at x.

compose(
    f::PiecewiseLinearFunction,
    g::PiecewiseLinearFunction;
    postprocess_breakpoints
) -> PiecewiseLinearFunction

Compute the composition f ∘ g of two piecewise linear functions. By default, the function checks for and removes redundant breakpoints in the output as a post processing step. For faster computation, set postprocess_breakpoints=false.

This method can also be called using Base.∘

compute_slopes(
    f::PiecewiseLinearFunction
) -> Vector{<:AbstractFloat}

Compute all the slopes of the piecewise linear function, from left to right.

convex_lower_bound(
    f::PiecewiseLinearFunction{T}
) -> PiecewiseLinearFunction

Compute the tightest convex lower bound of a piecewise linear function, using a convex hull algorithm.

convex_meet(
    f::PiecewiseLinearFunction,
    g::PiecewiseLinearFunction
) -> PiecewiseLinearFunction

A faster version of convex_meet that only works on convex functions.

is_convex(f::PiecewiseLinearFunction; eps) -> Any

Check if a piecewise linear function is convex.

is_convex_sequence(f, i1, i2, i3) -> Any

Check if the sequence of points (p1, p2, p3) forms a convex sequence.

old_convex_meet(
    f::PiecewiseLinearFunction{T},
    g::PiecewiseLinearFunction{T}
) -> PiecewiseLinearFunction

Compute the convex meet of two piecewise linear functions, i.e. the tightest convex lower bound.

remove_redundant_breakpoints(
    f::PiecewiseLinearFunction;
    atol
) -> PiecewiseLinearFunction

Return a new piecewise linear function with redundant breakpoints removed.