This page documents the various computations that can be used to define computed variables in the problem description file.

In this section:

Computation Library

The computation library includes common arithmetic computations. To use a computation from the library, simply enter its name in snake_case as the computation parameter of the computed variable.

Summation

summation

Supported Inputs Types

fixed (two or more), all

Inputs Order

Any

Formula

operand1 + operand2 + … + operandN

Examples

Subtraction

subtraction

Supported Inputs Types

fixed (exactly two)

Inputs Order

• minuend

• subtrahend

Formula

minuend - subtrahend

Example

Multiplication

multiplication

Supported Inputs Types

fixed (any number), all

Inputs Order

Any

Formula

operand1 * operand2 * … * operandN

Examples

Division

division

Supported Inputs Types

fixed (exactly two)

Inputs Order

• dividend

• divisor

Formula

dividend / divisor

Example

Exponentiation

exponentiation

Supported Inputs Types

fixed (exactly two)

Inputs Order

• base

• exponent

Formula

base ^ exponent

Example

Rated Summation

rated_summation

Supported Inputs Types

fixed (two or more)

Inputs Order

• base

• rate 1

• rate 2

• …

• rate N

Formula

base * (1 + rate1 + rate2+ ... + rateN)

which is equal to base + base * rate1 + base * rate2 + … + base * rateN

Example

Rated Subtraction

rated_subtraction

Supported Inputs Types

fixed (two or more)

Inputs Order

• base

• rate 1

• rate 2

• …

• rate N

Formula

base * (1 - (rate1 + rate2+ ... + rateN))

which is equal to base - base * rate1 - base * rate2 - … - base * rateN

Example

Minimum

minimum

Supported Inputs Types

fixed (two or more), all

Inputs Order

Any

Formula

Min(operand1, operand2, …, operandN)

Examples

Maximum

maximum

Supported Inputs Types

fixed (two or more), all

Inputs Order

Any

Formula

max(operand1, operand2, …, operandN)

Examples

Average

average

Supported Inputs Types

fixed (one or more), all

Inputs Order

Any

Formula

sum(operand1, operand2, …, operandN) / N

Examples

Standard Deviation

stdev

Supported Inputs Types

fixed (one or more), all

Inputs Order

Any

Formula

population_stdev = sqrt(sum((operand1 - avg)², (operand2 - avg)², …, (operandN - avg)²) / N)

Examples

Percentile

percentile

Warning: this computation may trigger some gradient losses and failures to find a solution if enough inputs share the same starting value. It is not recommended to use it in optimization use cases, but it is safe in simulation use cases.

Supported Inputs Types

fixed, all

Inputs Order

Any

Parameter

index: the index of the percentile to be outputed

Formula

Using Guava
Quantiles.percentiles().index(index).compute(dataset)

Which is a linearly interpolated percentile.

Examples

Rounding

type: rounding

Supported Inputs Types

fixed (exactly one)

Formula

round(value)

The rounding is performed following the list of rounding rules passed as parameter.

There are two types of roundig rules available to the user:

• slots: between lower_boundary (included) and upper_boundary (included), the output is only allowed to take values that end with the numbers specified in the slots list, repeating the rounding pattern with a specified period.

  • E.g. for prices between 0.19€ and 99.99€ to be rounded at 0.19, 0.49 and 0.99 the rule should be:

    • lower_boundary: 0.19

    • upper_boundary: 99.99

    • slots: [0.19, 0.49, 0.99]

    • period: 1

• uniform_increment: between the specified lower_boundary (included) and upper_boundary (included) the output is only allowed to take values that match lower_boundary + N * increment where N is a positive integer.

  • E.g. for prices to be rounded to .99€ between 99.99€ and 9999.99€ the rule should be:

    • lower_boundary: 99.99

    • upper_boundary: 9999.99

    • increment: 1

• The user also has to specify the general decimal precision, i.e. the maximum desired number of decimal digits among all the rounding rules.

  • This precision has to be compatible with the finest increments and slots. E.g. if the finest increment is 0.01, then the precision should be 2.

• The user can specify any number of rounding rules, the only conditions is for them to be contiguous, i.e. the upper_boundary of the first rule should be equal to the lower_boundary of the next rule.

Examples

Custom Computations

If you need more sophisticated computations, you can write your own expression inline in the description file. You can also provide a Java implementation but in this case, you need to have your own build of the Optimization Engine.

Inline Expressions

type: inline

This lets you specify a custom expression for your computation, as well as specific parameters that can either be inline or retrieved from data.

Supported Inputs Types

fixed (any number), all

Inputs Order

Any order, but the index is how you reference inputs in the expression.

Parameters Order

Any order, but the index is how you reference inputs in the expression.

Expression

A Java expression.

• java.lang.Math is supported

  • You may use other classes but you will need their fully qualified name.

• Access inputs with inputs.value(n), where n is the index of the input (the first index is 0).

• Access parameters with params.value(n), where n is the index of the parameter (the first index is 0).

• inputs and params also implement DoubleIterable asIterable() to easily iterate over the values or easily compute averages and such.

Example

Java Implementations

custom

This lets you use your own computation implemented as a Java class.

Supported Inputs Types

all, fixed (any number)

Inputs Order

Any order. The Java implementation will have a list in the same order.

Parameters Order

Any order, but the index is how you reference inputs in the expression.

Class

A fully qualified Java class name.

• java.lang.Math is supported

  • You may use other classes but you will need their fully qualified name.

• Access inputs with inputs.value(n), where n is the index of the input (the first index is 0).

• Access parameters with params.value(n), where n is the index of the parameter (the first index is 0).

• inputs and params also implement DoubleIterable asIterable() to easily iterate over the values or easily compute averages and such.

Prerequisites

• The Java class of your computation must be added to the OE classpath.

  • This means it only works with your own OE build for now.

• The Java class of your computation must either implement

  • net.pricefx.optimization.engine.description.CustomComputation (recommended)
    or

  • net.pricefx.optimization.declaration.Declaration.DeclaredComputation for advanced control over the instantiation of the computation, including accessing coordinates, inputs, and output agents.

• The Java class may declare a constructor with exactly one parameter of the type net.pricefx.optimization.mas.agent.computation.domain.Computation.InputValues to access the parameters defined in the description.

  • This constructor is mandatory if there are 1 or more declared parameters.

Example