new OMGOptimizer(dimensions, n_random_starts, mutation_rate)
Only Mutation Genetic Optimizer;
A class that performs optimization via random permutations to the best
found point thus far. Such approach in particular yields better results
than with crossover on the SigOpt's "evalset" set of problems.
Parameters:
| Name | Type | Default | Description |
|---|---|---|---|
dimensions |
Array | A list of dimensions or a Space object. Describes the space of values over which a function will be optimized. | |
n_random_starts |
Integer | 13 | Determines how many points wil be generated initially at random. The points are not generated at random after this number of evaluations has been reported to the optimizer. |
mutation_rate |
Number | 0.1 | A value in the range of (0.0, 1.0] |
Properties:
| Name | Type | Description |
|---|---|---|
X |
Array | An array of arguments tried. |
Y |
Array | An array of function values observed. The order corresponds to the order in arguments array. |
best_x |
Array | An argument that results in minimal objective function value. |
best_y |
Number | Minimal objective value observed. |
space |
Space | Optimization space over which the optimization is done. |
- Source:
Methods
ask() → {Array}
Generates the next point to evaluate. Different points will be generated for multiple calls, which can be used for parallelisation of optimization.
- Source:
Returns:
a point to evaluate.
- Type
- Array
rnd(p) → {Boolean}
Generates a boolean value at random. Is used for random mutations.
Parameters:
| Name | Type | Description |
|---|---|---|
p |
Number | Probability of generation of true value |
- Source:
Returns:
a randomly generated boolean value.
- Type
- Boolean
tell(X, Y)
Function for reporting of the observed function values
Parameters:
| Name | Type | Description |
|---|---|---|
X |
Array | Array of observed points. |
Y |
Array | Array of objective values corresponding to the points that were evaluated. |
- Source: