pyrfume.optimization module

class pyrfume.optimization.BetterFitness(values=())

Bases: deap.base.Fitness

class pyrfume.optimization.OdorantSetOptimizer(library: pandas.core.frame.DataFrame, n_desired: int, weights: list, keep_cids: bool = None, fitness=None, n_gen: int = 300, mu: int = 100, lamda: int = 200, p_cx: float = 0.4, p_mut: float = 0.4, sel='selBest', standardize_weights=False, npartitions=1)

Bases: object

compute_weight_stats()

Compute scores for a bunch of random individuals to use for standardizing fitness scores and thus making weights interpretable as # of sd’s better or worse than random individuals

crossover(ind1, ind2) → tuple

Apply a crossover operation on input sets.

Args:

ind1: The first individual ind2: The second individual

Returns:

tuple: Two new individuals

dask_map(f, x: list) → list

A mapping function for Dask. Used for multithreading.

Args:

f: Any function. x (list): A list of inputs to be sequentially passed

to that function.

Returns:

list: A list of outputs from that function.

eval_individual(individual, standardize: bool = None)

Evaluate the fitness of the odorant set.

Args:

individual: A DEAP individual as registered above. standardize (bool, optional): Whether to standardize the

resulting score.

Returns:

A fitness score.

eval_mean(individual, column: str) → float

Evaluation the mean of column for the selected individual

Args:

individual: An individual. column (str): A column name from the library dataframe.

Returns:

float: The mean of that column, for the subset

described by the individual

eval_sum(individual, column: str)

Evaluation the sum of column for the selected individual

Args:

individual: An individual. column (str): A column name from the library dataframe.

Returns:

float: The sum of that column, for the subset

described by the individual

force_keep(ind)

Ensure that the CIDs in self.keep get included

Args:

ind: An individual.

Returns:

An individual with certain additional items kept.

mutate(individual) → tuple

Mutation that pops or adds some elements.

Args:

individual: An individual

Returns:

tuple: A new individual (wrapped in a tuple)

plot_score_history()

Plot the history of each of the scores over generations.

run(pop=None, hof=None, quiet: bool = False) → tuple

Run the optimization

Args:

pop: The population to start with. hof: The Hall of Fame to start with quiet: (bool, optional): Whether to suppress stdout.

Defaults to False.

Returns:

tuple: The population, stats, Hall of Fame, and logboox

pyrfume.optimization.get_coverage(odorant_indices, space, sigma=2)

This function will be used to compute coverage of odorant space during optimization. We want non-selected odorants to be ‘covered’ as much as possible by selected ones

pyrfume.optimization.get_entropy(odorant_indices, space, bins_per_dim=10)

pyrfume.optimization.get_spacing(odorant_indices, space, n=5)

These function will be used to determine the spacing between the selected odorants in the low-d manifold. We don’t want to select odorants which are right next to each other in odorant space

pyrfume.optimization.suppress_stdout(on)