History Module

Note that this is the developer API reference for the internal history module. See history array for the user reference.

class history.History(alloc_specs, sim_specs, gen_specs, exit_criteria, H0)

The history class provides methods for managing the history array.

Object Attributes:

These are set on initialization.

Variables
  • H (numpy_structured_array) – History array storing rows for each point. Field names are in libensemble/tools/fields_keys.py

  • offset (int) – Starting index for this ensemble (after H0 read in)

  • index (int) – Index where libEnsemble should start filling in H

  • given_count (int) – Number of points given to sim functions (according to H)

  • returned_count (int) – Number of points evaluated (according to H)

Note that index, given_count and returned_count reflect the total number of points in H and therefore include those prepended to H in addition to the current run.

__init__(alloc_specs, sim_specs, gen_specs, exit_criteria, H0)

Forms the numpy structured array that records everything from the libEnsemble run

update_history_f(D, safe_mode)

Updates the history after points have been evaluated

update_history_x_out(q_inds, sim_worker)

Updates the history (in place) when new points have been given out to be evaluated

Parameters
  • q_inds (numpy array) – Row IDs for history array H

  • sim_worker (integer) – Worker ID

update_history_to_gen(q_inds)

Updates the history (in place) when points are given back to the gen

update_history_x_in(gen_worker, D, safe_mode)

Updates the history (in place) when new points have been returned from a gen

Parameters
  • gen_worker (integer) – The worker who generated these points

  • D (numpy array) – Output from gen_func

grow_H(k)

Adds k rows to H in response to gen_f producing more points than available rows in H.

Parameters

k (integer) – Number of rows to add to H

trim_H()

Returns truncated array