| title | Simple POMDP | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| type | reference | ||||||||
| status | stable | ||||||||
| updated | 2026-10-10 | ||||||||
| semantic_relations |
|
SimplePOMDP (code/Things/Simple_POMDP/simple_pomdp.py) is a compact seeded
discrete POMDP agent. Its configuration, given as a YAML path or a dict, has
model, state_space, observation_space, action_space, matrices
(A_matrix to E_matrix), inference, and visualization sections, plus an
optional seed. Unknown sections or fields are rejected. A and each B slice
are column-stochastic, preferences are finite, and D and E are normalized
distributions.
Methods: step, run, reset, save_state, load_state, and visualize.
visualize(plot_type) accepts belief_evolution, free_energy_landscape,
policy_evaluation, state_transitions, observation_likelihood, and
efe_components_detailed; any other value raises ValueError. A plot is saved
as <plot_type>.png under visualization.output_dir. The module also exports
SimplePOMDPState and compute_expected_free_energy.
Details and a runnable example: [[code/Things/Simple_POMDP/Simple_POMDP_README]].
The tests are in code/tests/test_simple_pomdp.py:
.venv/bin/python -m pytest code/tests/test_simple_pomdp.py -q