forked from google-deepmind/open_spiel
-
Notifications
You must be signed in to change notification settings - Fork 1
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
PiperOrigin-RevId: 477154449 Change-Id: I544aa226f7ca8fccc889402c7ffd53c7fd3c27fc
- Loading branch information
Showing
4 changed files
with
201 additions
and
65 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
86 changes: 86 additions & 0 deletions
86
open_spiel/python/mfg/algorithms/munchausen_mirror_descent.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,86 @@ | ||
| # Copyright 2022 DeepMind Technologies Limited | ||
| # | ||
| # Licensed under the Apache License, Version 2.0 (the "License"); | ||
| # you may not use this file except in compliance with the License. | ||
| # You may obtain a copy of the License at | ||
| # | ||
| # http://www.apache.org/licenses/LICENSE-2.0 | ||
| # | ||
| # Unless required by applicable law or agreed to in writing, software | ||
| # distributed under the License is distributed on an "AS IS" BASIS, | ||
| # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| # See the License for the specific language governing permissions and | ||
| # limitations under the License. | ||
| """Munchausen Online Mirror Descent.""" | ||
|
|
||
| from typing import Dict, List, Optional | ||
|
|
||
| import numpy as np | ||
|
|
||
| from open_spiel.python import policy as policy_lib | ||
| from open_spiel.python.mfg import value | ||
| from open_spiel.python.mfg.algorithms import mirror_descent | ||
| import pyspiel | ||
|
|
||
|
|
||
| class ProjectedPolicyMunchausen(mirror_descent.ProjectedPolicy): | ||
| """Project values on the policy simplex.""" | ||
|
|
||
| def __init__( | ||
| self, | ||
| game: pyspiel.Game, | ||
| player_ids: List[int], | ||
| state_value: value.ValueFunction, | ||
| learning_rate: float, | ||
| policy: policy_lib.Policy, | ||
| ): | ||
| """Initializes the projected policy. | ||
| Args: | ||
| game: The game to analyze. | ||
| player_ids: list of player ids for which this policy applies; each should | ||
| be in the range 0..game.num_players()-1. | ||
| state_value: The state value to project. | ||
| learning_rate: The learning rate. | ||
| policy: The policy to project. | ||
| """ | ||
| super().__init__(game, player_ids, state_value) | ||
| self._learning_rate = learning_rate | ||
| self._policy = policy | ||
|
|
||
| def action_probabilities(self, | ||
| state: pyspiel.State, | ||
| player_id: Optional[int] = None) -> Dict[int, float]: | ||
| del player_id | ||
| action_logit = [ | ||
| (a, self._learning_rate * self.value(state, action=a) + np.log(p)) | ||
| for a, p in self._policy.action_probabilities(state).items() | ||
| ] | ||
| action, logit = zip(*action_logit) | ||
| return dict(zip(action, mirror_descent.softmax_projection(logit))) | ||
|
|
||
|
|
||
| class MunchausenMirrorDescent(mirror_descent.MirrorDescent): | ||
| """Munchausen Online Mirror Descent algorithm. | ||
| This algorithm is equivalent to the online mirror descent algorithm but | ||
| instead of summing value functions, it directly computes the cumulative | ||
| Q-function using a penalty with respect to the previous policy. | ||
| """ | ||
|
|
||
| def eval_state(self, state: pyspiel.State, learning_rate: float): | ||
| """Evaluate the value of a state.""" | ||
| state_str = state.observation_string(pyspiel.PlayerId.DEFAULT_PLAYER_ID) | ||
| # Return the already calculated value if present. | ||
| if self._state_value.has(state_str): | ||
| return self._state_value(state_str) | ||
| # Otherwise, calculate the value of the state. | ||
| v = self.get_state_value(state, learning_rate) | ||
| self._state_value.set_value(state_str, v) | ||
| return v | ||
|
|
||
| def get_projected_policy(self) -> policy_lib.Policy: | ||
| """Returns the projected policy.""" | ||
| return ProjectedPolicyMunchausen(self._game, | ||
| list(range(self._game.num_players())), | ||
| self._state_value, self._lr, self._policy) |
44 changes: 44 additions & 0 deletions
44
open_spiel/python/mfg/algorithms/munchausen_mirror_descent_test.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,44 @@ | ||
| # Copyright 2022 DeepMind Technologies Limited | ||
| # | ||
| # Licensed under the Apache License, Version 2.0 (the "License"); | ||
| # you may not use this file except in compliance with the License. | ||
| # You may obtain a copy of the License at | ||
| # | ||
| # http://www.apache.org/licenses/LICENSE-2.0 | ||
| # | ||
| # Unless required by applicable law or agreed to in writing, software | ||
| # distributed under the License is distributed on an "AS IS" BASIS, | ||
| # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| # See the License for the specific language governing permissions and | ||
| # limitations under the License. | ||
| """Tests for Munchausen Online Mirror Descent.""" | ||
|
|
||
| from absl.testing import absltest | ||
| from absl.testing import parameterized | ||
|
|
||
| from open_spiel.python.mfg import value | ||
| from open_spiel.python.mfg.algorithms import munchausen_mirror_descent | ||
| from open_spiel.python.mfg.algorithms import nash_conv | ||
| from open_spiel.python.mfg.games import crowd_modelling # pylint: disable=unused-import | ||
| import pyspiel | ||
|
|
||
|
|
||
| class MunchausenMirrorDescentTest(parameterized.TestCase): | ||
|
|
||
| @parameterized.named_parameters(('python', 'python_mfg_crowd_modelling'), | ||
| ('cpp', 'mfg_crowd_modelling')) | ||
| def test_run(self, name): | ||
| """Checks if the algorithm works.""" | ||
| game = pyspiel.load_game(name) | ||
| md = munchausen_mirror_descent.MunchausenMirrorDescent( | ||
| game, value.TabularValueFunction(game)) | ||
| for _ in range(10): | ||
| md.iteration() | ||
| md_policy = md.get_policy() | ||
| nash_conv_md = nash_conv.NashConv(game, md_policy) | ||
|
|
||
| self.assertAlmostEqual(nash_conv_md.nash_conv(), 2.27366, places=5) | ||
|
|
||
|
|
||
| if __name__ == '__main__': | ||
| absltest.main() |