MCPcopy Create free account
hub / github.com/lazyprogrammer/machine_learning_examples / windy_grid

Function windy_grid

rl/grid_world.py:199–255  ·  view source on GitHub ↗
()

Source from the content-addressed store, hash-verified

197
198
199def windy_grid():
200 g = WindyGrid(3, 4, (2, 0))
201 rewards = {(0, 3): 1, (1, 3): -1}
202 actions = {
203 (0, 0): ('D', 'R'),
204 (0, 1): ('L', 'R'),
205 (0, 2): ('L', 'D', 'R'),
206 (1, 0): ('U', 'D'),
207 (1, 2): ('U', 'D', 'R'),
208 (2, 0): ('U', 'R'),
209 (2, 1): ('L', 'R'),
210 (2, 2): ('L', 'R', 'U'),
211 (2, 3): ('L', 'U'),
212 }
213
214 # p(s' | s, a) represented as:
215 # KEY: (s, a) --> VALUE: {s': p(s' | s, a)}
216 probs = {
217 ((2, 0), 'U'): {(1, 0): 1.0},
218 ((2, 0), 'D'): {(2, 0): 1.0},
219 ((2, 0), 'L'): {(2, 0): 1.0},
220 ((2, 0), 'R'): {(2, 1): 1.0},
221 ((1, 0), 'U'): {(0, 0): 1.0},
222 ((1, 0), 'D'): {(2, 0): 1.0},
223 ((1, 0), 'L'): {(1, 0): 1.0},
224 ((1, 0), 'R'): {(1, 0): 1.0},
225 ((0, 0), 'U'): {(0, 0): 1.0},
226 ((0, 0), 'D'): {(1, 0): 1.0},
227 ((0, 0), 'L'): {(0, 0): 1.0},
228 ((0, 0), 'R'): {(0, 1): 1.0},
229 ((0, 1), 'U'): {(0, 1): 1.0},
230 ((0, 1), 'D'): {(0, 1): 1.0},
231 ((0, 1), 'L'): {(0, 0): 1.0},
232 ((0, 1), 'R'): {(0, 2): 1.0},
233 ((0, 2), 'U'): {(0, 2): 1.0},
234 ((0, 2), 'D'): {(1, 2): 1.0},
235 ((0, 2), 'L'): {(0, 1): 1.0},
236 ((0, 2), 'R'): {(0, 3): 1.0},
237 ((2, 1), 'U'): {(2, 1): 1.0},
238 ((2, 1), 'D'): {(2, 1): 1.0},
239 ((2, 1), 'L'): {(2, 0): 1.0},
240 ((2, 1), 'R'): {(2, 2): 1.0},
241 ((2, 2), 'U'): {(1, 2): 1.0},
242 ((2, 2), 'D'): {(2, 2): 1.0},
243 ((2, 2), 'L'): {(2, 1): 1.0},
244 ((2, 2), 'R'): {(2, 3): 1.0},
245 ((2, 3), 'U'): {(1, 3): 1.0},
246 ((2, 3), 'D'): {(2, 3): 1.0},
247 ((2, 3), 'L'): {(2, 2): 1.0},
248 ((2, 3), 'R'): {(2, 3): 1.0},
249 ((1, 2), 'U'): {(0, 2): 0.5, (1, 3): 0.5},
250 ((1, 2), 'D'): {(2, 2): 1.0},
251 ((1, 2), 'L'): {(1, 2): 1.0},
252 ((1, 2), 'R'): {(1, 3): 1.0},
253 }
254 g.set(rewards, actions, probs)
255 return g
256

Calls 2

setMethod · 0.95
WindyGridClass · 0.85

Tested by

no test coverage detected