forked from FlorianMarquardt/machine-learning-for-physicists
-
Notifications
You must be signed in to change notification settings - Fork 0
/
08_tutorial_MazeDeepPolicyGradient.py
934 lines (726 loc) · 35.8 KB
/
08_tutorial_MazeDeepPolicyGradient.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
# coding: utf-8
# # Reinforcement Learning: Using "Deep" Policy Gradient
#
# This uses a (possibly deep) neural network to turn the image of a grid-world (maybe a maze) into action probabilities.
# Example code for the lecture series "Machine Learning for Physicists" by Florian Marquardt
#
# Lecture 8, Tutorial (this is discussed in session 8)
#
# See https://machine-learning-for-physicists.org and the current course website linked there!
#
# This notebook is distributed under the Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license:
#
# https://creativecommons.org/licenses/by-sa/4.0/
# This notebook shows how to:
# - use neural-network-based policy gradient reinforcement learning
#
# It is an extension of the tutorial in session 7, where table-based policy gradient was employed to solve a fixed maze (the same maze in every trial; the state space is the space of all positions).
# Note: At the moment, this only successfully (sometimes) finds the best strategy in an empty playing field (no maze), going from a fixed initial position to a fixed treasure chest. Still, the input are images (of the maze, the robot, and the map containing the location of the chest), instead of a table; so a convolutional network is used to transform that input into a policy.
# In[1]:
from tensorflow.keras import Sequential
from tensorflow.keras import Model
from tensorflow.keras.layers import Dense, Conv2D, AveragePooling2D, Flatten
from tensorflow.keras import optimizers # to choose more advanced optimizers like 'adam'
import numpy as np
import matplotlib.pyplot as plt
import matplotlib
matplotlib.rcParams['figure.dpi']=300 # highres display
# for subplots within subplots:
from matplotlib import gridspec
from IPython.display import clear_output
from time import sleep
# ## Maze generation algorithm
# In[2]:
# Maze generation algorithm from wikipedia
# the code was removed in January 2020, but you can still
# access it under this link:
# https://en.wikipedia.org/w/index.php?title=Maze_generation_algorithm&oldid=930153705
def maze(width=81, height=51, complexity=.75, density=.75):
# Only odd shapes
shape = ((height // 2) * 2 + 1, (width // 2) * 2 + 1)
# Adjust complexity and density relative to maze size
complexity = int(complexity * (5 * (shape[0] + shape[1]))) # number of components
density = int(density * ((shape[0] // 2) * (shape[1] // 2))) # size of components
# Build actual maze
Z = np.zeros(shape, dtype=bool)
# Fill borders
Z[0, :] = Z[-1, :] = 1
Z[:, 0] = Z[:, -1] = 1
# Make aisles
for i in range(density):
x, y = np.random.randint(0, shape[1] // 2) * 2, np.random.randint(0, shape[0] // 2) * 2 # pick a random position
Z[y, x] = 1
for j in range(complexity):
neighbours = []
if x > 1: neighbours.append((y, x - 2))
if x < shape[1] - 2: neighbours.append((y, x + 2))
if y > 1: neighbours.append((y - 2, x))
if y < shape[0] - 2: neighbours.append((y + 2, x))
if len(neighbours):
y_,x_ = neighbours[np.random.randint(0, len(neighbours) - 1)]
if Z[y_, x_] == 0:
Z[y_, x_] = 1
Z[y_ + (y - y_) // 2, x_ + (x - x_) // 2] = 1
x, y = x_, y_
return Z
# ## Now: policy gradient
# $$\delta\theta = \eta R \sum_{s,a} {\partial \over \partial \theta} \ln \pi_{\theta}(s,a)$$
#
# Here $\theta$ stands for the parameters controlling the policy probabilities $\pi_{\theta}(s,a)$, s is the state, a the action, and the sum runs over all state-action pairs that were encountered in the given trajectory, which led to an overall return (sum of rewards) R.
#
# Now $\pi_{\theta}(s,a)$ will be produced by a neural network. The state s (which is the input to the neural network) will be a picture of the maze (including the location of the chests and the location of the robot, in different channels). The actions a are just the four directions, as usual, and therefore the output of the neural network has four neurons (and the last layer should be softmax, to make sure these four neurons represent a probability distribution).
#
# To implement the policy gradient update, as explained in the lecture, one just needs to use categorical cross-entropy as the cost function, and provide as the 'correct' answer for each time step: (0,R,0,0), where the return R for the full trajectory is inserted at the location of the action (=direction) actually taken in this time step. This will re-inforce that action, according to the return.
#
# ## The deep policy gradient algorithm: slow version (does not train in parallel on many mazes, just one maze at a time)
#
# In 169 lines of code, including visualization and comments.
#
# This is provided here because it is as close as possible to the table-based policy gradient employed in the previous notebook!
# (I haven't gotten it to be successful in this version, although it runs...)
#
# In[5]:
# an empty playground... (even that turns out to be difficult enough currently)
def empty_maze(width,height):
mymaze=np.zeros([width,height],dtype='int')
mymaze[:,0]=1
mymaze[:,-1]=1
mymaze[0,:]=1
mymaze[-1,:]=1
return(mymaze)
# In[6]:
# full policy gradient RL for picking up 'treasure chests'
# in automatically generated mazes; 2020 by F.M.
# this one uses deep neural networks
M=5 # the size of the world: M x M
eta=0.01 # the learning rate
num_chests=1 # maximum number of treasure chests
# define the policy neural network
num_channels=10 # number of channels used in convolutional layers
Policy=Sequential()
Policy.add(Conv2D(num_channels,kernel_size=5,input_shape=(M,M,3),
activation="relu",padding="same"))
# needed to indicate input shape (at least: number of channels)
# 3 input channels: 0=walls, 1=chests, 2=robot
Policy.add(Conv2D(num_channels,kernel_size=5,activation="relu",padding="same"))
Policy.add(Conv2D(num_channels,kernel_size=5,activation="relu",padding="same"))
# now: one giant down-sampling step, to a 1x1 image
Policy.add(AveragePooling2D(pool_size=(M,M),padding="same"))
# now go from convolutional to dense:
Policy.add(Flatten(input_shape=(1,1,num_channels)))
# one last computation step:
Policy.add(Dense(20,activation="relu"))
Policy.add(Dense(4,activation="softmax"))
# the output layer with the probabilities
# we have to use categorical cross entropy for the
# policy gradient update rule!
# also, we should not use adam (which tries to be adaptive),
# but just plain old simple stochastic gradient descent,
# if we want the policy gradient update to be correctly
# implemented!
Policy.compile(loss='categorical_crossentropy',
optimizer=optimizers.SGD(learning_rate=eta))
# In[8]:
# NOW: the actual training loop
# the four directions of motion [delta_jx,delta_jy]:
directions=np.array([[0,1],[0,-1],[1,0],[-1,0]])
# steps inside one trajectory
nsteps=5
# total number of trials, i.e. trajectories
ntrials=100
skipsteps=5 # don't plot every trial
# storing all the state/action pairs of the current trajectory
states=np.zeros([nsteps,2], dtype='int')
actions=np.zeros(nsteps, dtype='int')
# storing all the returns, for all trials:
Returns=np.zeros(ntrials)
# store cost function for the training step (out of curiosity)
costs=np.zeros(ntrials)
# try many trajectories:
for trial in range(ntrials):
# make a new maze (a new one in each trial!)
world=np.array(empty_maze(width=M,height=M),dtype='int')
# a map of rewards (the 'chests' are here!)
reward=np.zeros([M,M])
# random selection of reward sites (treasure chests)
for n in range(num_chests):
jx_target,jy_target=np.random.randint(M,size=2)
if world[jx_target,jy_target]==0: # empty, keep it!
reward[jx_target,jy_target]+=1
# set return to zero for this trajectory:
R=0
# initial position:
jx,jy=1,1
# prepare an input image for the network
input_image=np.zeros([1,M,M,3]) # first index would be batchsize
input_image[0,:,:,0]=world
# we can set this now, since it will not change during this trial
# we need to store all the input images
# for a whole trajectory, for convenience
# so we can afterwards make the training step
# on this set of images as one batch!
input_images=np.zeros([nsteps,M,M,3])
# go through all time steps
for t in range(nsteps):
# Obtain the policy prediction given the current
# situation:
# the maze map is already stored
# inside input_image[0,:,:,0]
# but the treasure map and
# the robot position need to be updated
input_image[0,:,:,1]=reward # current treasure map!
input_image[0,jx,jy,2]=1 # indicate position of robot!
# now: evaluate policy network:
policy_p=Policy.predict_on_batch(input_image)[0]
# now policy_p is an array of 4 probabilities
# [0] was needed to get rid of first index, which
# would be the batchsize (but we have a batchsize of 1)
# make a random step, according to the policy distribution
p=np.random.uniform()
cumulative_distribution=np.cumsum(policy_p)
for pick in range(4):
if p<cumulative_distribution[pick]:
break
# record the move
states[t,0]=jx
states[t,1]=jy
actions[t]=pick
input_images[t,:,:,:]=input_image[0,:,:,:] # store for later
# now make the move
jx_new,jy_new=np.array([jx,jy])+directions[pick]
# really make it if there is no wall:
if world[jx_new,jy_new]==0: # is empty, can move!
input_image[0,jx,jy,2]=0 # delete robot position in map
jx,jy=jx_new,jy_new
# get a reward if on a treasure chest!
r=reward[jx,jy]
if r>0:
reward[jx,jy]-=1 # delete treasure!
R+=r
# store the return
Returns[trial]=R
# use policy gradient update rule to adjust
# probabilities!
# make an array for the target policy distributions
# for all time steps
target_distributions=np.zeros([nsteps,4])
for t in range(nsteps): # go through the trajectory again
a=actions[t] # remember the action taken at step t
target_distributions[t,a]=R # reinforce that action!
costs[trial]=Policy.train_on_batch(input_images,target_distributions)
# visualize!
if trial%skipsteps==0 or trial==ntrials-1:
# show what's happened in this trajectory
clear_output(wait=True)
fig,ax=plt.subplots(ncols=2,nrows=1,figsize=(8,4))
ax[0].plot(Returns) # all the returns, in all trials
ax[0].set_title("Return vs. trial")
picture=np.zeros([M,M,3]) # last index: red/green/blue
picture[:,:,0]=world # walls are red
for j in range(nsteps): # highlight trajectory
picture[states[j,0],states[j,1],1]=0.5*(1.0+(1.0*j)/nsteps)
# put a bright pixel at the positions visited
picture[:,:,2]+=0.5*reward # highlight the target sites!
picture[:,:,0]+=0.5*reward
# show picture (transpose is needed because
# otherwise the first coordinate jx is plotted upwards,
# not to the right)
ax[1].imshow(np.transpose(picture,[1,0,2]),origin='lower')
ax[1].axis('off')
#ax[1,0].imshow(np.transpose(policy[:,:,0]),origin='lower')
#ax[1,0].axis('off')
#ax[1,0].set_title("prob(move up)")
#ax[1,1].imshow(np.transpose(policy[:,:,2]),origin='lower')
#ax[1,1].set_title("prob(move right)")
#ax[1,1].axis('off')
plt.show()
# ## The deep policy gradient algorithm: 'fast' version (trains in parallel on many mazes, which accelerates the policy network evaluation, since this can be done on a whole batch of input images)
#
# In[3]:
# an empty playground... (even that turns out to be difficult enough currently)
def empty_maze(width,height):
mymaze=np.zeros([width,height],dtype='int')
mymaze[:,0]=1
mymaze[:,-1]=1
mymaze[0,:]=1
mymaze[-1,:]=1
return(mymaze)
# In[4]:
def run_one_trial(Policy,M,nsteps,input_image,input_images,reward,jx,jy,
actions,position,world,allsamples,batchsize,
directions,
delete_treasure=True):
"""
Run one trial starting at position jx,jy, storing returns in R, storing
actions and position, using input_image and world.
This will return R,input_images.
"""
# set return to zero for this trajectory:
R=np.zeros([batchsize])
# go through all time steps
for t in range(nsteps):
# Obtain the policy prediction given the current
# situation:
# the maze map is already stored
# inside input_image[0,:,:,0]
# but the treasure map and
# the robot position need to be updated
input_image[:,:,:,1]=reward # current treasure map!
input_image[allsamples,jx,jy,2]=1
# last line indicates position of robot(s)! (for whole batch)
# note: this uses advanced numpy indexing, so we use
# three equal-length integer arrays a,b,c to do the
# equivalent of
# for k in range(n):
# input_image[a[k],b[k],c[k],2]=1
#
# now: evaluate policy network on the whole batch:
# this is where we gain efficiency!
policy_p=Policy.predict_on_batch(input_image)
#print(policy_p)
if np.any(policy_p>1.001):
print("Oops, policy out of range:",policy_p)
# now policy_p is an array of [batchsize,4] probabilities
# make a random step, according to the policy distribution
p=np.random.uniform(size=batchsize)
cumulative_distribution=np.cumsum(policy_p,axis=1) # note axis argument
pick=np.argmax(cumulative_distribution>p[:,None],axis=1)
# will give the index of the first entry that exceeds p, which is
# exactly what we need (pick will still be a 1D array of 'batchsize' length)
# record the move
#print(pick[0],jx[0],jy[0])
actions[:,t]=pick
position[:,t,0]=jx
position[:,t,1]=jy
input_images[:,t,:,:,:]=np.copy(input_image) # store for later
# now make the move
jx_new,jy_new=jx+directions[pick][:,0],jy+directions[pick][:,1]
# really make it if there is no wall:
# again, we do this for all samples in parallel,
# thanks to advanced numpy array indexing
can_move=np.array(world[allsamples,jx_new,jy_new]==0,dtype='int') # 1 if can move
input_image[allsamples,jx,jy,2]-=can_move # delete old position only if we can move
jx,jy=jx*(1-can_move)+jx_new*can_move,jy*(1-can_move)+jy_new*can_move
#print("-->",jx[0],jy[0])
# get a reward if on a treasure chest!
r=reward[allsamples,jx,jy] # will be array of batchsize length
if delete_treasure:
reward[allsamples,jx,jy]-=1*(r>0) # delete treasure!
R+=r
return(R)
# In[5]:
def prepare_batch(batchsize,world,M,num_chests,jx,jy,full_maze=False):
"""
prepare one batch, return:
reward,input_image
Arrays world and jx,jy need already to exist, but will
be filled with values.
"""
# a map of rewards (the 'chests' are here!)
reward=np.zeros([batchsize,M,M],dtype='float')
for sample in range(batchsize):
# make a new maze (a new one in each trial!)
if full_maze:
world[sample,:,:]=np.array(maze(width=M,height=M),dtype='int')
else:
world[sample,:,:]=np.array(empty_maze(width=M,height=M),dtype='int')
# random selection of reward sites (treasure chests)
for n in range(num_chests):
while True:
jx_target,jy_target=np.random.randint(low=2,high=M-2,size=2) # avoid close to boundary!
if world[sample,jx_target,jy_target]==0: # empty, keep it!
reward[sample,jx_target,jy_target]+=1
break
# pick random initial position:
for sample in range(batchsize):
while True:
jx_try,jy_try=np.random.randint(low=2,high=M-2,size=2) # avoid close to boundary!
if world[sample,jx_try,jy_try]==0:
jx[sample],jy[sample]=jx_try,jy_try
break
# prepare an input image for the network
input_image=np.zeros([batchsize,M,M,3],dtype='float')
input_image[:,:,:,0]=world
# we can set this now, since it will not change during this trial
return(reward,input_image)
# In[9]:
def try_pos(x,y,Policy,M,world,reward):
try_image=np.zeros([1,M,M,3])
try_image[0,x,y,2]=1.0
try_image[0,:,:,0]=world[0,:,:]
try_image[0,:,:,1]=reward[0,:,:]
return( Policy.predict_on_batch(try_image)[0] )
def plot_try_pos(M,num_actions,Policy,world,reward,
directions,ax=None,target=None):
P=np.zeros([M,M,num_actions])
for x in range(M):
for y in range(M):
P[x,y,:]=try_pos(x,y,Policy,M,world,reward)
if ax is None:
fig,ax=plt.subplots(ncols=num_actions,nrows=1)
ax_was_none=True
else:
ax_was_none=False
for n in range(num_actions):
ax[n].imshow(np.transpose(P[:,:,n]+world[0,:,:]),origin='lower',vmin=0.0,vmax=1.0)
ax[n].axis('off')
ax[n].set_title(str(directions[n]))
if target is not None:
ax[n].scatter([target[0]],[target[1]],c="orange")
if np.any(np.abs(np.sum(P,axis=2)-1.0)>0.01):
print("WARNING: Probabilities do not sum up to 1!", np.sum(P,axis=1))
if ax_was_none:
plt.show()
# In[10]:
def run_RL(M=7,eta=0.0001,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=50,test_batchsize=5,
nsteps=15,ntrials=1000,skipsteps=20,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=False,choose_random_positions_for_single_maze=False,
place_chest=None,kernel_size=5,num_channels=5,
full_maze=False):
"""
Run the full policy gradient RL algorithm, for a number of trials.
"""
global orig_test_reward,orig_test_world
# first define the policy neural network
Policy=Sequential()
Policy.add(Conv2D(num_channels,kernel_size=kernel_size,input_shape=(M,M,3),
activation="elu",padding="same"))
Policy.add(Conv2D(num_channels,kernel_size=kernel_size,
activation="elu",padding="same"))
Policy.add(Conv2D(num_channels,kernel_size=kernel_size,
activation="elu",padding="same"))
Policy.add(Conv2D(num_channels,kernel_size=kernel_size,
activation="elu",padding="same"))
Policy.add(Flatten())
Policy.add(Dense(num_actions,activation="softmax"))
# the output layer with the probabilities
# we have to use categorical cross entropy for the
# policy gradient update rule!
# also, we should not use adam (which tries to be adaptive),
# but just plain old simple stochastic gradient descent,
# if we want the policy gradient update to be correctly
# implemented!
if try_adam:
Policy.compile(loss='categorical_crossentropy',
optimizer=optimizers.Adam(learning_rate=eta))
else:
Policy.compile(loss='categorical_crossentropy',
optimizer=optimizers.SGD(learning_rate=eta,clipnorm=1.0))
# NOW: the setup
# the four directions of motion [delta_jx,delta_jy]:
directions=np.array([[1,0],[0,-1],[0,1],[-1,0],[0,0]])
# storing all the actions of the current trajectory
actions=np.zeros([batchsize,nsteps], dtype='int')
position=np.zeros([batchsize,nsteps,2], dtype='int')
# same for test batch (which is smaller):
test_actions=np.zeros([test_batchsize,nsteps], dtype='int')
test_position=np.zeros([test_batchsize,nsteps,2], dtype='int')
# we need to store all the input images
# for a whole trajectory, for convenience
# so we can afterwards make the training step
# on this set of images as one batch!
input_images=np.zeros([batchsize,nsteps,M,M,3],dtype='float')
test_input_images=np.zeros([test_batchsize,nsteps,M,M,3],dtype='float')
# the current position, for all samples of the batch in parallel:
jx,jy=np.zeros([batchsize],dtype='int'),np.zeros([batchsize],dtype='int')
test_jx,test_jy=np.zeros([test_batchsize],dtype='int'),np.zeros([test_batchsize],dtype='int')
# storing all the returns, for all trials:
Returns=np.zeros([batchsize,ntrials])
# the maze maps, for all samples in parallel:
world=np.zeros([batchsize,M,M],dtype='int')
test_world=np.zeros([test_batchsize,M,M],dtype='int')
# store cost function for the training step (out of curiosity)
costs=np.zeros(ntrials)
# useful for some advanced array indexing:
allsamples=np.arange(0,batchsize)
test_allsamples=np.arange(0,test_batchsize)
# generate test mazes on which the strategy will be
# repeatedly illustrated:
test_reward,test_input_image=prepare_batch(batchsize=test_batchsize,
world=test_world,M=M,
num_chests=num_chests,jx=test_jx,jy=test_jy,
full_maze=full_maze)
orig_test_reward=np.copy(test_reward)
orig_test_world=np.copy(test_world)
# special case: a single maze (not a fresh maze for every sample)
if single_maze:
if place_chest is not None:
test_reward[0,:,:]=0
test_reward[0,place_chest[0],place_chest[1]]=1.0
for n in range(1,test_batchsize):
test_reward[n,:,:]=test_reward[0,:,:]
test_input_image[n,:,:,:]=test_input_image[0,:,:,:]
test_world[n,:,:]=test_world[0,:,:]
reward,input_image=prepare_batch(batchsize=batchsize,
world=world,M=M,num_chests=num_chests,jx=jx,jy=jy,
full_maze=full_maze)
for n in range(batchsize):
reward[n,:,:]=test_reward[0,:,:]
input_image[n,:,:,:]=test_input_image[0,:,:,:]
world[n,:,:]=test_world[0,:,:]
w=np.where(test_reward[0,:,:]>0.5)
chest_x=w[0][0] # position of treasure chest
chest_y=w[1][0]
costs=np.zeros(ntrials)
# try many trajectories:
for trial in range(ntrials):
# prepare the mazes of the batch (and starting positions, and treasure distribution):
if not single_maze: # generate a fresh maze for every sample in the batch
reward,input_image=prepare_batch(batchsize=batchsize,
world=world,M=M,num_chests=num_chests,jx=jx,jy=jy,
full_maze=full_maze)
else: # keep the one single maze
for sample in range(batchsize):
jx[sample],jy[sample]=1,1
input_image[sample,:,:,2]=0
input_image[sample,jx[sample],jy[sample],2]=1
# the following good for starting from random positions, but the same maze
if choose_random_positions_for_single_maze:
while True:
jx_try,jy_try=np.random.randint(M,size=2)
jx_try,jy_try=np.random.randint(M,size=2)
if world[sample,jx_try,jy_try]==0:
jx[sample],jy[sample]=jx_try,jy_try
input_image[sample,:,:,2]=0
input_image[sample,jx[sample],jy[sample],2]=1 # place robot
break
# run a single full trial
# as usual, all arrays are passed by reference and will be filled
# with new values in this routine
R=run_one_trial(Policy,M,nsteps,input_image=input_image,
input_images=input_images,
reward=reward, jx=jx, jy=jy,
actions=actions,position=position,world=world,
allsamples=allsamples,batchsize=batchsize,
directions=directions,
delete_treasure=delete_treasure)
# store the return
Returns[:,trial]=R
# w=np.argmax(R)
# for t in range(nsteps):
# print(position[w,t,0],position[w,t,1],end=" | ")
# print("R = ",R[w])
# use policy gradient update rule to adjust
# probabilities!
# first: make an array for the target policy distributions
# for all time steps (those that contain 'R' in the slot
# of the action that was actually taken!)
target_distributions=np.zeros([batchsize,nsteps,num_actions])
for t in range(nsteps): # go through the trajectory again
a=actions[:,t] # remember the action taken at step t
target_distributions[allsamples,t,a]=R # reinforce that action!
costs[trial]=Policy.train_on_batch(np.reshape(input_images,[batchsize*nsteps,M,M,3]),
np.reshape(target_distributions,[batchsize*nsteps,num_actions]))
# we needed the reshape to make sure the input is still
# of shape [total_batchsize,M,M,3]
# ...only now total_batchsize is larger!
# this will run through the update algorithm a larger number
# of samples in parallel, which again is very efficient!
# visualize!
if do_visualize and trial%skipsteps==0 or trial==ntrials-1:
if do_show_test_batch:
current_test_reward=np.copy(test_reward) # avoid changes in the test_reward!
R_test=run_one_trial(Policy,M,nsteps,input_image=test_input_image,
input_images=test_input_images,
reward=current_test_reward, jx=test_jx, jy=test_jy,
actions=test_actions,position=test_position,
world=test_world,allsamples=test_allsamples,
batchsize=test_batchsize,directions=directions,
delete_treasure=delete_treasure)
clear_output(wait=True)
fig=plt.figure(constrained_layout=True,
figsize=(num_actions,3))
gs=fig.add_gridspec(ncols=num_actions,nrows=3)
returns_plot=fig.add_subplot(gs[0:2,:])
# show all the returns so far (averaged over the batch for each trial)
returns_plot.plot(np.average(Returns,axis=0))
# all the returns, in all trials (averaged over batch!)
returns_plot.set_title("Return vs trial ("+str(batchsize)+" trajs/trial)")
if do_show_test_batch:
n_test_plots=test_batchsize
else:
n_test_plots=num_actions
test_plot=[]
for n in range(n_test_plots):
test_plot.append(fig.add_subplot(gs[2,n]))
if not do_show_test_batch:
plot_try_pos(M,num_actions,Policy,orig_test_world,orig_test_reward,
directions,ax=test_plot,target=[chest_x,chest_y])
else:
# show what's happened in this test trial
for n in range(test_batchsize):
# draw the trajectory of current batch sample 0!
picture=np.zeros([M,M,3]) # last index: red/green/blue
picture[:,:,0]=test_world[n,:,:] # walls are red
for j in range(nsteps): # highlight trajectory
picture[test_position[n,j,0],test_position[n,j,1],1]=0.5*(1.0+(1.0*j)/nsteps)
# put a bright pixel at the positions visited
# highlight the target sites!
picture[:,:,2]+=1*(test_reward[n,:,:]>0)
# show picture (transpose is needed because
# otherwise the first coordinate jx is plotted upwards,
# not to the right)
test_plot[n].imshow(np.transpose(picture,[1,0,2]),origin='lower')
test_plot[n].axis('off')
plt.show()
# ## Tests (in an empty maze: just trying to find the right path from a fixed starting point to a fixed treasure chest...)
# In[381]:
run_RL(M=5,eta=0.01,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=1000,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=False,choose_random_positions_for_single_maze=False)
# In[383]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.01,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=False,choose_random_positions_for_single_maze=False,
place_chest=[2,2])
# In[384]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.01,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=False,choose_random_positions_for_single_maze=False,
place_chest=[2,2])
# In[385]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.01,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=False,choose_random_positions_for_single_maze=False,
place_chest=[2,2])
# In[387]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.01,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=False,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# In[389]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.01,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=False,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# In[390]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.01,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=False,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# ## Now: using adam
# In[392]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.001,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# In[393]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.001,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# In[394]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.0005,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# In[395]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.0002,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# In[396]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.0002,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# In[425]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.0002,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# In[426]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.0001,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=1000,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# In[427]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.0001,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=1000,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# In[11]:
# the same again, with the same position of the chest:
run_RL(M=5,eta=0.0001,num_chests=1,delete_treasure=False,
single_maze=True,batchsize=10,test_batchsize=5,
nsteps=7,ntrials=1000,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=[2,2], kernel_size=3)
# # Tutorial: Explore parameters to get better performance!
#
# These would include:
# - the learning rate eta
# - the batchsize
# - whether to choose adam or stochastic gradient descent
# - the size M of the maze and the number of time steps (nsteps) in a trajectory
# - the layout of the neural network: kernel_size, num_channels, or possibly more changes in the network structure itself (inside run_RL)
# ## Train on arbitrary mazes (does not yet work)
# In[420]:
# for fun, try arbitrary mazes (arbitrary position of the chest, random initial pos.)
# this does not yet really work...probably a question of finding
# the right parameters...
run_RL(M=9,eta=0.0005,num_chests=1,delete_treasure=False,
single_maze=False ,full_maze=True,
batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=None, kernel_size=3)
# In[421]:
# for fun, try arbitrary mazes (arbitrary position of the chest, random initial pos.)
# this does not yet really work...probably a question of finding
# the right parameters...
run_RL(M=9,eta=0.0005,num_chests=1,delete_treasure=False,
single_maze=False ,full_maze=True,
batchsize=10,test_batchsize=5,
nsteps=7,ntrials=500,skipsteps=5,do_visualize=True,
num_actions=5,do_show_test_batch=False,
try_adam=True,choose_random_positions_for_single_maze=False,
place_chest=None, kernel_size=3)