-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathOldCode.py
354 lines (311 loc) · 13.5 KB
/
OldCode.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
class GradientDescent(object):
def alpha( self, t ):
raise NotImplementedError, "Cannot call abstract method"
theta_0=None
Threshold=None
T = -1
sign = None
def run( self, f_grad, f_proj=None, b_norm=False ): #grad is a function of theta
theta = self.theta_0.copy()
best_theta = theta.copy()
best_norm = float("inf")
best_iter = 0
t=0
while True:#Do...while loop
t+=1
DeltaTheta = f_grad( theta )
current_norm = norm( DeltaTheta )
if b_norm and current_norm > 0.:
DeltaTheta /= norm( DeltaTheta )
theta = theta + self.sign * self.alpha( t )*DeltaTheta
if f_proj:
theta = f_proj( theta )
print "Norme du gradient : "+str(current_norm)+", pas : "+str(self.alpha(t))+", iteration : "+str(t)
if current_norm < best_norm:
best_norm = current_norm
best_theta = theta.copy()
best_iter = t
if norm < self.Threshold or (self.T != -1 and t >= self.T):
break
print "Gradient de norme : "+str(best_norm)+", a l'iteration : "+str(best_iter)
return best_theta
class StructuredClassifier(GradientDescent):
sign=-1.
Threshold=0.1 #Sensible default
T=40 #Sensible default
phi=None
phi_xy=None
inputs=None
labels=None
label_set=None
dic_data={}
x_dim=None
def alpha(self, t):
return 3./(t+1)#Sensible default
def __init__(self, data, x_dim, phi, phi_dim, Y):
self.x_dim=x_dim
self.inputs = data[:,:-1]
shape = list(data.shape)
shape[-1] = 1
self.labels = data[:,-1].reshape(shape)
self.phi=phi
self.label_set = Y
self.theta_0 = zeros((phi_dim,1))
self.phi_xy = self.phi(data)
for x,y in zip(self.inputs,self.labels):
self.dic_data[str(x)] = y
print self.inputs.shape
def structure(self, xy):
return 0. if xy[-1] == self.dic_data[str(xy[:-1])] else 1.
def structured_decision(self, theta):
def decision( x ):
score = lambda xy: dot(theta.transpose(),self.phi(xy)) + self.structure(xy)
input_label_couples = [hstack([x,y]) for y in self.label_set]
best_label = argmax(input_label_couples, score)[-1]
return best_label
vdecision = non_scalar_vectorize(decision, (self.x_dim,), (1,1))
return lambda x: vdecision(x).reshape(x.shape[:-1]+(1,))
def gradient(self, theta):
classif_rule = self.structured_decision(theta)
y_star = classif_rule(self.inputs)
#print "Gradient : "+str(y_star)
#print str(self.labels)
phi_star = self.phi(hstack([self.inputs,y_star]))
return mean(phi_star-self.phi_xy,axis=0)
def run(self):
f_grad = lambda theta: self.gradient(theta)
theta = super(StructuredClassifier,self).run( f_grad, b_norm=True)
classif_rule = greedy_policy(theta,self.phi,self.label_set)
return classif_rule,theta
def least_squares_regressor(data, psi, x_dim, _lambda=0.1):
"""Return a function that given a x, will output a y based on a least square regression of the data provided in argument"""
X = data[:,0:x_dim]
Y = data[:,x_dim:]
psi_X = squeeze(psi(X))
#print psi_X
omega = dot(dot(inv(dot(psi_X.transpose(),psi_X)+_lambda*identity(psi_X.shape[1])),psi_X.transpose()),Y)
#omega = dot(inv(dot(psi_X.transpose(),psi_X)+_lambda*identity(psi_X.shape[1])),psi_X.transpose())
#omega = inv(dot(psi_X.transpose(),psi_X)+_lambda*identity(psi_X.shape[1]))
#omega = dot(psi_X.transpose(),psi_X)+_lambda*identity(psi_X.shape[1])
#print psi_X.shape
#omega = dot(psi_X.transpose(),psi_X)
regression = lambda x:dot(omega.transpose(),psi(x))
return regression
def CSI(data, classifier, regressor, A, s_dim=1, gamma=0.9):#FIXME7: does not work for vectorial actions
column_shape = (len(data),1)
s = data[:,0:s_dim]
a = data[:,s_dim:s_dim+1].reshape(column_shape)
sa = data[:,0:s_dim+1]
s_dash = data[:,s_dim+1:s_dim+1+s_dim]
pi_c,q = classifier(hstack([s,a]))
a_dash = pi_c(s_dash).reshape(column_shape)
sa_dash = hstack([s_dash,a_dash])
hat_r = (q(sa)-gamma*q(sa_dash)).reshape(column_shape)
print "Shapes : s, a, s_dash, a_dash, sa, sa_dash, hat_r"+str([x.shape for x in [s,a,s_dash,a_dash,sa, sa_dash,hat_r]])
r_min = min(hat_r)-ones(column_shape)
regression_input_matrices = [hstack([s,action*ones(column_shape)]) for action in A]
def add_output_column( reg_mat ):
actions = reg_mat[:,-1].reshape(column_shape)
hat_r_bool_table = array(actions==a)
r_min_bool_table = array(hat_r_bool_table==False) #"not hat_r_bool_table" does not work as I expected
output_column = hat_r_bool_table*hat_r+r_min_bool_table*r_min
return hstack([reg_mat,output_column])
regression_matrix = vstack(map(add_output_column,regression_input_matrices))
return regressor( regression_matrix )
def get_structured_classifier_for_SCI(s_dim, phi, phi_dim, A):
"""Return a classifier function as expected by the CSI algorithm using the StructuredClassifier class"""
def classifier(data):
structured_classifier = StructuredClassifier(data, s_dim, phi, phi_dim, A)
classif_rule,omega_classif = structured_classifier.run()
score = lambda sa: squeeze(dot(omega_classif.transpose(),phi(sa)))
return classif_rule,score
return classifier
def get_least_squares_regressor_for_SCI(phi, sa_dim):
#ax=b,find x
def regressor(data):
a=squeeze(phi(data[:,0:sa_dim]))
b=data[:,sa_dim:]
print "Regressor : data, a and b shapes and s_dim: "+str(data.shape)+str(a.shape)+str(b.shape)+str(sa_dim)
x=lstsq(a,b)[0]
print "Regressor : x.shape : "+str(x.shape)
return lambda sa:dot(x.transpose(),phi(sa))
return lambda data: regressor(data)
#VRAC
states = rand(1000,2)*6-3
actions = policy(states)
data = hstack([states,actions])
classifier = get_structured_classifier_for_SCI(2, inverted_pendulum_phi, 30, ACTION_SPACE)
classif_rule,score = classifier(data)
data_0 = array([l for l in data if l[2]==0.])
data_1 = array([l for l in data if l[2]==1.])
data_2 = array([l for l in data if l[2]==2.])
inverted_pendulum_plot_policy(policy)
plot(data_0[:,0],data_0[:,1],ls='',marker='o')
plot(data_1[:,0],data_1[:,1],ls='',marker='o')
plot(data_2[:,0],data_2[:,1],ls='',marker='o')
inverted_pendulum_plot_policy( classif_rule )
two_args_score = lambda p,s: score(zip_stack(p,s,squeeze(classif_rule(zip_stack(p,s)))))
inverted_pendulum_plot(two_args_score)
reg_S = rand(1000,2)*2-1
reg_A = policy(reg_S)
reg_X = hstack([reg_S,reg_A])
test_omega = rand(30,1)-0.5
reg_Y = dot(test_omega.transpose(),inverted_pendulum_phi(reg_X)).reshape(1000,1)
regressor = get_least_squares_regressor_for_SCI(inverted_pendulum_phi, 3)
reg = regressor(hstack([reg_X,reg_Y]))
inverted_pendulum_plot_SAReward( reg, policy)
X = linspace(-pi,pi,30)
Y = X
X,Y = meshgrid(X,Y)
XY=zip_stack(X,Y)
A = squeeze(policy(XY))
B = zip_stack(X,Y,A)
Z = squeeze(reg(B))
from matplotlib import cm
fig = figure()
ax=Axes3D(fig)
ax.plot_surface(X,Y,Z, cmap=cm.jet,cstride=1,rstride=1)
ax.scatter(reg_X[:,0],reg_X[:,1],reg_Y,c=reg_Y)
show()
#Déroulage "à la main" de l'algo CSI, pour comprendre ce qui cloche
#def CSI(data, classifier, regressor, A, s_dim=1, gamma=0.9):#FIXME7: does not work for vectorial actions
#data = data_expert[:50,:]
data = rand(2000,2)*2*pi - pi
data_classif = hstack([data,policy(data)])
data_classif = hstack([data_classif,zeros((2000,2))])
for index, line in enumerate(data_classif):
data_classif[index,-2:] = inverted_pendulum_next_state(data_classif[index,:2],data_classif[index,2])
random_policy= lambda s:choice(ACTION_SPACE)
vrandom_policy = non_scalar_vectorize( random_policy, (2,), (1,1) )
pi_r = lambda state: vrandom_policy(state).reshape(state.shape[:-1]+(1,))
data_regress = hstack([data,pi_r(data)])
data_regress = hstack([data_regress,zeros((2000,2))])
for index, line in enumerate(data_regress):
data_regress[index,-2:] = inverted_pendulum_next_state(data_regress[index,:2],data_regress[index,2])
gamma=0.9
#classifier, regressor, deja definis
A = ACTION_SPACE
s_dim=2
phi=inverted_pendulum_phi
psi=inverted_pendulum_psi
#Algo
column_shape = (len(data_classif),1)
s = data_classif[:,0:s_dim]
a = data_classif[:,s_dim:s_dim+1].reshape(column_shape)
sa = data_classif[:,0:s_dim+1]
s_dash = data_classif[:,s_dim+1:s_dim+1+s_dim]
#from sklearn import svm
#clf = svm.SVC(C=1000., probability=True)
#clf.fit(squeeze(psi(s)), a)
#clf_predict= lambda state : clf.predict(squeeze(psi(state)))
#vpredict = non_scalar_vectorize( clf_predict, (2,), (1,1) )
#pi_c = lambda state: vpredict(state).reshape(state.shape[:-1]+(1,))
#clf_score = lambda sa : squeeze(clf.predict_proba(squeeze(psi(sa[:2]))))[sa[-1]]
#vscore = non_scalar_vectorize( clf_score,(3,),(1,1) )
#q = lambda sa: vscore(sa).reshape(sa.shape[:-1])
pi_c,q=policy,lambda sa: squeeze(dot(omega.transpose(),phi(sa)))
#pi_c,q = classifier(hstack([s,a]))
#Plots de la politique de l'expert, des données fournies par l'expert, de la politique du classifieur
inverted_pendulum_plot_policy(policy)
scatter(data_classif[:,0],data_classif[:,1],c=data_classif[:,2])
figure()
scatter(data_classif[:,0],data_classif[:,1],c=data_classif[:,2])
inverted_pendulum_plot_policy(pi_c)
##Plots de Q et de la fonction de score du classifieur et évaluation de la politique du classifieur
#phi=inverted_pendulum_phi
Q = lambda sa: squeeze(dot(omega.transpose(),phi(sa)))
Q_0 = lambda p,s:Q(zip_stack(p,s,0*ones(p.shape)))
Q_1 = lambda p,s:Q(zip_stack(p,s,1*ones(p.shape)))
Q_2 = lambda p,s:Q(zip_stack(p,s,2*ones(p.shape)))
q_0 = lambda p,s:q(zip_stack(p,s,0*ones(p.shape)))
q_1 = lambda p,s:q(zip_stack(p,s,1*ones(p.shape)))
q_2 = lambda p,s:q(zip_stack(p,s,2*ones(p.shape)))
inverted_pendulum_plot(Q_0)
inverted_pendulum_plot(Q_1)
inverted_pendulum_plot(Q_2)
inverted_pendulum_plot(q_0)
inverted_pendulum_plot(q_1)
inverted_pendulum_plot(q_2)
##FIXME: combien de fois le classifieur tombe-t-il ?
#On continue CSI
s = data_regress[:,0:s_dim]
a = data_regress[:,s_dim:s_dim+1].reshape(column_shape)
sa = data_regress[:,0:s_dim+1]
s_dash = data_regress[:,s_dim+1:s_dim+1+s_dim]
a_dash = pi_c(s_dash).reshape(column_shape)
sa_dash = hstack([s_dash,a_dash])
hat_r = (q(sa)-gamma*q(sa_dash)).reshape(column_shape)
r_min = min(hat_r)-1.*ones(column_shape)
#Plot des samples hat_r Pour chacune des 3 actions
sar = hstack([sa,hat_r])
for action in ACTION_SPACE:
sr = array([l for l in sar if l[2]==action])
axis([-pi,pi,-pi,pi])
scatter(sr[:,0],sr[:,1],s=20,c=sr[:,3], marker = 'o', cmap = cm.jet );
colorbar()
figure()
#On continue SCI
##Avec l'heuristique :
#regression_input_matrices = [hstack([s,action*ones(column_shape)]) for action in A]
#def add_output_column( reg_mat ):
# actions = reg_mat[:,-1].reshape(column_shape)
# hat_r_bool_table = array(actions==a)
# r_min_bool_table = array(hat_r_bool_table==False) #"not hat_r_bool_table" does not work as I expected
# output_column = hat_r_bool_table*hat_r+r_min_bool_table*r_min
# return hstack([reg_mat,output_column])
#regression_matrix = vstack(map(add_output_column,regression_input_matrices))
#On plotte les mêmes données que juste précedemment, mais avec l'heuristique en prime
#for action in ACTION_SPACE:
# sr = array([l for l in regression_matrix if l[2]==action])
# axis([-pi,pi,-pi,pi])
# scatter(sr[:,0],sr[:,1],s=20,c=sr[:,3], marker = 'o', cmap = cm.jet );
# colorbar()
# figure()
##Sans heuristique :
regression_matrix=sar
#On continue CSI
def regressor(data):
a=squeeze(phi(data[:,0:3]))
#a=data[:,0:3]
b=data[:,3:]
# print x.shape
#return lambda sa:dot(x,phi(sa))
##Trichons
def triche(sa):
s_dash = inverted_pendulum_next_state(sa[:2],sa[-1])
a_dash = pi_c(s_dash)
sa_dash=hstack([s_dash,a_dash])
return q(sa)-gamma*q(sa_dash)
vtriche=non_scalar_vectorize(triche, (3,), (1,1))
return lambda sa:vtriche(sa)
reg = regressor( regression_matrix )
CSI_reward = lambda sas: squeeze(reg(sas[:3]))
#On plotte les rewards en fonction de l'action
for action in ACTION_SPACE:
sr = array([l for l in regression_matrix if l[2]==action])
R = lambda p,s: squeeze( reg(zip_stack(p,s,action*ones(p.shape))))
pos = linspace(-pi,pi,30)
speed = linspace(-pi,pi,30)
pos,speed = meshgrid(pos,speed)
Z = R(pos,speed)
figure()
contourf(pos,speed,Z,50)
scatter(sr[:,0],sr[:,1],s=20,c=sr[:,3], marker = 'o', )#cmap = cm.jet );
clim(vmin=min(Z.reshape(-1)),vmax=max(Z.reshape(-1)))
colorbar()
def mean_reward(s,p):
actions = [a*ones(s.shape) for a in ACTION_SPACE]
matrices = [zip_stack(s,p,a) for a in actions]
return mean(array([squeeze(reg(m)) for m in matrices]), axis=0)
inverted_pendulum_plot(mean_reward)
def inverted_pendulum_expert_trace( reward ):
data = inverted_pendulum_random_trace(reward=reward)
policy,omega = lspi( data, s_dim=2,a_dim=1, A=ACTION_SPACE, phi=inverted_pendulum_phi, phi_dim=30, iterations_max=10 )
inverted_pendulum_plot(inverted_pendulum_V(omega))
two_args_pol = lambda p,s:squeeze(policy(zip_stack(p,s)))
inverted_pendulum_plot(two_args_pol,contour_levels=3)
return inverted_pendulum_trace( policy,run_length=EXPERT_RUN_LENGTH ),policy,omega
data_expert,policy,omega = inverted_pendulum_expert_trace(inverted_pendulum_reward)
#data_CSI,policy_CSI,omega_CSI = inverted_pendulum_expert_trace(CSI_reward)
CSI_falls_rate = - mean( data_CSI[:,5] )
print "Rate of falls for IRL controller "+str(CSI_falls_rate)