def estimate_log_evidence_pm_pendulum(load_path="./result_apply/",
save_path="./result_apply/",
loop_num=0,
i=5):
gather_sample.gather_real_world_sample(loop_num=loop_num)
X = np.loadtxt(load_path + 'real_world_samples_input.csv', delimiter=',')
Y = np.loadtxt(load_path + 'real_world_samples_output.csv', delimiter=',')
X = np.delete(X, i, 0)
Y = np.delete(Y, i, 0)
pm_class = pendulum_pm.PendulumPM(X, Y)
gather_sample.gather_real_world_sample(loop_num=loop_num)
X = np.loadtxt(load_path + 'real_world_samples_input.csv', delimiter=',')
Y = np.loadtxt(load_path + 'real_world_samples_output.csv', delimiter=',')
X = np.delete(X, i, 0)
Y = np.delete(Y, i, 0)
X2 = np.loadtxt(load_path + 'real_world_samples_input.csv', delimiter=',')
Y2 = np.loadtxt(load_path + 'real_world_samples_output.csv', delimiter=',')
print("prev_log_evidence =", pm_class.log_evidence())
print("debug_log_evidence =", pm_class.log_evidence_of_new_data(X, Y))
print("new_log_evidence =", pm_class.log_evidence_of_new_data(X2, Y2))
print("diff =",
pm_class.log_evidence_of_new_data(X2, Y2) - pm_class.log_evidence())
temp_data = np.array(
[pm_class.log_evidence(),
pm_class.log_evidence_of_new_data(X2, Y2)])
return (pm_class.log_evidence_of_new_data(X2, Y2) -
pm_class.log_evidence())
def estimate_log_evidence_pm_cartpole(load_path="./result_apply/",
save_path="./result_apply/",
loop_num=0):
gather_sample.gather_real_world_sample(loop_num=loop_num)
X = np.loadtxt(load_path+'real_world_samples_input.csv', delimiter=',')
Y = np.loadtxt(load_path+'real_world_samples_output.csv', delimiter=',')
pm_class = cartpole_pm.CartPolePM(X,Y)
X2 = np.r_[X, np.loadtxt(load_path+"policy"+str(loop_num)+'/simulation2/simulation_samples_input.csv' , delimiter=',')]
Y2 = np.r_[Y, np.loadtxt(load_path+"policy"+str(loop_num)+'/simulation2/simulation_samples_output.csv', delimiter=',')]
print("prev_log_evidence =",pm_class.log_evidence())
print("debug_log_evidence =",pm_class.log_evidence_of_new_data(X,Y))
print("new_log_evidence =",pm_class.log_evidence_of_new_data(X2,Y2))
temp_data = np.array([pm_class.log_evidence(), pm_class.log_evidence_of_new_data(X2,Y2)])
np.savetxt(save_path+"policy"+str(loop_num)+'/simulation2/log_evidence.csv',temp_data,delimiter=',')
np.savetxt(save_path+"policy"+str(loop_num)+'/log_evidence_for_learn2.csv',np.array([pm_class.log_evidence()]),delimiter=',', fmt ='%.6f')
return temp_data[1]
def estimate_log_evidence_npm_pendulum(load_path="./result_apply/",
save_path="./result_apply/",
loop_num=0,
i=5):
gather_sample.gather_real_world_sample(loop_num=loop_num)
X = np.loadtxt(load_path + 'real_world_samples_input.csv', delimiter=',')
Y = np.loadtxt(load_path + 'real_world_samples_output.csv', delimiter=',')
X = np.delete(X, i, 0)
Y = np.delete(Y, i, 0)
print("X.shape =", X.shape)
npm_class = pendulum_npm.PendulumNPM(X, Y)
gather_sample.gather_real_world_sample(loop_num=loop_num)
X = np.loadtxt(load_path + 'real_world_samples_input.csv', delimiter=',')
Y = np.loadtxt(load_path + 'real_world_samples_output.csv', delimiter=',')
X = np.delete(X, i, 0)
Y = np.delete(Y, i, 0)
X = np.insert(X, 1, np.sin(X[:, 0]), axis=1) # [th,sinth,thdot,u]
X[:, 0] = np.cos(X[:, 0]) # [costh,sinth,thdot,u]
gprd_m1 = gpr_without_optimize(X, Y[:, 0].reshape((Y[:, 0].shape[0]), 1))
gprd_m2 = gpr_without_optimize(X, Y[:, 1].reshape((Y[:, 1].shape[0]), 1))
gprd_m1[:] = npm_class.gpr_m1.param_array
gprd_m2[:] = npm_class.gpr_m2.param_array
print("prev_log_evidence =", npm_class.log_evidence())
print("debug_log_evidence =",
gprd_m1.log_likelihood() + gprd_m2.log_likelihood())
gather_sample.gather_real_world_sample(loop_num=loop_num)
X = np.loadtxt(load_path + 'real_world_samples_input.csv', delimiter=',')
Y = np.loadtxt(load_path + 'real_world_samples_output.csv', delimiter=',')
X = np.delete(X, i, 0)
Y = np.delete(Y, i, 0)
X2 = np.loadtxt(load_path + 'real_world_samples_input.csv', delimiter=',')
Y2 = np.loadtxt(load_path + 'real_world_samples_output.csv', delimiter=',')
X2 = np.insert(X2, 1, np.sin(X2[:, 0]), axis=1) # [th,sinth,thdot,u]
X2[:, 0] = np.cos(X2[:, 0]) # [costh,sinth,thdot,u]
gpr_m1 = gpr_without_optimize(X2, Y2[:, 0].reshape((Y2[:, 0].shape[0]), 1))
gpr_m2 = gpr_without_optimize(X2, Y2[:, 1].reshape((Y2[:, 1].shape[0]), 1))
gpr_m1[:] = npm_class.gpr_m1.param_array
gpr_m2[:] = npm_class.gpr_m2.param_array
print("prev_log_evidence =", npm_class.log_evidence())
print("debug_log_evidence =",
gprd_m1.log_likelihood() + gprd_m2.log_likelihood())
temp_le = gpr_m1.log_likelihood() + gpr_m2.log_likelihood()
print("new_log_evidence =", temp_le)
print("diff =", temp_le - npm_class.log_evidence())
return (temp_le - npm_class.log_evidence())
def estimate_log_evidence_npm_cartpole(load_path="./result_apply/",
save_path="./result_apply/",
loop_num=0):
gather_sample.gather_real_world_sample(loop_num=loop_num)
X = np.loadtxt(load_path+'real_world_samples_input.csv', delimiter=',')
Y = np.loadtxt(load_path+'real_world_samples_output.csv', delimiter=',')
npm_class = cartpole_npm.CartPoleNPM(X,Y)
gather_sample.gather_real_world_sample(loop_num=loop_num)
X = np.loadtxt(load_path+'real_world_samples_input.csv', delimiter=',')
Y = np.loadtxt(load_path+'real_world_samples_output.csv', delimiter=',')
X = np.insert(X, 3, np.sin(X[:,2]), axis=1) # [z,zdot,th,sinth,thdot,u]
X[:,2] = np.cos(X[:,2]) # [z,zdot,costh,sinth,thdot,u]
gprd_m1 = gpr_without_optimize(X, Y[:,0].reshape((Y[:,0].shape[0]),1))
gprd_m2 = gpr_without_optimize(X, Y[:,1].reshape((Y[:,1].shape[0]),1))
gprd_m3 = gpr_without_optimize(X, Y[:,2].reshape((Y[:,2].shape[0]),1))
gprd_m4 = gpr_without_optimize(X, Y[:,3].reshape((Y[:,3].shape[0]),1))
gprd_m1[:] = npm_class.gpr_m1.param_array
gprd_m2[:] = npm_class.gpr_m2.param_array
gprd_m3[:] = npm_class.gpr_m3.param_array
gprd_m4[:] = npm_class.gpr_m4.param_array
gather_sample.gather_real_world_sample(loop_num=loop_num)
X = np.loadtxt(load_path+'real_world_samples_input.csv', delimiter=',')
Y = np.loadtxt(load_path+'real_world_samples_output.csv', delimiter=',')
X2 = np.r_[X, np.loadtxt(load_path+"policy"+str(loop_num)+'/simulation1/simulation_samples_input.csv' , delimiter=',')]
Y2 = np.r_[Y, np.loadtxt(load_path+"policy"+str(loop_num)+'/simulation1/simulation_samples_output.csv', delimiter=',')]
X2 = np.insert(X2, 3, np.sin(X2[:,2]), axis=1) # [z,zdot,th,sinth,thdot,u]
X2[:,2] = np.cos(X2[:,2]) # [z,zdot,costh,sinth,thdot,u]
gpr_m1 = gpr_without_optimize(X2, Y2[:,0].reshape((Y2[:,0].shape[0]),1))
gpr_m2 = gpr_without_optimize(X2, Y2[:,1].reshape((Y2[:,1].shape[0]),1))
gpr_m3 = gpr_without_optimize(X2, Y2[:,2].reshape((Y2[:,2].shape[0]),1))
gpr_m4 = gpr_without_optimize(X2, Y2[:,3].reshape((Y2[:,3].shape[0]),1))
gpr_m1[:] = npm_class.gpr_m1.param_array
gpr_m2[:] = npm_class.gpr_m2.param_array
gpr_m3[:] = npm_class.gpr_m3.param_array
gpr_m4[:] = npm_class.gpr_m4.param_array
print("prev_log_evidence =",npm_class.log_evidence())
print("debug_log_evidence =",gprd_m1.log_likelihood() + gprd_m2.log_likelihood() + gprd_m3.log_likelihood() + gprd_m4.log_likelihood())
temp_le = gpr_m1.log_likelihood() + gpr_m2.log_likelihood() + gpr_m3.log_likelihood() + gpr_m4.log_likelihood()
print("new_log_evidence =",temp_le)
temp_data = np.array([npm_class.log_evidence(), temp_le])
np.savetxt(save_path+"policy"+str(loop_num)+'/simulation1/log_evidence.csv',temp_data,delimiter=',')
np.savetxt(save_path+"policy"+str(loop_num)+'/log_evidence_for_learn1.csv',np.array([npm_class.log_evidence()]),delimiter=',', fmt ='%.6f')
return temp_data[1]