apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-11 13:02:38 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

self.num_data and self.input_dim are set dynamically in class GP() after the shape of X. In MRD, the user-specific values are passed around until X is defined.

Browse source
This commit is contained in:
Masha Naslidnyk 🦉 2020-01-10 12:42:03 +00:00
parent 1e06c6ce2f
commit 7871af8dec
2 changed files with 18 additions and 17 deletions
Download patch file Download diff file Expand all files Collapse all files

11
GPy/core/gp.py
View file

@ -43,8 +43,6 @@ class GP(Model):
self.X = X.copy() self.X = X.copy()
else: self.X = ObsAr(X) else: self.X = ObsAr(X)
self.num_data, self.input_dim = self.X.shape
assert Y.ndim == 2 assert Y.ndim == 2
logger.info("initializing Y") logger.info("initializing Y")
@ -199,6 +197,14 @@ class GP(Model):
def _predictive_variable(self): def _predictive_variable(self):
return self.X return self.X
@property
def num_data(self):
return self.X.shape[0]
@property
def input_dim(self):
return self.X.shape[1]
def set_XY(self, X=None, Y=None): def set_XY(self, X=None, Y=None):
""" """
Set the input / output data of the model Set the input / output data of the model
@ -236,7 +242,6 @@ class GP(Model):
else: else:
self.X = ObsAr(X) self.X = ObsAr(X)
self.num_data, self.input_dim = self.X.shape
self.update_model(True) self.update_model(True)
def set_X(self,X): def set_X(self,X):

24
GPy/models/mrd.py
View file

@ -63,7 +63,6 @@ class MRD(BayesianGPLVMMiniBatch):
Ynames=None, normalizer=False, stochastic=False, batchsize=10): Ynames=None, normalizer=False, stochastic=False, batchsize=10):
self.logger = logging.getLogger(self.__class__.__name__) self.logger = logging.getLogger(self.__class__.__name__)
self.input_dim = input_dim
self.num_inducing = num_inducing self.num_inducing = num_inducing
if isinstance(Ylist, dict): if isinstance(Ylist, dict):
@ -87,11 +86,11 @@ class MRD(BayesianGPLVMMiniBatch):
self.inference_method = inference_method self.inference_method = inference_method
if X is None: if X is None:
X, fracs = self._init_X(initx, Ylist) X, fracs = self._init_X(input_dim, initx, Ylist)
else: else:
fracs = [X.var(0)]*len(Ylist) fracs = [X.var(0)]*len(Ylist)
Z = self._init_Z(initz, X) Z = self._init_Z(initz, X, input_dim)
self.Z = Param('inducing inputs', Z) self.Z = Param('inducing inputs', Z)
self.num_inducing = self.Z.shape[0] # ensure M==N if M>N self.num_inducing = self.Z.shape[0] # ensure M==N if M>N
@ -128,7 +127,6 @@ class MRD(BayesianGPLVMMiniBatch):
self.unlink_parameter(self.likelihood) self.unlink_parameter(self.likelihood)
self.unlink_parameter(self.kern) self.unlink_parameter(self.kern)
self.num_data = Ylist[0].shape[0]
if isinstance(batchsize, int): if isinstance(batchsize, int):
batchsize = itertools.repeat(batchsize) batchsize = itertools.repeat(batchsize)
@ -187,32 +185,32 @@ class MRD(BayesianGPLVMMiniBatch):
def log_likelihood(self): def log_likelihood(self):
return self._log_marginal_likelihood return self._log_marginal_likelihood
def _init_X(self, init='PCA', Ylist=None): def _init_X(self, input_dim, init='PCA', Ylist=None):
if Ylist is None: if Ylist is None:
Ylist = self.Ylist Ylist = self.Ylist
if init in "PCA_concat": if init in "PCA_concat":
X, fracs = initialize_latent('PCA', self.input_dim, np.hstack(Ylist)) X, fracs = initialize_latent('PCA', input_dim, np.hstack(Ylist))
fracs = [fracs]*len(Ylist) fracs = [fracs]*len(Ylist)
elif init in "PCA_single": elif init in "PCA_single":
X = np.zeros((Ylist[0].shape[0], self.input_dim)) X = np.zeros((Ylist[0].shape[0], input_dim))
fracs = np.empty((len(Ylist), self.input_dim)) fracs = np.empty((len(Ylist), input_dim))
for qs, Y in zip(np.array_split(np.arange(self.input_dim), len(Ylist)), Ylist): for qs, Y in zip(np.array_split(np.arange(input_dim), len(Ylist)), Ylist):
x, frcs = initialize_latent('PCA', len(qs), Y) x, frcs = initialize_latent('PCA', len(qs), Y)
X[:, qs] = x X[:, qs] = x
fracs[:, qs] = frcs fracs[:, qs] = frcs
else: # init == 'random': else: # init == 'random':
X = np.random.randn(Ylist[0].shape[0], self.input_dim) X = np.random.randn(Ylist[0].shape[0], input_dim)
fracs = X.var(0) fracs = X.var(0)
fracs = [fracs]*len(Ylist) fracs = [fracs]*len(Ylist)
X -= X.mean() X -= X.mean()
X /= X.std() X /= X.std()
return X, fracs return X, fracs
def _init_Z(self, init, X): def _init_Z(self, init, X, input_dim):
if init in "permute": if init in "permute":
Z = np.random.permutation(X.copy())[:self.num_inducing] Z = np.random.permutation(X.copy())[:self.num_inducing]
elif init in "random": elif init in "random":
Z = np.random.randn(self.num_inducing, self.input_dim) * X.var() Z = np.random.randn(self.num_inducing, input_dim) * X.var()
return Z return Z
def predict(self, Xnew, full_cov=False, Y_metadata=None, kern=None, Yindex=0): def predict(self, Xnew, full_cov=False, Y_metadata=None, kern=None, Yindex=0):
@ -350,5 +348,3 @@ class MRD(BayesianGPLVMMiniBatch):
print('# Private dimensions model ' + str(i) + ':' + str(privateDims[i])) print('# Private dimensions model ' + str(i) + ':' + str(privateDims[i]))
return sharedDims, privateDims return sharedDims, privateDims