def calPredictions(thisTree, Wcat, Wv, indicies, sStr, params):
wsz = params.wordSize
# vecPath - contains the vectors along the node path
# nodePath - contains the nodes along the node path
thisTree.nodePath = findPath(thisTree, indicies)
# Get the internal features
thisTree.features = getInternalFeatures(thisTree.nodePath, sStr, indicies, params.features_mean, params.features_std)
# Add only the top node
thisTree.nodePath = np.max(thisTree.nodePath)
thisTree.pooledVecPath = thisTree.nodeAct_a[:,thisTree.nodePath].reshape(wsz,1)
# Add the elements
thisTree.pooledVecPath = np.hstack((thisTree.pooledVecPath, thisTree.nodeAct_a[:,[indicies[0]]], thisTree.nodeAct_a[:,[indicies[1]]]))
thisTree.nodePath = np.array([thisTree.nodePath, indicies[0], indicies[1]], dtype='int32')
thisTree.poolMatrix = np.eye(3*wsz)
# add outer context words
thisTree = addOuterContext(thisTree, indicies, len(sStr), params);
# add inner context words
thisTree = addInnerContext(thisTree,indicies, Wv, len(sStr), params);
# add in nearest neighbors
# thisTree = addNN(thisTree, Wv, sNN, params); #bhanu
catInput = np.concatenate((thisTree.pooledVecPath.flatten(), thisTree.features))#, thisTree.NN_vecs.flatten())) #bhanu
thisTree.score = np.dot(Wcat, np.transpose(catInput)) #score for each label/tag
num = np.exp(thisTree.score)
thisTree.y = np.divide(num, np.sum(num)) #conditional probs for each tag/label
def _setInternalFeatureStats(self, data):
# features [path length, path depth1, path depth2, sentence length, length between elements]
numFeatures = 5;
self.features_mean = 0;
self.features_std = np.ones(numFeatures);
totaltrees = 0
for s in range(len(data.allSStr)):
nverbs = len(data.verbIndices[s].flatten())
sentLength = len(data.allSStr[s].flatten())
totaltrees += nverbs*sentLength
allFeatures = np.zeros((totaltrees,numFeatures));
thisTree = Tree()
for s in range(len(data.allSStr)):
thisTree.pp = data.allSTree[s].flatten()
verbIndexesList = data.verbIndices[s].flatten()
for vid in verbIndexesList:
for wid in range(len(data.allSStr[s])):
indices = [vid, wid]
fullPath = findPath(thisTree, indices)
allFeatures[s,:] = getInternalFeatures(fullPath, data.allSNum[s], indices, self.features_mean, self.features_std);
# Find the mean
self.features_mean = np.mean(allFeatures,axis=0); # check this
allFeatures = allFeatures - self.features_mean
# Find the standard deviation
self.features_std = np.std(allFeatures, axis=0)
def calPredictions(thisTree, Wcat, Wv, indicies, sStr, params):
wsz = params.wordSize
# vecPath - contains the vectors along the node path
# nodePath - contains the nodes along the node path
thisTree.nodePath = findPath(thisTree, indicies)
# Get the internal features
thisTree.features = getInternalFeatures(thisTree.nodePath, sStr, indicies,
params.features_mean,
params.features_std)
# Add only the top node
thisTree.nodePath = np.max(thisTree.nodePath)
thisTree.pooledVecPath = thisTree.nodeAct_a[:, thisTree.nodePath].reshape(
wsz, 1)
# Add the elements
thisTree.pooledVecPath = np.hstack(
(thisTree.pooledVecPath, thisTree.nodeAct_a[:, [indicies[0]]],
thisTree.nodeAct_a[:, [indicies[1]]]))
thisTree.nodePath = np.array([thisTree.nodePath, indicies[0], indicies[1]],
dtype='int32')
thisTree.poolMatrix = np.eye(3 * wsz)
# add outer context words
thisTree = addOuterContext(thisTree, indicies, len(sStr), params)
# add inner context words
thisTree = addInnerContext(thisTree, indicies, Wv, len(sStr), params)
# add in nearest neighbors
# thisTree = addNN(thisTree, Wv, sNN, params); #bhanu
catInput = np.concatenate(
(thisTree.pooledVecPath.flatten(),
thisTree.features)) #, thisTree.NN_vecs.flatten())) #bhanu
thisTree.score = np.dot(Wcat,
np.transpose(catInput)) #score for each label/tag
num = np.exp(thisTree.score)
thisTree.y = np.divide(num,
np.sum(num)) #conditional probs for each tag/label