import autograd.numpy as np
from autograd import grad
from util import shl, shm, shs
from datasets import ToyDataset
np.random.seed(10)
ds = ToyDataset()
xv, yv = ds.next_train_batch()
batch_size = xv.shape[0]
def cost(y):
x_gram = np.dot(xv, xv.T)
y_gram = np.dot(y, y.T)
return np.sum(np.square(y_gram - x_gram))
grad_cost = grad(cost)
n_comp = 2
yit = np.random.random((batch_size, n_comp))
for it in xrange(200):
cost_val = cost(yit)
dy = grad_cost(yit)
import numpy as np
import pickle
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import cv2
for j in range(10):
# NOTE! change this to not overwrite all log data when you train the model:
model_id = "Ensemble-MAP-Adam-Fixed_%d_M4" % (j + 1)
num_epochs = 150
batch_size = 32
learning_rate = 0.001
train_dataset = ToyDataset()
N = float(len(train_dataset))
print(N)
alpha = 1.0
num_train_batches = int(len(train_dataset) / batch_size)
print("num_train_batches:", num_train_batches)
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
batch_size=batch_size,
shuffle=True)
init_param_values = {}
network = ToyNet(model_id,
project_dir="/root/evaluating_bdl/toyRegression").cuda()
]
classifiers = [
KNeighborsClassifier(3),
SVC(kernel="linear", C=0.025),
SVC(gamma=2, C=1),
GaussianProcessClassifier(1.0 * RBF(1.0), warm_start=True),
DecisionTreeClassifier(max_depth=5),
RandomForestClassifier(max_depth=5, n_estimators=10, max_features=1),
MLPClassifier(max_iter=1000),
AdaBoostClassifier(),
GaussianNB(),
QuadraticDiscriminantAnalysis()
]
ds = ToyDataset()
# model = DecisionTreeClassifier()
for model_name, model in zip(names, classifiers):
x_train, y_train = ds.next_train_batch()
y_train = np.argmax(y_train, axis=1)
x_test, y_test = ds.next_test_batch()
y_test = np.argmax(y_test, axis=1)
model.fit(x_train, y_train)
predicted = model.predict(x_test)
print '%s, predicting, classification error=%f' % (