def runModel_i(input_vector):
hls, lr, ne, rs = input_vector
model = init_two_layer_model(
32 * 32 * 3, int(hls),
10) # input size, hidden size, number of classes
trainer = ClassifierTrainer()
best_model = {}
loss_history = []
train_acc_history = []
val_acc_history = []
best_model, loss_history, train_acc_history, val_acc_history = trainer.train(
X_train,
y_train,
X_val,
y_val,
model,
two_layer_net,
num_epochs=int(ne),
reg=rs,
momentum=0.9,
learning_rate_decay=0.95,
learning_rate=lr,
verbose=True)
print("Inside model I")
print(val_acc_history)
return best_model, loss_history, train_acc_history, val_acc_history
hidden_size = [600, 800, 1000, 1200, 1400, 2000]
training_epochs = [45]
regs = [1e-2, 1e-3, 5e-3, 1e-4, 5e-4, 2]
learning_rate = [1e-3, 1e-4, 5e-5, 1e-6, 5e-6]
best = 0
bmodel = None
best_para = {}
for hs in hidden_size:
for ep in training_epochs:
for reg in regs:
for lr in learning_rate:
print "test on param hs :", hs, " ep: ", ep, " reg: ", reg, " lr:", lr
model = init_two_layer_model(
32 * 32 * 3, hs,
10) # input size, hidden size, number of classes
trainer = ClassifierTrainer()
best_model, loss_history, train_acc, val_acc = trainer.train(
X_train,
y_train,
X_val,
y_val,
model,
two_layer_net,
num_epochs=ep,
reg=reg,
momentum=0.9,
learning_rate_decay=0.95,
learning_rate=lr,
verbose=True)
plt.subplot(2, 1, 1)
print('%s max relative error: %e' %
(param_name, rel_error(param_grad_num, grads[param_name])))
###############################################################################
# # Train the network
# To train the network we will use SGD with Momentum.
#Open the file `classifier_trainer.py` and familiarize yourself with the `ClassifierTrainer`
#class. It performs optimization given an arbitrary cost function data, and model.
#By default it uses vanilla SGD, which you need to implement.
#First, run the optimization below using Vanilla SGD:
from cs231n.classifier_trainer import ClassifierTrainer
model = init_toy_model()
trainer = ClassifierTrainer()
# call the trainer to optimize the loss
# Notice that we're using sample_batches=False, so we're performing Gradient Descent (no sampled batches of data)
best_model, loss_history, _, _ = trainer.train(X,
y,
X,
y,
model,
two_layer_net,
reg=0.001,
learning_rate=1e-1,
momentum=0.0,
learning_rate_decay=1,
update='sgd',
sample_batches=False,
num_epochs=100,
