程序代写代做代考 Coursework_1
Coursework_1
Machine Learning Practical: Coursework 1¶
Release date: Monday 10th October 2016
Due date: 16:00 Thursday 27th October 2016
Instructions for the coursework are available as a PDF here.
Part 1: Learning rate schedules¶
In [ ]:
# The below code will set up the data providers, random number
# generator and logger objects needed for training runs. As
# loading the data from file take a little while you generally
# will probably not want to reload the data providers on
# every training run. If you wish to reset their state you
# should instead use the .reset() method of the data providers.
import numpy as np
import logging
from mlp.data_providers import MNISTDataProvider
# Seed a random number generator
seed = 10102016
rng = np.random.RandomState(seed)
# Set up a logger object to print info about the training run to stdout
logger = logging.getLogger()
logger.setLevel(logging.INFO)
logger.handlers = [logging.StreamHandler()]
# Create data provider objects for the MNIST data set
train_data = MNISTDataProvider(‘train’, batch_size=50, rng=rng)
valid_data = MNISTDataProvider(‘valid’, batch_size=50, rng=rng)
In [ ]:
# The model set up code below is provided as a starting point.
# You will probably want to add further code cells for the
# different experiments you run.
from mlp.layers import AffineLayer, SoftmaxLayer, SigmoidLayer
from mlp.errors import CrossEntropySoftmaxError
from mlp.models import MultipleLayerModel
from mlp.initialisers import ConstantInit, GlorotUniformInit
input_dim, output_dim, hidden_dim = 784, 10, 100
weights_init = GlorotUniformInit(rng=rng)
biases_init = ConstantInit(0.)
model = MultipleLayerModel([
AffineLayer(input_dim, hidden_dim, weights_init, biases_init),
SigmoidLayer(),
AffineLayer(hidden_dim, hidden_dim, weights_init, biases_init),
SigmoidLayer(),
AffineLayer(hidden_dim, output_dim, weights_init, biases_init)
])
error = CrossEntropySoftmaxError()
Part 2: Momentum learning rule¶
In [ ]:
# The below code will set up the data providers, random number
# generator and logger objects needed for training runs. As
# loading the data from file take a little while you generally
# will probably not want to reload the data providers on
# every training run. If you wish to reset their state you
# should instead use the .reset() method of the data providers.
import numpy as np
import logging
from mlp.data_providers import MNISTDataProvider
# Seed a random number generator
seed = 10102016
rng = np.random.RandomState(seed)
# Set up a logger object to print info about the training run to stdout
logger = logging.getLogger()
logger.setLevel(logging.INFO)
logger.handlers = [logging.StreamHandler()]
# Create data provider objects for the MNIST data set
train_data = MNISTDataProvider(‘train’, batch_size=50, rng=rng)
valid_data = MNISTDataProvider(‘valid’, batch_size=50, rng=rng)
In [ ]:
# The model set up code below is provided as a starting point.
# You will probably want to add further code cells for the
# different experiments you run.
from mlp.layers import AffineLayer, SoftmaxLayer, SigmoidLayer
from mlp.errors import CrossEntropySoftmaxError
from mlp.models import MultipleLayerModel
from mlp.initialisers import ConstantInit, GlorotUniformInit
input_dim, output_dim, hidden_dim = 784, 10, 100
weights_init = GlorotUniformInit(rng=rng)
biases_init = ConstantInit(0.)
model = MultipleLayerModel([
AffineLayer(input_dim, hidden_dim, weights_init, biases_init),
SigmoidLayer(),
AffineLayer(hidden_dim, hidden_dim, weights_init, biases_init),
SigmoidLayer(),
AffineLayer(hidden_dim, output_dim, weights_init, biases_init)
])
error = CrossEntropySoftmaxError()
Part 3: Adaptive learning rules¶
In [ ]:
# The below code will set up the data providers, random number
# generator and logger objects needed for training runs. As
# loading the data from file take a little while you generally
# will probably not want to reload the data providers on
# every training run. If you wish to reset their state you
# should instead use the .reset() method of the data providers.
import numpy as np
import logging
from mlp.data_providers import MNISTDataProvider
# Seed a random number generator
seed = 10102016
rng = np.random.RandomState(seed)
# Set up a logger object to print info about the training run to stdout
logger = logging.getLogger()
logger.setLevel(logging.INFO)
logger.handlers = [logging.StreamHandler()]
# Create data provider objects for the MNIST data set
train_data = MNISTDataProvider(‘train’, batch_size=50, rng=rng)
valid_data = MNISTDataProvider(‘valid’, batch_size=50, rng=rng)
In [ ]:
# The model set up code below is provided as a starting point.
# You will probably want to add further code cells for the
# different experiments you run.
from mlp.layers import AffineLayer, SoftmaxLayer, SigmoidLayer
from mlp.errors import CrossEntropySoftmaxError
from mlp.models import MultipleLayerModel
from mlp.initialisers import ConstantInit, GlorotUniformInit
input_dim, output_dim, hidden_dim = 784, 10, 100
weights_init = GlorotUniformInit(rng=rng)
biases_init = ConstantInit(0.)
model = MultipleLayerModel([
AffineLayer(input_dim, hidden_dim, weights_init, biases_init),
SigmoidLayer(),
AffineLayer(hidden_dim, hidden_dim, weights_init, biases_init),
SigmoidLayer(),
AffineLayer(hidden_dim, output_dim, weights_init, biases_init)
])
error = CrossEntropySoftmaxError()
