def breast_cancer(x_train, y_train, x_val, y_val, params):
print("Iteration parameters: ", params)
def weights_init_uniform_rule(m):
classname = m.__class__.__name__
if classname.find('Linear') != -1:
n = m.in_features
y = 1.0 / np.sqrt(n)
m.weight.data.uniform_(-y, y)
m.bias.data.fill_(0)
manager = DataManager.from_numpy(train_inputs=x_train,
train_labels=y_train,
batch_size=params["batch_size"],
validation_inputs=x_val,
validation_labels=y_val)
net = BreastCancerNet(n_feature=x_train.shape[1],
first_neuron=params["first_neuron"],
second_neuron=params["second_neuron"],
dropout=params["dropout"])
net.apply(weights_init_uniform_rule)
net.init_history()
model = DeepLearningInterface(model=net,
optimizer_name=params["optimizer_name"],
learning_rate=params["learning_rate"],
loss_name=params["loss_name"],
metrics=["accuracy"])
model.add_observer("after_epoch", update_talos_history)
model.training(manager=manager,
nb_epochs=params["epochs"],
checkpointdir=None,
fold_index=0,
with_validation=True)
return net, net.parameters()
def setUp(self):
""" Setup test.
"""
data = fetch_cifar(datasetdir="/tmp/cifar")
self.manager = DataManager(input_path=data.input_path,
labels=["label"],
metadata_path=data.metadata_path,
number_of_folds=10,
batch_size=10,
stratify_label="category",
test_size=0.1,
sample_size=0.01)
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(3, 6, 5)
self.pool = nn.MaxPool2d(2, 2)
self.conv2 = nn.Conv2d(6, 16, 5)
self.fc1 = nn.Linear(16 * 5 * 5, 120)
self.fc2 = nn.Linear(120, 84)
self.fc3 = nn.Linear(84, 10)
def forward(self, x):
x = self.pool(func.relu(self.conv1(x)))
x = self.pool(func.relu(self.conv2(x)))
x = x.view(-1, 16 * 5 * 5)
x = func.relu(self.fc1(x))
x = func.relu(self.fc2(x))
x = self.fc3(x)
return x
self.cl = DeepLearningInterface(model=Net(),
optimizer_name="SGD",
momentum=0.9,
learning_rate=0.001,
loss_name="CrossEntropyLoss",
metrics=["accuracy"])
# You may need to change the 'datasetdir' parameter.
import os
import numpy as np
from pynet.datasets import DataManager, fetch_echocardiography
from pynet.plotting import plot_data
from pynet.utils import setup_logging
setup_logging(level="info")
data = fetch_echocardiography(datasetdir="/tmp/echocardiography")
manager = DataManager(input_path=data.input_path,
metadata_path=data.metadata_path,
output_path=data.output_path,
number_of_folds=2,
stratify_label="label",
sampler="random",
batch_size=10,
test_size=0.1,
sample_size=0.2)
dataset = manager["test"]
print(dataset.inputs.shape, dataset.outputs.shape)
data = np.concatenate((dataset.inputs, dataset.outputs), axis=1)
plot_data(data, nb_samples=5)
#############################################################################
# Optimisation
# ------------
#
# From the available models load the UNet, and start the training.
# You may need to change the 'outdir' parameter.
#############################################################################
# Import a pynet dataset
# ----------------------
#
# Use a fetcher to retrieve some data and use generic interface to import and
# split this dataset: train, test and validation.
# You may need to change the 'datasetdir' parameter.
from pynet.datasets import DataManager, fetch_cifar
data = fetch_cifar(datasetdir="/tmp/cifar")
manager = DataManager(input_path=data.input_path,
labels=["label"],
metadata_path=data.metadata_path,
number_of_folds=10,
batch_size=50,
stratify_label="category",
test_size=0.1)
#############################################################################
# We have now a test, and multiple folds with train-validation datasets that
# can be used to train our network using cross-validation.
import numpy as np
from pynet.plotting import plot_data
print("Nb folds: ", manager.number_of_folds)
dataloader = manager.get_dataloader(train=True,
validation=False,
test=False,
[np.round(np.mean(labels[tri])) for tri in ico_triangles])
plot_trisurf(fig, ax, ico_vertices, ico_triangles, tri_texture)
data = np.zeros((N_SAMPLES, N_CLASSES, len(labels)), dtype=float)
for klass in (0, 1):
k_indices = np.argwhere(labels == 0).squeeze()
for loc, scale in SAMPLES[klass]:
data[:, klass, k_indices] = np.random.normal(loc=loc,
scale=scale,
size=len(k_indices))
labels = np.ones((N_SAMPLES, 1)) * labels
print("dataset: x {0} - y {1}".format(data.shape, labels.shape))
# Create data manager
manager = DataManager.from_numpy(train_inputs=data,
train_labels=labels,
test_inputs=data,
test_labels=labels,
batch_size=BATCH_SIZE)
# Create model
net_params = pynet.NetParameters(in_order=ICO_ORDER,
in_channels=2,
out_channels=N_CLASSES,
depth=3,
start_filts=32,
conv_mode="1ring",
up_mode="transpose",
cachedir=os.path.join(OUTDIR, "cache"))
model = SphericalUNetEncoder(net_params,
optimizer_name="SGD",
learning_rate=0.1,
# neural network is evaluated on the validation set, but not trained on it.
# If the validation loss starts to grow, it means that the network is
# overfitting the training set, and that it is time to stop the training.
#
# The following cell create stratified test, train, and validation loaders.
from pynet.datasets import fetch_orientation
from pynet.datasets import DataManager
data = fetch_orientation(
datasetdir="/tmp/orientation",
flatten=True)
manager = DataManager(
input_path=data.input_path,
labels=["label"],
metadata_path=data.metadata_path,
number_of_folds=10,
batch_size=1000,
stratify_label="label",
test_size=0.1)
#############################################################################
# Displaying some images of the test dataset.
from pynet.plotting import plot_data
dataset = manager["test"]
sample = dataset.inputs.reshape(-1, data.height, data.width)
sample = np.expand_dims(sample, axis=1)
plot_data(sample, nb_samples=5)
# Show example noisy training data that have the signatures applied.
# It's not obvious to the human eye the subtle differences, but the cross
# row and column above perturbed the below matrices with the y weights.
# Show in the title how much each signature is weighted by.
plt.figure(figsize=(16, 4))
for idx in range(3):
plt.subplot(1, 3, idx + 1)
plt.imshow(np.squeeze(x_train[idx]), interpolation="None")
plt.colorbar()
plt.title(y_train[idx])
manager = DataManager.from_numpy(train_inputs=x_train,
train_labels=y_train,
validation_inputs=x_valid,
validation_labels=y_valid,
test_inputs=x_test,
test_labels=y_test,
batch_size=128,
continuous_labels=True)
interfaces = pynet.get_interfaces()["graph"]
net_params = pynet.NetParameters(input_shape=(90, 90),
in_channels=1,
num_classes=2,
nb_e2e=32,
nb_e2n=64,
nb_n2g=30,
dropout=0.5,
leaky_alpha=0.1,
twice_e2e=False,
dense_sml=True)
my_loss = pynet.get_tools()["losses"]["MSELoss"]()
the activation map of the last convolutional layer in our model.
Load the data
-------------
Load some images and apply the ImageNet transformation.
You may need to change the 'datasetdir' parameter.
"""
from pynet.datasets import DataManager, fetch_gradcam
from pynet.plotting import plot_data
data = fetch_gradcam(datasetdir="/tmp/gradcam")
manager = DataManager(input_path=data.input_path,
metadata_path=data.metadata_path,
number_of_folds=2,
batch_size=5,
test_size=1)
dataset = manager["test"]
print(dataset.inputs.shape)
plot_data(dataset.inputs, nb_samples=5, random=False, rgb=True)
#############################################################################
# Explore different architectures
# -------------------------------
#
# Let's automate this procedure for different networks.
# We need to reload the data for the inception network.
# You may need to change the 'datasetdir' parameter.
import os
masker = MultiNiftiMasker(mask_img=mask_img, standardize=True)
masker.fit()
if not os.path.isfile(DATAFILE):
y = np.concatenate(masker.transform(func_filenames), axis=0)
print(y.shape)
np.save(DATAFILE, y)
else:
y = np.load(DATAFILE)
iterator = masker.inverse_transform(y).get_fdata()
iterator = iterator.transpose((3, 0, 1, 2))
iterator = np.expand_dims(iterator, axis=1)
print(iterator.shape)
# Data iterator
manager = DataManager.from_numpy(train_inputs=iterator,
batch_size=BATCH_SIZE,
add_input=True)
# Create model
name = "ResAENet"
model_weights = os.path.join(WORKDIR, "checkpoint_" + name,
"model_0_epoch_{0}.pth".format(EPOCH))
if os.path.isfile(model_weights):
pretrained = model_weights
else:
pretrained = None
params = NetParameters(input_shape=(61, 73, 61),
cardinality=1,
layers=[3, 4, 6, 3],
n_channels_in=1,
decode=True)
# Show in the title how much each signature is weighted by.
plt.figure(figsize=(16, 4))
for idx in range(3):
plt.subplot(1, 3, idx + 1)
plt.imshow(np.squeeze(x_train[idx]), interpolation="None")
plt.colorbar()
plt.title(y_train[idx])
data = np.concatenate(data, axis=0)
labels = np.asarray(labels)
print("dataset: x {0} - y {1}".format(data.shape, labels.shape))
# Create data manager
manager = DataManager.from_numpy(
train_inputs=data, train_labels=np.zeros(labels.shape),
batch_size=BATCH_SIZE)
class FKmeans(object):
def __init__(self, n_clusters):
self.n_clusters = n_clusters
def fit(self, data):
n_data, d = data.shape
self.clus = faiss.Kmeans(d, self.n_clusters)
self.clus.seed = np.random.randint(1234)
self.clus.niter = 20
self.clus.max_points_per_centroid = 10000000
self.clus.train(data)
"""
#############################################################################
# Import the dataset
# ------------------
#
# You may need to change the 'datasetdir' parameter.
import numpy as np
from pynet.datasets import DataManager, fetch_echocardiography
from pynet.plotting import plot_data
data = fetch_echocardiography(datasetdir="/tmp/echocardiography")
manager = DataManager(input_path=data.input_path,
metadata_path=data.metadata_path,
output_path=data.output_path,
number_of_folds=10,
batch_size=10,
test_size=0.1)
dataset = manager["test"]
data = np.concatenate((dataset.inputs, dataset.outputs), axis=1)
plot_data(data, nb_samples=5)
#############################################################################
# Optimisation
# ------------
#
# From the available models load the UNet, and start the training.
# You may need to change the 'outdir' parameter.
import os
import torch
from pynet.losses import get_vae_loss
from pynet.models.vae.utils import (reconstruct_traverse, make_mosaic_img,
add_labels)
# Global parameters
WDIR = "/tmp/beta_vae_disentangling"
BATCH_SIZE = 64
N_EPOCHS = 30
ADAM_LR = 5e-4
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
DISPLAY = False
# Load the data
dataset = DSprites(WDIR)
manager = DataManager.from_dataset(train_dataset=dataset,
batch_size=BATCH_SIZE,
sampler="random")
# Test different losses
loss_params = {
"betah": {
"beta": 4,
"steps_anneal": 0,
"use_mse": True
},
"betab": {
"C_init": 0.5,
"C_fin": 25,
"gamma": 100,
"steps_anneal": 100000,
board = Board(port=8097, host="http://localhost", env="data-augmentation")
compose_transforms = Transformer()
compose_transforms.register(flip,
probability=0.5,
axis=0,
apply_to=["input", "output"])
compose_transforms.register(add_blur,
probability=1,
sigma=4,
apply_to=["input"])
manager = DataManager(input_path=data.input_path,
metadata_path=data.metadata_path,
output_path=data.output_path,
number_of_folds=2,
batch_size=2,
test_size=0.1,
sample_size=0.1,
sampler=None,
add_input=True,
data_augmentation_transforms=[compose_transforms])
loaders = manager.get_dataloader(train=True, validation=False, fold_index=0)
for dataitem in loaders.train:
print("-" * 50)
print(dataitem.inputs.shape, dataitem.outputs.shape, dataitem.labels)
images = [
dataitem.inputs[0, 0].numpy(), dataitem.inputs[0, 1].numpy(),
dataitem.outputs[0, 0].numpy(), dataitem.outputs[0, 1].numpy(),
dataitem.outputs[0, 4].numpy(), dataitem.outputs[0, 5].numpy()
]
images = np.asarray(images)
images = np.expand_dims(images, axis=1)
import os
import sys
from pynet.datasets import DataManager, fetch_height_biobank
from pynet.utils import setup_logging
# This example cannot run in CI : it accesses NS intra filesystems
if "CI_MODE" in os.environ:
sys.exit(0)
setup_logging(level="info")
data = fetch_height_biobank(datasetdir="/neurospin/tmp/height_bb")
manager = DataManager(input_path=data.input_path,
labels=["Height"],
metadata_path=data.metadata_path,
number_of_folds=2,
batch_size=5,
test_size=0.2,
continuous_labels=True)
#############################################################################
# Basic inspection
import numpy as np
import matplotlib.pyplot as plt
train_dataset = manager["train"][0]
X_train = train_dataset.inputs[train_dataset.indices]
y_train = train_dataset.labels[train_dataset.indices]
test_dataset = manager["test"]
X_test = test_dataset.inputs[test_dataset.indices]
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
losses = pynet.get_tools(tool_name="losses")
setup_logging(level="info")
#############################################################################
# Kang dataset
# ------------
#
# Fetch & load the Kang dataset:
data, trainset, testset, membership_mask = fetch_kang(datasetdir=datasetdir,
random_state=0)
gtpath = os.path.join(datasetdir, "kang_recons.h5ad")
manager = DataManager.from_numpy(train_inputs=trainset,
validation_inputs=testset,
test_inputs=data.X,
batch_size=batch_size,
sampler="random",
add_input=True)
#############################################################################
# Training
# --------
#
# Create/train the model:
if checkpointdir is not None:
weights_filename = os.path.join(
checkpointdir, "model_0_epoch_{0}.pth".format(nb_epochs - 1))
params = NetParameters(membership_mask=membership_mask,
latent_dim=latent_dim,
hidden_layers=[12],
#
# Use the fetcher of the pynet package.
from pynet.datasets import DataManager, fetch_brats
from pynet.plotting import plot_data
from pynet.transforms import RandomFlipDimensions, Offset
data = fetch_brats(datasetdir="/neurospin/nsap/datasets/brats")
manager = DataManager(
input_path=data.input_path,
metadata_path=data.metadata_path,
output_path=data.output_path,
projection_labels=None,
number_of_folds=10,
batch_size=1,
stratify_label="grade",
#input_transforms=[
# RandomFlipDimensions(ndims=3, proba=0.5, with_channels=True),
# Offset(nb_channels=4, factor=0.1)],
sampler="random",
add_input=True,
test_size=0.1,
pin_memory=True)
dataset = manager["test"][:1]
print(dataset.inputs.shape, dataset.outputs.shape)
plot_data(dataset.inputs, channel=1, nb_samples=5)
plot_data(dataset.outputs, channel=1, nb_samples=5)
#############################################################################
# Training
# --------
import matplotlib.pyplot as plt
setup_logging(level="info")
logger = logging.getLogger("pynet")
use_toy = False
dtype = "all"
data = fetch_impac(datasetdir="/neurospin/nsap/datasets/impac",
mode="train",
dtype=dtype)
nb_features = data.nb_features
manager = DataManager(input_path=data.input_path,
labels=["participants_asd"],
metadata_path=data.metadata_path,
number_of_folds=3,
batch_size=128,
sampler="random",
test_size=2,
sample_size=1)
if use_toy:
toy_data = {}
nb_features = 50
for name, nb_samples in (("train", 1000), ("test", 2)):
x1 = torch.randn(nb_samples, 50)
x2 = torch.randn(nb_samples, 50) + 1.5
x = torch.cat([x1, x2], dim=0)
y1 = torch.zeros(nb_samples, 1)
y2 = torch.ones(nb_samples, 1)
y = torch.cat([y1, y2], dim=0)
toy_data[name] = (x, y)
if not isinstance(imgtype, list):
imgtype = [imgtype]
imgtype = [typemap[key] for key in imgtype]
transformed_data = []
for channel_id in range(len(data)):
if channel_id not in imgtype:
continue
arr = data[channel_id]
transformed_data.append(downsample(arr, scale=3))
return np.asarray(transformed_data)
manager = DataManager(input_path=data.input_path,
metadata_path=data.metadata_path,
stratify_label="grade",
number_of_folds=10,
batch_size=batch_size,
test_size=0,
input_transforms=[transformer],
sample_size=0.2)
########################
# Loss
# ----
def calc_gradient_penalty(model, x, x_gen, w=10):
""" WGAN-GP gradient penalty.
"""
assert (x.size() == x_gen.size()), "Real and sampled sizes do not match."
alpha_size = tuple((len(x), *(1, ) * (x.dim() - 1)))
alpha_t = torch.cuda.FloatTensor if x.is_cuda else torch.Tensor
#
# A validation step is a useful way to avoid overfitting. At each epoch, the
# neural network is evaluated on the validation set, but not trained on it.
# If the validation loss starts to grow, it means that the network is
# overfitting the training set, and that it is time to stop the training.
#
# The following cell create stratified test, train, and validation loaders.
from pynet.datasets import fetch_orientation
from pynet.datasets import DataManager
data = fetch_orientation(datasetdir="/tmp/orientation", flatten=True)
manager = DataManager(input_path=data.input_path,
labels=["label"],
metadata_path=data.metadata_path,
number_of_folds=10,
batch_size=1000,
stratify_label="label",
test_size=0.1,
sample_size=(1 if "CI_MODE" not in os.environ else 0.1))
#############################################################################
# Displaying some images of the test dataset.
from pynet.plotting import plot_data
dataset = manager["test"]
sample = dataset.inputs.reshape(-1, data.height, data.width)
sample = np.expand_dims(sample, axis=1)
plot_data(sample, nb_samples=5)
#############################################################################
from pynet.plotting import Board, update_board
#############################################################################
# The model will be trained on MNIST - handwritten digits dataset. The input
# is an image in R(28×28).
def flatten(arr):
return arr.flatten()
data = fetch_minst(datasetdir="/neurospin/nsap/datasets/minst")
manager = DataManager(
input_path=data.input_path,
metadata_path=data.metadata_path,
stratify_label="label",
number_of_folds=10,
batch_size=64,
test_size=0,
input_transforms=[flatten],
add_input=True,
sample_size=0.05)
#############################################################################
# The Model
# ---------
#
# The model is composed of two sub-networks:
#
# 1. Given x (image), encode it into a distribution over the latent space -
# referred to as Q(z|x).
# 2. Given z in latent space (code representation of an image), decode it into
import os
import numpy as np
from pynet.datasets import DataManager, fetch_echocardiography
from pynet.plotting import plot_data
from pynet.utils import setup_logging
setup_logging(level="info")
data = fetch_echocardiography(
datasetdir="/tmp/echocardiography")
manager = DataManager(
input_path=data.input_path,
metadata_path=data.metadata_path,
output_path=data.output_path,
number_of_folds=2,
stratify_label="label",
sampler="weighted_random",
batch_size=10,
test_size=0.1,
sample_size=(1 if "CI_MODE" not in os.environ else 0.05))
dataset = manager["test"]
print(dataset.inputs.shape, dataset.outputs.shape)
data = np.concatenate((dataset.inputs, dataset.outputs), axis=1)
plot_data(data, nb_samples=5)
#############################################################################
# Optimisation
# ------------
#
# From the available models load the UNet, and start the training.
Load some data.
You may need to change the 'datasetdir' parameter.
"""
import os
from pynet.datasets import DataManager, fetch_genomic_pred
from pynet.utils import setup_logging
setup_logging(level="info")
data = fetch_genomic_pred(datasetdir="/tmp/genomic_pred")
manager = DataManager(input_path=data.input_path,
labels=["env0"],
metadata_path=data.metadata_path,
number_of_folds=2,
batch_size=5,
test_size=0.2,
continuous_labels=True)
#############################################################################
# Basic inspection
import numpy as np
from sklearn.decomposition import PCA
import matplotlib.pyplot as plt
train_dataset = manager["train"][0]
X_train = train_dataset.inputs[train_dataset.indices]
y_train = train_dataset.labels[train_dataset.indices]
test_dataset = manager["test"]
from pynet.plotting import plot_history
from pynet.history import History
from pynet.losses import MSELoss, NCCLoss, RCNetLoss
import matplotlib.pyplot as plt
setup_logging(level="debug")
logger = logging.getLogger("pynet")
outdir = "/neurospin/nsap/tmp/registration"
data = fetch_registration(
datasetdir=outdir)
manager = DataManager(
input_path=data.input_path,
metadata_path=data.metadata_path,
number_of_folds=10,
batch_size=1,
sampler="random",
#stratify_label="centers",
test_size=0.1,
add_input=True,
sample_size=1)
#############################################################################
# Training
# --------
#
# From the available models load the VoxelMorphRegister, VTNetRegister or
# ADDNet and start the training.
# Note that the two first estimate a non linear deformation and require
# the input data to be afinely registered. The ADDNet estimate an affine
# transform. We will see in the next section how to combine them in an
# efficient way.
from pynet.history import History
from pynet.losses import MSELoss, NCCLoss, RCNetLoss, PCCLoss
from pynet.plotting import Board, update_board
import matplotlib.pyplot as plt
setup_logging(level="debug")
logger = logging.getLogger("pynet")
losses = pynet.get_tools(tool_name="losses")
outdir = "/neurospin/nsap/tmp/registration"
data = fetch_registration(datasetdir=outdir)
manager = DataManager(input_path=data.input_path,
metadata_path=data.metadata_path,
number_of_folds=2,
batch_size=8,
sampler="random",
stratify_label="studies",
projection_labels={"studies": ["abide"]},
test_size=0.1,
add_input=True,
sample_size=0.1)
#############################################################################
# Training
# --------
#
# From the available models load the VoxelMorphRegister, VTNetRegister or
# ADDNet and start the training.
# Note that the two first estimate a non linear deformation and require
# the input data to be afinely registered. The ADDNet estimate an affine
# transform. We will see in the next section how to combine them in an
# efficient way.
n_feats=n_feats,
n_classes=n_classes,
train=True,
snr=snr)
ds_val = SyntheticDataset(n_samples=n_samples,
lat_dim=true_lat_dims,
n_feats=n_feats,
n_classes=n_classes,
train=False,
snr=snr)
image_datasets = {"train": ds_train, "val": ds_val}
manager = DataManager.from_numpy(train_inputs=ds_train.data,
train_outputs=None,
train_labels=ds_train.labels,
validation_inputs=ds_val.data,
validation_outputs=None,
validation_labels=ds_val.labels,
batch_size=batch_size,
sampler="random",
add_input=True)
print("- datasets:", image_datasets)
print("- shapes:", ds_train.data.shape, ds_val.data.shape)
# Display generated data
method = manifold.TSNE(n_components=2, init="pca", random_state=0)
y_train = method.fit_transform(ds_train.data)
y_val = method.fit_transform(ds_val.data)
fig, axs = plt.subplots(nrows=3, ncols=2)
for cnt, (name, y, labels) in enumerate(
(("train", y_train, ds_train.labels), ("val", y_val, ds_val.labels))):
colors = labels.astype(float)