75 lines
1.7 KiB
Python
75 lines
1.7 KiB
Python
|
"""LVQ1 example using the Iris dataset."""
|
||
|
|
|
||
|
|
import prototorch as pt
|
||
|
|
import pytorch_lightning as pl
|
||
|
|
import torch
|
||
|
|
|
||
|
|
if __name__ == "__main__":
|
||
|
|
# Acquire data
|
||
|
|
from sklearn.datasets import load_iris
|
||
|
|
x_train, y_train = load_iris(return_X_y=True)
|
||
|
|
x_train = x_train[:, [0, 2]]
|
||
|
|
|
||
|
|
# Relabel classes
|
||
|
|
# y_train[y_train == 0] = 3
|
||
|
|
# y_train[y_train == 1] = 4
|
||
|
|
# y_train[y_train == 2] = 6
|
||
|
|
|
||
|
|
# Dataset
|
||
|
|
train_ds = pt.datasets.NumpyDataset(x_train, y_train)
|
||
|
|
|
||
|
|
# Dataloaders
|
||
|
|
train_loader = torch.utils.data.DataLoader(train_ds, shuffle=True)
|
||
|
|
|
||
|
|
# Hyperparameters
|
||
|
|
num_classes = 3
|
||
|
|
prototypes_per_class = 10
|
||
|
|
hparams = dict(
|
||
|
|
distribution={
|
||
|
|
# class_label: num_prototypes
|
||
|
|
# 3: 1,
|
||
|
|
# 4: 2,
|
||
|
|
# 6: 3,
|
||
|
|
0: 1,
|
||
|
|
2: 2,
|
||
|
|
3: 3,
|
||
|
|
},
|
||
|
|
lr=0.001,
|
||
|
|
)
|
||
|
|
|
||
|
|
# Initialize the model
|
||
|
|
model = pt.models.LVQ1(
|
||
|
|
hparams,
|
||
|
|
prototypes_initializer=pt.initializers.SMCI(train_ds),
|
||
|
|
)
|
||
|
|
|
||
|
|
# Check if `num_classes` is correct
|
||
|
|
print(f"{model.num_classes=}")
|
||
|
|
assert model.num_classes == 3
|
||
|
|
|
||
|
|
# Compute intermediate input and output sizes
|
||
|
|
model.example_input_array = torch.zeros(4, 2)
|
||
|
|
|
||
|
|
# Model summary
|
||
|
|
print(model)
|
||
|
|
|
||
|
|
# Callbacks
|
||
|
|
vis = pt.models.VisGLVQ2D(data=(x_train, y_train),
|
||
|
|
cmap="viridis",
|
||
|
|
resolution=200,
|
||
|
|
block=False)
|
||
|
|
|
||
|
|
# Setup trainer
|
||
|
|
trainer = pl.Trainer(
|
||
|
|
gpus=0,
|
||
|
|
max_epochs=50,
|
||
|
|
callbacks=[vis],
|
||
|
|
# fast_dev_run=1,
|
||
|
|
)
|
||
|
|
|
||
|
|
# Get prototype labels
|
||
|
|
print(f"Protoype Labels are: ", model.prototype_labels.tolist())
|
||
|
|
|
||
|
|
# Training loop
|
||
|
|
trainer.fit(model, train_loader)
|