river-torch 0.0.14

Online Deep Learning for river

DeepRiver is a Python library for online deep learning. DeepRivers ambition is to enable online machine learning for neural networks. It combines the river API with the capabilities of designing neural networks based on PyTorch.

💈 Installation

pip install river-torch

You can install the latest development version from GitHub as so:

pip install https://github.com/online-ml/river-torch/archive/refs/heads/master.zip

🍫 Quickstart

We build the development of neural networks on top of the river API and refer to the rivers design principles. The following example creates a simple MLP architecture based on PyTorch and incrementally predicts and trains on the website phishing dataset. For further examples check out the Documentation.

Classification

>>> from river.datasets import Phishing
>>> from river import metrics
>>> from river import preprocessing
>>> from river import compose
>>> from river_torch import classification
>>> from torch import nn
>>> from torch import optim
>>> from torch import manual_seed

>>> _ = manual_seed(42)


>>> def build_torch_mlp_classifier(n_features):
...     net = nn.Sequential(
...         nn.Linear(n_features, 5),
...         nn.ReLU(),
...         nn.Linear(5, 2),
...         nn.Softmax(-1)
...     )
...     return net


>>> model = compose.Pipeline(
...     preprocessing.StandardScaler(),
...     classification.Classifier(build_fn=build_torch_mlp_classifier, loss_fn='binary_cross_entropy', optimizer_fn=optim.Adam, lr=1e-3)
... )

>>> dataset = Phishing()
>>> metric = metrics.Accuracy()

>>> for x, y in dataset:
...     y_pred = model.predict_one(x)  # make a prediction
...     metric = metric.update(y, y_pred)  # update the metric
...     model = model.learn_one(x, y)  # make the model learn

>>> print(f'Accuracy: {metric.get()}')
Accuracy: 0.8336

Anomaly Detection

>>> import math
>>> from river import metrics, preprocessing
>>> from river.datasets import CreditCard
>>> from river_torch.anomaly import Autoencoder
>>> from river_torch.utils import get_activation_fn
>>> from torch import manual_seed, nn

>>> _ = manual_seed(42)

>>> def get_ae(n_features=3, dropout=0.1):
...     latent_dim = math.ceil(n_features / 2)
...     net = nn.Sequential(
...         nn.Dropout(p=dropout),
...         nn.Linear(in_features=n_features, out_features=latent_dim),
...         nn.ReLU(),
...         nn.Linear(in_features=latent_dim, out_features=n_features),
...         nn.Sigmoid()
...     )
...     return net

>>> dataset = CreditCard().take(5000)
>>> metric = metrics.ROCAUC()
>>> scaler = preprocessing.MinMaxScaler()

>>> model = Autoencoder(build_fn=get_ae, lr=0.01)

>>> for x, y in dataset:
...     x = scaler.learn_one(x).transform_one(x)
...     score = model.score_one(x)
...     metric = metric.update(y_true=y, y_pred=score)
...     model = model.learn_one(x=x)

🏫 Affiliations