Probability weighted Autoencoder¶

In [1]:

from river import compose, preprocessing, metrics, datasets
from deep_river.anomaly import ProbabilityWeightedAutoencoder
from torch import nn, manual_seed

from river import compose, preprocessing, metrics, datasets from deep_river.anomaly import ProbabilityWeightedAutoencoder from torch import nn, manual_seed

In [2]:

_ = manual_seed(42)
dataset = datasets.CreditCard().take(5000)
metric = metrics.RollingROCAUC(window_size=5000)


class MyAutoEncoder(nn.Module):
    def __init__(self, n_features, latent_dim=3):
        super(MyAutoEncoder, self).__init__()
        self.linear1 = nn.Linear(n_features, latent_dim)
        self.nonlin = nn.LeakyReLU()
        self.linear2 = nn.Linear(latent_dim, n_features)
        self.sigmoid = nn.Sigmoid()

    def forward(self, X, **kwargs):
        X = self.linear1(X)
        X = self.nonlin(X)
        X = self.linear2(X)
        return self.sigmoid(X)


model_pipeline = compose.Pipeline(
    preprocessing.MinMaxScaler(),
    ProbabilityWeightedAutoencoder(module=MyAutoEncoder, lr=0.01),
)
model_pipeline

_ = manual_seed(42) dataset = datasets.CreditCard().take(5000) metric = metrics.RollingROCAUC(window_size=5000) class MyAutoEncoder(nn.Module): def __init__(self, n_features, latent_dim=3): super(MyAutoEncoder, self).__init__() self.linear1 = nn.Linear(n_features, latent_dim) self.nonlin = nn.LeakyReLU() self.linear2 = nn.Linear(latent_dim, n_features) self.sigmoid = nn.Sigmoid() def forward(self, X, **kwargs): X = self.linear1(X) X = self.nonlin(X) X = self.linear2(X) return self.sigmoid(X) model_pipeline = compose.Pipeline( preprocessing.MinMaxScaler(), ProbabilityWeightedAutoencoder(module=MyAutoEncoder, lr=0.01), ) model_pipeline

Out[2]:

MinMaxScaler

MinMaxScaler ()

ProbabilityWeightedAutoencoder

ProbabilityWeightedAutoencoder ( module=None loss_fn="mse" optimizer_fn="sgd" lr=0.01 device="cpu" seed=42 skip_threshold=0.9 window_size=250 )

In [3]:

for x, y in dataset:
    score = model_pipeline.score_one(x)
    metric.update(y_true=y, y_pred=score)
    model_pipeline.learn_one(x=x)
print(f"ROCAUC: {metric.get():.4f}")

for x, y in dataset: score = model_pipeline.score_one(x) metric.update(y_true=y, y_pred=score) model_pipeline.learn_one(x=x) print(f"ROCAUC: {metric.get():.4f}")

ROCAUC: 0.8813