def load_autoencoder():
ae_kwargs = {}
ae_kwargs["latent_dim"] = 2
ae_kwargs["hidden_dim"] = [15, 7]
ae_kwargs["epochs"] = 14
ae_kwargs["batch_size"] = 128
ae = AutoencoderModel(in_train.shape[1], **ae_kwargs)
ae.load_model()
return ae
def load_autoencoder():
ae_kwargs = {}
ae_kwargs["latent_dim"] = 2
ae_kwargs["hidden_dim"] = [15, 7]
ae_kwargs["epochs"] = 14
ae_kwargs["batch_size"] = 128
ae = AutoencoderModel(in_train.shape[1], **ae_kwargs)
ae.load_model()
metrics = load_metrics("metrics/" + ae.model_name + "/metrics.json")
return ae, metrics
def train_autoencoder():
# Instantiate and Train Autoencoder
ae_kwargs = {}
ae_kwargs["latent_dim"] = 2
ae_kwargs["hidden_dim"] = [15, 7]
ae_kwargs["epochs"] = 14
ae_kwargs["batch_size"] = 128
# ae_kwargs["model_path"] = ae_model_path
ae = AutoencoderModel(in_train.shape[1], **ae_kwargs)
ae.train(in_train, in_test)
ae.save_model()
inlier_scores = ae.compute_anomaly_score(in_test)
outlier_scores = ae.compute_anomaly_score(out_test)
print(inlier_scores)
print(outlier_scores)
metrics = eval_utils.evaluate_model(
inlier_scores, outlier_scores, model_name="ae", show_plot=False)
print(metrics)
return metrics
# DATA.
#
# ###########################################################################
from models.ae import AutoencoderModel
import numpy as np
from utils.eval_utils import load_metrics
import json
ae_kwargs = {}
in_shape = 18
ae_kwargs["latent_dim"] = 2
ae_kwargs["hidden_dim"] = [15, 7]
ae_kwargs["epochs"] = 14
ae_kwargs["batch_size"] = 128
ae = AutoencoderModel(in_shape, **ae_kwargs)
ae.load_model()
metrics = load_metrics("metrics/" + ae.model_name + "/metrics.json")
def predict(args):
data = np.array(args)
if data.shape[1] != 18:
return {"status": "input data should have 18 features"}
scores = ae.compute_anomaly_score(data)
predictions = (scores > metrics["threshold"])
result = {"scores": scores.tolist(),
"predictions": list(predictions.tostring())
}