> For the complete documentation index, see [llms.txt](https://dashpay.gitbook.io/roboflux-whitepaper/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://dashpay.gitbook.io/roboflux-whitepaper/5.-anomaly-detection-subsystem/5.1-tensor-based-anomaly-detection.md).

# 5.1 Tensor-Based Anomaly Detection

Utilizing multi-layer convolutional recurrent neural networks (CRNNs) with attention mechanisms, the ADS processes high-dimensional time-series data streams to detect anomalous patterns.

**Model Architecture:**

* **Input Tensor Shape:** (Batch, Channels, Timesteps, Features)
* **Convolutional Layers:** Extract localized feature embeddings from multivariate sequences.
* **Bidirectional LSTM Stack:** Capture long-range temporal dependencies bidirectionally.
* **Self-Attention Layer:** Weight sequence elements by contextual relevance to anomaly likelihood.
* **Output Softmax Layer:** Classify sequences into normal and anomalous categories.