# Metrics and evaluation

Choose metrics according to the task and the meaning of the model outputs. Accuracy, signal reconstruction error and model compute describe different properties; none substitutes for the others.

## Select a metric

| Task | API | Interpretation |
|---|---|---|
| Classification with extra spatial or temporal axes | [MultiF1Score](https://ambiqai.github.io/helia-edge/reference/api/helia_edge/metrics/fscore/) | Collapses leading dimensions and treats the last dimension as classes. |
| Classification error analysis | [ConfusionMatrix](https://ambiqai.github.io/helia-edge/reference/api/helia_edge/metrics/confusion_matrix/) | Inspect errors between classes, alongside aggregate scores. |
| Signal reconstruction | [Snr](https://ambiqai.github.io/helia-edge/reference/api/helia_edge/metrics/snr/) | Signal-to-noise ratio in dB, with the reference signal as `y_true`. |
| Reconstruction distortion | [PRD and TruePRD](https://ambiqai.github.io/helia-edge/reference/api/helia_edge/metrics/prd/) | Check the denominator and normalization definition before comparing values. |
| Confidence filtering | [Threshold helpers](https://ambiqai.github.io/helia-edge/reference/api/helia_edge/metrics/threshold/) | Select predictions according to their probability threshold. |
| Static compute analysis | [get_flops](https://ambiqai.github.io/helia-edge/reference/api/helia_edge/metrics/flops/) | TensorFlow-based operation counting; it does not measure device latency. |

For a reconstruction task, the standalone SNR metric can be updated from held-out reference and predicted signals:

```python
from helia_edge.metrics import Snr

metric = Snr()
metric.update_state(reference_signals, predicted_signals)
print(metric.result())
metric.reset_state()
```

The arrays must represent aligned samples with the same signal interpretation. Consult each metric's API for weighting and reduction behavior.

## Plot the result

In a source checkout, run `uv sync --extra plotting` alongside your backend extra to visualize training progress and classification results.

[Training history](https://ambiqai.github.io/helia-edge/reference/api/helia_edge/plotting/history/)
[Confusion matrices](https://ambiqai.github.io/helia-edge/reference/api/helia_edge/plotting/cm/)
[ROC curves](https://ambiqai.github.io/helia-edge/reference/api/helia_edge/plotting/roc/)

Keep task metrics from held-out data separate from conversion checks. After export, compare converted predictions with the source model, then evaluate the converted model on your task's validation data.

[Export and quantization](https://ambiqai.github.io/helia-edge/guide/export/)