# How-to guides

How-to guides answer a specific job you already know you want to do. They
assume you have finished Quickstart so that the package installs cleanly and
the stage names in the logs are familiar. If you have never run ARCADE at all,
return to {doc}`../quickstart/index` first, then come back here with a
concrete goal.

Each page below is procedural. It tells you which YAML keys to set, which
commands to run, and what artifact to inspect when you finish. For design
intent and metric definitions, open the sibling Explanation section under
Guides in the sidebar, starting from {doc}`../explanation/index`.

```{toctree}
:maxdepth: 1

compare_classifiers
add_classifier
enable_optimization
enable_qec
```

## Choose a guide by goal

| Goal | Guide |
|------|-------|
| Rank methods on your own HDF5 under one frozen recipe | {doc}`compare_classifiers` |
| Build your own classifier, including a transformer-style custom class | {doc}`add_classifier` |
| Turn on pruning, distillation, NAS, quantization, or Optuna tuning | {doc}`enable_optimization` |
| Map classifier accuracy into logical error rate and cycle time | {doc}`enable_qec` |

### Build your own classifier

The {doc}`add_classifier` guide is the extension path. It shows how to register
a class, how to point YAML at a registered name or a dotted import path, and how
to plug a custom transformer into the same report as zoo baselines. Start from
the nearest-mean toy if you only need filter features. Jump to the transformer
section when your method needs time-resolved sequences.

When you need exact field names or registry strings, leave Guides and open the
API Reference. When you want a paced first sitting rather than a job recipe,
open Tutorials after you have skimmed Quickstart.