You want a first model baseline with little manual effort. What do you use?
Why D is the answer
Automated machine learning
AutoML tries algorithms and settings to explore good models.
In detail
- Automated machine learning tries algorithms and settings for you.
- Use the baseline to decide if custom work is worth it.
Use case
An illustrative scenario, not a real incident.
A team needs a first baseline quickly. Automated machine learning tries algorithms and gives a benchmark to beat.
See it in different ways
mindmap
root((Reproducible run))
Code
Git
Data
Versioned data assets
Environment
Registered environment
Compute
Cluster that scales to zero
Results
MLflowflowchart LR S0["Provide data"] --> S1["Automated ML"] --> S2["Compare models"] --> S3["Baseline"]
Key point: AutoML gives a quick baseline.
Why the other options are wrong
- A. A hand-written training loop for each algorithm
- That is the manual effort you want to avoid.
- B. A batch endpoint with no registered model
- A batch endpoint needs a model already.
- C. A fine-tuning job on a language model
- Fine-tuning a language model is a different task.