Is One Epoch All You Need For Multi-Fidelity Hyperparameter Optimization?

TL;DR

This paper shows that a simple one-epoch baseline for multi-fidelity hyperparameter optimization performs comparably to complex methods, highlighting the importance of benchmarking practices and the presence of dominant learning curves.

Contribution

The paper demonstrates that a straightforward one-epoch discarding baseline is as effective as advanced MF-HPO methods on classical benchmarks, challenging current assumptions.

Findings

The baseline achieves similar results to complex MF-HPO methods.

Dominant learning curves explain the baseline's success.

Researchers should include this baseline in future benchmarks.

Abstract

Hyperparameter optimization (HPO) is crucial for fine-tuning machine learning models but can be computationally expensive. To reduce costs, Multi-fidelity HPO (MF-HPO) leverages intermediate accuracy levels in the learning process and discards low-performing models early on. We compared various representative MF-HPO methods against a simple baseline on classical benchmark data. The baseline involved discarding all models except the Top-K after training for only one epoch, followed by further training to select the best model. Surprisingly, this baseline achieved similar results to its counterparts, while requiring an order of magnitude less computation. Upon analyzing the learning curves of the benchmark data, we observed a few dominant learning curves, which explained the success of our baseline. This suggests that researchers should (1) always use the suggested baseline in benchmarks and (2) broaden the diversity of MF-HPO benchmarks to include more complex cases.