Semantic Scholar Open Access 2021 184 sitasi

Accounting for Variance in Machine Learning Benchmarks

Xavier Bouthillier Pierre Delaunay Mirko Bronzi Assya Trofimov B. Nichyporuk +12 lainnya

Abstrak

Strong empirical evidence that one machine-learning algorithm A outperforms another one B ideally calls for multiple trials optimizing the learning pipeline over sources of variation such as data sampling, data augmentation, parameter initialization, and hyperparameters choices. This is prohibitively expensive, and corners are cut to reach conclusions. We model the whole benchmarking process, revealing that variance due to data sampling, parameter initialization and hyperparameter choice impact markedly the results. We analyze the predominant comparison methods used today in the light of this variance. We show a counter-intuitive result that adding more sources of variation to an imperfect estimator approaches better the ideal estimator at a 51 times reduction in compute cost. Building on these results, we study the error rate of detecting improvements, on five different deep-learning tasks/architectures. This study leads us to propose recommendations for performance comparisons.

Penulis (17)

X

Xavier Bouthillier

P

Pierre Delaunay

M

Mirko Bronzi

A

Assya Trofimov

B

B. Nichyporuk

J

Justin Szeto

N

Naz Sepah

E

Edward Raff

K

Kanika Madan

V

Vikram S. Voleti

S

Samira Ebrahimi Kahou

V

Vincent Michalski

D

Dmitriy Serdyuk

T

T. Arbel

C

C. Pal

G

G. Varoquaux

P

Pascal Vincent

Format Sitasi

Bouthillier, X., Delaunay, P., Bronzi, M., Trofimov, A., Nichyporuk, B., Szeto, J. et al. (2021). Accounting for Variance in Machine Learning Benchmarks. https://www.semanticscholar.org/paper/9ceae85a0bd4231cd2efe14884c40b7bc04d3dac

Akses Cepat

PDF tidak tersedia langsung

Cek di sumber asli →
Lihat di Sumber
Informasi Jurnal
Tahun Terbit
2021
Bahasa
en
Total Sitasi
184×
Sumber Database
Semantic Scholar
Akses
Open Access ✓