Erratum to: Minimizing finite sums with the stochastic average gradient

Mark Schmidt, Nicolas Le Roux, Francis R. Bach

TL;DR

This erratum corrects a minor but important detail in a previous publication: the spelling of the second author's first name. The original paper, "Minimizing finite sums with the stochastic average gradient," likely presented methods for optimizing functions that are sums of many smaller functions, a common problem in machine learning and other fields. This correction ensures proper attribution and clarity for future references.

What Problem Does This Paper Address?

This document is an erratum, meaning it's a correction to a previously published paper. The original paper, "Minimizing finite sums with the stochastic average gradient," likely addressed the challenge of efficiently optimizing objective functions that are composed of a large number of individual, simpler functions. This is a prevalent problem in areas like machine learning, where training models often involves minimizing a loss function that is an average over a vast dataset. Standard optimization techniques can be computationally prohibitive in such scenarios, necessitating specialized algorithms like stochastic gradient descent (SGD) and its variants, such as stochastic average gradient (SAG). The core problem the original paper aimed to solve was likely developing or analyzing methods to speed up this optimization process.

Methods

As this is an erratum, it does not introduce new methods. It serves to correct a factual error in the original publication. The methods discussed in the original paper, which this erratum refers to, would have likely involved algorithms for optimizing large-scale finite sums. This could include variations of stochastic gradient descent, such as the Stochastic Average Gradient (SAG) method, or other related techniques designed to handle the computational burden of large datasets. The novelty would have been in the specific algorithmic improvements, theoretical analysis, or convergence guarantees provided for these methods.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • The Importance of Accurate Attribution: This erratum highlights how even seemingly small errors, like a misspelling of a name, can impact the clarity and integrity of scientific communication. Ensuring correct author names is crucial for proper citation and recognition of contributions.
  • The Need for Precision in Scientific Records: The correction underscores the importance of maintaining accurate records in scientific publications. This ensures that future researchers can reliably refer back to the original work and its authors.

Key Results

The key "result" of this erratum is the correction of a factual error: the first name of the second author was misspelled as "Nicholas" in the original publication and should be "Nicolas." The original article has reportedly been updated to reflect this correction.

Limitations and Future Work

Not discussed in the paper.

Summary updated

BibTeX
@article{schmidt2017erratum,
  author = {Mark Schmidt and Nicolas Le Roux and Francis R. Bach},
  title = {{Erratum to: Minimizing finite sums with the stochastic average gradient}},
  journal = {Math. Program.},
  year = {2017},
  doi = {10.1007/s10107-016-1051-1},
}
Figure