louis, ard a and Keinath-Esmail, zaman (2025) The Uses and Limitations of Occam Algorithms: a response to Herrmann. [Preprint]

Text The_Uses_and_Limitations_of_Occam_Algorithms__response_to_Herrmann__For_Philosophy_of_Science (5).pdf - Draft Version Available under License Creative Commons Attribution. Download (191kB)

Abstract

In a recent paper, Daniel Herrman uses probably approximately correct (PAC) learning theory to argue that Occam algorithms do not justify a preference for simpler hypotheses. He claims to derive equally efficient "Anti-Occam" algorithms favouring the most complex hypotheses. We argue that Herrmann's analysis omits key elements of Occam algorithms, which eliminate the possibility of "Anti-Occam" algorithms and counter many of his arguments. These elements clarify the intrinsic connection of Occam algorithms to theories of learnability. Occam algorithms are not a failed epistemic justification of Occam's razor but rather a pragmatic base for practical algorithms

---

Export/Citation:

EndNote | BibTeX | Dublin Core | ASCII/Text Citation (Chicago) | HTML Citation | OpenURL

Social Networking:

Share |

---

Item Type:	Preprint
Creators:	Creators Email ORCID louis, ard a ard.louis@physics.ox.ac.uk 0000-0002-8438-910X Keinath-Esmail, zaman zfkeinathesmail@gmail.com
Keywords:	PAC learning, Occam's razor, learning theory
Subjects:	General Issues > Formal Learning Theory Specific Sciences > Artificial Intelligence > Machine Learning
Depositing User:	Prof Ard Louis
Date Deposited:	17 Mar 2025 16:04
Last Modified:	17 Mar 2025 16:04
Item ID:	24910
Subjects:	General Issues > Formal Learning Theory Specific Sciences > Artificial Intelligence > Machine Learning
Date:	13 March 2025
URI:	https://philsci-archive.pitt.edu/id/eprint/24910

Monthly Views for the past 3 years

Monthly Downloads for the past 3 years

Plum Analytics

Actions (login required)

View Item