Deep Model Merging: The Sister of Neural Network Interpretability -- A Survey

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Khan, Arham, Nief, Todd, Hudson, Nathaniel, Sakarvadia, Mansi, Grzenda, Daniel, Ajith, Aswathy, Pettyjohn, Jordan, Chard, Kyle, Foster, Ian
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916659337363456
author Khan, Arham
Nief, Todd
Hudson, Nathaniel
Sakarvadia, Mansi
Grzenda, Daniel
Ajith, Aswathy
Pettyjohn, Jordan
Chard, Kyle
Foster, Ian
author_facet Khan, Arham
Nief, Todd
Hudson, Nathaniel
Sakarvadia, Mansi
Grzenda, Daniel
Ajith, Aswathy
Pettyjohn, Jordan
Chard, Kyle
Foster, Ian
contents We survey the model merging literature through the lens of loss landscape geometry to connect observations from empirical studies on model merging and loss landscape analysis to phenomena that govern neural network training and the emergence of their inner representations. We distill repeated empirical observations from the literature in these fields into descriptions of four major characteristics of loss landscape geometry: mode convexity, determinism, directedness, and connectivity. We argue that insights into the structure of learned representations from model merging have applications to model interpretability and robustness, subsequently we propose promising new research directions at the intersection of these fields.
format Preprint
id arxiv_https___arxiv_org_abs_2410_12927
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Deep Model Merging: The Sister of Neural Network Interpretability -- A Survey
Khan, Arham
Nief, Todd
Hudson, Nathaniel
Sakarvadia, Mansi
Grzenda, Daniel
Ajith, Aswathy
Pettyjohn, Jordan
Chard, Kyle
Foster, Ian
Machine Learning
Artificial Intelligence
We survey the model merging literature through the lens of loss landscape geometry to connect observations from empirical studies on model merging and loss landscape analysis to phenomena that govern neural network training and the emergence of their inner representations. We distill repeated empirical observations from the literature in these fields into descriptions of four major characteristics of loss landscape geometry: mode convexity, determinism, directedness, and connectivity. We argue that insights into the structure of learned representations from model merging have applications to model interpretability and robustness, subsequently we propose promising new research directions at the intersection of these fields.
title Deep Model Merging: The Sister of Neural Network Interpretability -- A Survey
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2410.12927