StitchNet: Composing Neural Networks from Pre-Trained Fragments

We propose StitchNet, a novel neural network cre-ation paradigm that stitches together fragments (one or more consecutive network layers) from multiple pre-trained neural networks. StitchNet allows the creation of high-performing neural networks without the large compute and data requirements needed...

Full description

Saved in:
Bibliographic Details
Main Authors: Teerapittayanon, Surat, Comiter, Marcus, McDanel, Bradley, Kung, H.T.
Format: Conference Proceeding
Language:English
Subjects:
Centered Kernel Alignment (CKA)
Codes
Computational modeling
Data models
Deep Learning
Libraries
Machine learning
Neural networks
Reusable Network Compo-nents
StitchNet
Training
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We propose StitchNet, a novel neural network cre-ation paradigm that stitches together fragments (one or more consecutive network layers) from multiple pre-trained neural networks. StitchNet allows the creation of high-performing neural networks without the large compute and data requirements needed under traditional model creation processes via backprop-agation training. We leverage Centered Kernel Alignment (CKA) as a compatibility measure to efficiently guide the selection of these fragments in composing a network for a given task tailored to specific accuracy needs and computing resource constraints. We then show that these fragments can be stitched together to create neural networks with accuracy comparable to that of traditionally trained networks at a fraction of computing resource and data requirements. Finally, we explore a novel on-the-fly personalized model creation and inference application enabled by this new paradigm. The code is available at https://github.com/steerapi/stitchnet.
ISSN:1946-0759
DOI:10.1109/ICMLA58977.2023.00017