Saved in:
Bibliographic Details
Main Authors: Zhang, Tunhou, Li, Shiyu, Cheng, Hsin-Pai, Yan, Feng, Li, Hai, Chen, Yiran
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2404.17152
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910890527293440
author Zhang, Tunhou
Li, Shiyu
Cheng, Hsin-Pai
Yan, Feng
Li, Hai
Chen, Yiran
author_facet Zhang, Tunhou
Li, Shiyu
Cheng, Hsin-Pai
Yan, Feng
Li, Hai
Chen, Yiran
contents Exploring dense connectivity of convolutional operators establishes critical "synapses" to communicate feature vectors from different levels and enriches the set of transformations on Computer Vision applications. Yet, even with heavy-machinery approaches such as Neural Architecture Search (NAS), discovering effective connectivity patterns requires tremendous efforts due to either constrained connectivity design space or a sub-optimal exploration process induced by an unconstrained search space. In this paper, we propose CSCO, a novel paradigm that fabricates effective connectivity of convolutional operators with minimal utilization of existing design motifs and further utilizes the discovered wiring to construct high-performing ConvNets. CSCO guides the exploration via a neural predictor as a surrogate of the ground-truth performance. We introduce Graph Isomorphism as data augmentation to improve sample efficiency and propose a Metropolis-Hastings Evolutionary Search (MH-ES) to evade locally optimal architectures and advance search quality. Results on ImageNet show ~0.6% performance improvement over hand-crafted and NAS-crafted dense connectivity. Our code is publicly available.
format Preprint
id arxiv_https___arxiv_org_abs_2404_17152
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle CSCO: Connectivity Search of Convolutional Operators
Zhang, Tunhou
Li, Shiyu
Cheng, Hsin-Pai
Yan, Feng
Li, Hai
Chen, Yiran
Computer Vision and Pattern Recognition
Exploring dense connectivity of convolutional operators establishes critical "synapses" to communicate feature vectors from different levels and enriches the set of transformations on Computer Vision applications. Yet, even with heavy-machinery approaches such as Neural Architecture Search (NAS), discovering effective connectivity patterns requires tremendous efforts due to either constrained connectivity design space or a sub-optimal exploration process induced by an unconstrained search space. In this paper, we propose CSCO, a novel paradigm that fabricates effective connectivity of convolutional operators with minimal utilization of existing design motifs and further utilizes the discovered wiring to construct high-performing ConvNets. CSCO guides the exploration via a neural predictor as a surrogate of the ground-truth performance. We introduce Graph Isomorphism as data augmentation to improve sample efficiency and propose a Metropolis-Hastings Evolutionary Search (MH-ES) to evade locally optimal architectures and advance search quality. Results on ImageNet show ~0.6% performance improvement over hand-crafted and NAS-crafted dense connectivity. Our code is publicly available.
title CSCO: Connectivity Search of Convolutional Operators
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2404.17152