CoChat

Learning With Sharing: An Edge-Optimized Incremental Learning Method for Deep Neural Networks

Open access
Hussain, Muhammad Awais, Huang, Shih-An, Tsai, Tsung-Han · 2023 · IEEE Transactions on Emerging Topics in Computing

The DOI matches its Crossref record; 39 of the 40 references it cites were found in a registry or an index.

Abstract

Incremental learning techniques aim to increase the capability of Deep Neural Network (DNN) model to add new classes in the pre-trained model. However, DNNs suffer from catastrophic forgetting during the incremental learning process. Existing incremental learning techniques try to reduce the effect of catastrophic forgetting by either using previous samples of data while adding new classes in the model or designing complex model architectures. This leads to high design complexity and memory requirements which make incremental learning impossible to implement on edge devices that have limited memory and computation resources. So, we propose a new incremental learning technique Learning with Sharing (LwS) based on the concept of transfer learning. The main aims of LwS are reduction in training complexity and storage memory requirements while achieving high accuracy during the incremental learning process. We perform cloning and sharing of full connected (FC) layers to add new classes in the model incrementally. Our proposed technique can preserve the knowledge of existing classes and add new classes without storing data from the previous classes. We show that our proposed technique outperforms the state-of-the-art techniques in accuracy comparison for Cifar-100, Caltech-101, and UCSD Birds datasets.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/tetc.2022.3210905

Cites

39 of 39 references verified0 citation errors · 0 not checkable

Statistics

Figures

FIGURE 1 In incremental learning, different data classes arrive at different times. Incremental model is able to learn about new classes with their arrival at any given time.
FIGURE 1 In incremental learning, different data classes arrive at different times. Incremental model is able to learn about new classes with their arrival at any given time.CC-BY · From the PDF
Image 2
CC-BY · From the PDF
FIGURE 2 Incremental learning method to accommodate new classes in the base network. (a) Y new classes are added to the model by cloning the whole DNN model expect output layers of base network. In the cloned network, FC layers and output layer are trained for Y classes (shown as blue box in target model). (b) In the inference stage, the newly trained FC layers block (FC Block 2) and the previously trained FC layers block (FC Block2) are combined with the frozen DNN model and combined inference is performed on all blocks.
FIGURE 2 Incremental learning method to accommodate new classes in the base network. (a) Y new classes are added to the model by cloning the whole DNN model expect output layers of base network. In the cloned network, FC layers and output layer are trained for Y classes (shown as blue box in target model). (b) In the inference stage, the newly trained FC layers block (FC Block 2) and the previously trained FC layers block (FC Block2) are combined with the frozen DNN model and combined inference is performed on all blocks.CC-BY · From the PDF
Image 4
CC-BY · From the PDF
FIGURE 3 The base model architecture used in the paper. The output layer of DNN model is removed and two fully connected layers with the size of N1 and N2 are inserted into the model. The number of output classes C depends upon the training data, which keeps changing according to different experiments setup.
FIGURE 3 The base model architecture used in the paper. The output layer of DNN model is removed and two fully connected layers with the size of N1 and N2 are inserted into the model. The number of output classes C depends upon the training data, which keeps changing according to different experiments setup.CC-BY · From the PDF
FIGURE 4 Validation accuracy results of Cifar-100 dataset using the proposed technique. Each base model's accuracy is compared with the incrementally learned model. Red curve represents the accuracy of base model at different number of classes. Green curve represents the accuracy of incrementally learned model.
FIGURE 4 Validation accuracy results of Cifar-100 dataset using the proposed technique. Each base model's accuracy is compared with the incrementally learned model. Red curve represents the accuracy of base model at different number of classes. Green curve represents the accuracy of incrementally learned model.CC-BY · From the PDF
Image 7
CC-BY · From the PDF
FIGURE 5 Increase in memory size for different models when multiple classes are added to the model incrementally. x -axis represents the number of added classes incrementally in DNN model and y -axis represent the increase in DNN model size due to addition of new parameters for incrementally added classes.
FIGURE 5 Increase in memory size for different models when multiple classes are added to the model incrementally. x -axis represents the number of added classes incrementally in DNN model and y -axis represent the increase in DNN model size due to addition of new parameters for incrementally added classes.CC-BY · From the PDF
Image 9
CC-BY · From the PDF
FIGURE 7 Accuracy comparison for Caltech-101 dataset between LwS and LwM [12]. Base model is trained with 10 classes and 10 classes are added to the models at each incremental step.
FIGURE 7 Accuracy comparison for Caltech-101 dataset between LwS and LwM [12]. Base model is trained with 10 classes and 10 classes are added to the models at each incremental step.CC-BY · From the PDF
FIGURE 8 AccuracycomparisonforCUB-200-2011datasetbetween LwS and related works. Base model is trained with 10 classes and 10classes are added to the model at each incremental step.
FIGURE 8 AccuracycomparisonforCUB-200-2011datasetbetween LwS and related works. Base model is trained with 10 classes and 10classes are added to the model at each incremental step.CC-BY · From the PDF
Image 12
CC-BY · From the PDF
Image 13
CC-BY · From the PDF
Image 14
CC-BY · From the PDF

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 39 references

Cited by

See all 11

Metadata

Authors
Hussain, M. A., Huang, S.-A., & Tsai, T.-H.
Affiliations
  • National Central University (TW)
  • National Central University, Taoyuan, Taiwan
Venue
IEEE Transactions on Emerging Topics in Computing
Published
2023-04-01
Volume
11
Issue
2
Pages
461-473
ISSN
  • 2168-6750 (ISSN-L)
  • 2376-4562
OpenAlex
W4312935497
Cited by
11
References
40
Citations by year
2025: 7 · 2024: 3 · 2023: 1
FWCI
0.90
Type
journal article
Language
en
Open access
Locations
IEEE Transactions on Emerging Topics in Computing (publishedVersion, cc-by)
Funders
Industrial Technology Research Institute
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
  • Physical Sciences › Computer Science › Artificial Intelligence › Data Stream Mining Techniques
Keywords
incremental learning, catastrophic forgetting, edge devices, transfer learning, memory-efficient training, deep neural network

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Learning With Sharing: An Edge-Optimized Incremental Learning Method for Deep Neural Networks. https://app.cochat.ai/papers/10.1109/tetc.2022.3210905
The paper
Hussain, M. A., Huang, S.-A., & Tsai, T.-H. (2023). Learning With Sharing: An Edge-Optimized Incremental Learning Method for Deep Neural Networks. IEEE Transactions on Emerging Topics in Computing. https://doi.org/10.1109/tetc.2022.3210905

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.