-
Online Continual Learning on Hierarchical Label Expansion
Authors:
Byung Hyun Lee,
Okchul Jung,
Jonghyun Choi,
Se Young Chun
Abstract:
Continual learning (CL) enables models to adapt to new tasks and environments without forgetting previously learned knowledge. While current CL setups have ignored the relationship between labels in the past task and the new task with or without small task overlaps, real-world scenarios often involve hierarchical relationships between old and new tasks, posing another challenge for traditional CL…
▽ More
Continual learning (CL) enables models to adapt to new tasks and environments without forgetting previously learned knowledge. While current CL setups have ignored the relationship between labels in the past task and the new task with or without small task overlaps, real-world scenarios often involve hierarchical relationships between old and new tasks, posing another challenge for traditional CL approaches. To address this challenge, we propose a novel multi-level hierarchical class incremental task configuration with an online learning constraint, called hierarchical label expansion (HLE). Our configuration allows a network to first learn coarse-grained classes, with data labels continually expanding to more fine-grained classes in various hierarchy depths. To tackle this new setup, we propose a rehearsal-based method that utilizes hierarchy-aware pseudo-labeling to incorporate hierarchical class information. Additionally, we propose a simple yet effective memory management and sampling strategy that selectively adopts samples of newly encountered classes. Our experiments demonstrate that our proposed method can effectively use hierarchy on our HLE setup to improve classification accuracy across all levels of hierarchies, regardless of depth and class imbalance ratio, outperforming prior state-of-the-art works by significant margins while also outperforming them on the conventional disjoint, blurry and i-Blurry CL setups.
△ Less
Submitted 28 August, 2023;
originally announced August 2023.
-
Adaptive GLCM sampling for transformer-based COVID-19 detection on CT
Authors:
Okchul Jung,
Dong Un Kang,
Gwanghyun Kim,
Se Young Chun
Abstract:
The world has suffered from COVID-19 (SARS-CoV-2) for the last two years, causing much damage and change in people's daily lives. Thus, automated detection of COVID-19 utilizing deep learning on chest computed tomography (CT) scans became promising, which helps correct diagnosis efficiently. Recently, transformer-based COVID-19 detection method on CT is proposed to utilize 3D information in CT vol…
▽ More
The world has suffered from COVID-19 (SARS-CoV-2) for the last two years, causing much damage and change in people's daily lives. Thus, automated detection of COVID-19 utilizing deep learning on chest computed tomography (CT) scans became promising, which helps correct diagnosis efficiently. Recently, transformer-based COVID-19 detection method on CT is proposed to utilize 3D information in CT volume. However, its sampling method for selecting slices is not optimal. To leverage rich 3D information in CT volume, we propose a transformer-based COVID-19 detection using a novel data curation and adaptive sampling method using gray level co-occurrence matrices (GLCM). To train the model which consists of CNN layer, followed by transformer architecture, we first executed data curation based on lung segmentation and utilized the entropy of GLCM value of every slice in CT volumes to select important slices for the prediction. The experimental results show that the proposed method improve the detection performance with large margin without much difficult modification to the model.
△ Less
Submitted 4 July, 2022;
originally announced July 2022.
-
Jet-driven viscous locomotion of confined thermoresponsive microgels
Authors:
Ivan Tanasijević,
Oliver Jung,
Lyndon Koens,
Ahmed Mourran,
Eric Lauga
Abstract:
We consider the dynamics of micro-sized, asymmetrically-coated thermoresponsive hydrogel ribbons (microgels) under periodic heating and cooling in the confined space between two planar surfaces. As the result of the temperature changes, the volume and thus the shape of the slender microgel change, which lead to repeated cycles of bending and elastic relaxation, and to net locomotion. Small devices…
▽ More
We consider the dynamics of micro-sized, asymmetrically-coated thermoresponsive hydrogel ribbons (microgels) under periodic heating and cooling in the confined space between two planar surfaces. As the result of the temperature changes, the volume and thus the shape of the slender microgel change, which lead to repeated cycles of bending and elastic relaxation, and to net locomotion. Small devices designed for biomimetic locomotion need to exploit flows that are not symmetric in time (non-reciprocal) to escape the constraints of the scallop theorem and undergo net motion. Unlike other biological slender swimmers, the non-reciprocal bending of the gel centreline is not sufficient here to explain for the overall swimming motion. We show instead that the swimming of the gel results from the flux of water periodically emanating from (or entering) the gel itself due to its shrinking (or swelling). The associated flows induce viscous stresses that lead to a net propulsive force on the gel. We derive a theoretical model for this hypothesis of jet-driven propulsion, which leads to excellent agreement with our experiments.
△ Less
Submitted 8 February, 2022;
originally announced February 2022.
