arXiv subjects

Yizhu Chen

Publications and source records attributed to Yizhu Chen.

6 recordsLinked to original sources

arXiv2025

REAL: Representation Enhanced Analytic Learning for Exemplar-free Class-incremental Learning

Exemplar-free class-incremental learning (EFCIL) aims to mitigate catastrophic forgetting in class-incremental learning (CIL) without available historical training samples as exemplars. Compared with its exemplar-based CIL counterpart that stores exemplars, EFCIL suffers more from forgetting issues. Recently, a new EFCIL branch named Analytic Continual Learning (ACL) introduces a gradient-free paradigm via Recursive Least-Square, achieving a forgetting-resistant classifier training with a frozen backbone during CIL. However, existing ACL suffers from ineffective representations and insufficient utilization of backbone knowledge. In this paper, we propose a representation-enhanced analytic learning (REAL) to address these problems. To enhance the representation, REAL constructs a dual-stream base pretraining followed by representation enhancing distillation process. The dual-stream base pretraining combines self-supervised contrastive learning for general features and supervised learning for class-specific knowledge, followed by the representation enhancing distillation to merge both streams, enhancing representations for subsequent CIL paradigm. To utilize more knowledge from the backbone, REAL presents a feature fusion buffer to multi-layer backbone features, providing informative features for the subsequent classifier training. Our method can be incorporated into existing ACL techniques and provides more competitive performance. Empirical results demonstrate that, REAL achieves state-of-the-art performance on CIFAR-100, ImageNet-100 and ImageNet-1k benchmarks, outperforming exemplar-free methods and rivaling exemplar-based approaches.

cs.LG↗

arXiv2025

Wave-Particle (Continuous-Discrete) Dualistic Visual Tokenization for Unified Understanding and Generation

The unification of understanding and generation within a single multi-modal large model (MLLM) remains one significant challenge, largely due to the dichotomy between continuous and discrete visual tokenizations. Continuous tokenizer (CT) achieves strong performance by bridging multiple independently-trained understanding modules and generation modules, but suffers from complex multi-stage pipelines and substantial engineering overhead. Conversely, discrete tokenizers (DT) offer a conceptually elegant idea by quantizing each image into a primitive, but inevitably leading to information loss and performance degradation. To resolve this tension, we question the binary choice between CT and DT, inspired by the wave-particle duality of light, and propose the Continuous-Discrete Dualistic Visual Tokenizer (CDD-VT). We treat visual data as a flexible composition of image primitives derived from quantized codebooks, with the crucial insight that the primitive number assigned to each visual sample is adaptively determined according to its complexity: simple instances use a few primitives, emulating discrete tokenization, while complex instances use many, approximating continuous tokenization. Two core components are designed: Diverse Quantitative Primitives, which encourage primitives orthogonality to better populate information space, and Dynamic Primitive Allocator, which assesses sample complexity to determine the optimal set of primitives. Extensive experiments on reconstruction, retrieval and classification show that CDD-VT achieves superior performance over to specialized CT and DT, effectively getting strong result within a concise and scalable MLLM.

cs.CV↗

arXiv2024

Graph Pooling by Local Cluster Selection

Graph pooling is a family of operations which take graphs as input and produce shrinked graphs as output. Modern graph pooling methods are trainable and, in general inserted in Graph Neural Networks (GNNs) architectures as graph shrinking operators along the (deep) processing pipeline. This work proposes a novel procedure for pooling graphs, along with a node-centred graph pooling operator.

cs.LG↗

arXiv2024

GACL: Exemplar-Free Generalized Analytic Continual Learning

Class incremental learning (CIL) trains a network on sequential tasks with separated categories in each task but suffers from catastrophic forgetting, where models quickly lose previously learned knowledge when acquiring new tasks. The generalized CIL (GCIL) aims to address the CIL problem in a more real-world scenario, where incoming data have mixed data categories and unknown sample size distribution. Existing attempts for the GCIL either have poor performance or invade data privacy by saving exemplars. In this paper, we propose a new exemplar-free GCIL technique named generalized analytic continual learning (GACL). The GACL adopts analytic learning (a gradient-free training technique) and delivers an analytical (i.e., closed-form) solution to the GCIL scenario. This solution is derived via decomposing the incoming data into exposed and unexposed classes, thereby attaining a weight-invariant property, a rare yet valuable property supporting an equivalence between incremental learning and its joint training. Such an equivalence is crucial in GCIL settings as data distributions among different tasks no longer pose challenges to adopting our GACL. Theoretically, this equivalence property is validated through matrix analysis tools. Empirically, we conduct extensive experiments where, compared with existing GCIL methods, our GACL exhibits a consistently leading performance across various datasets and GCIL settings. Source code is available at https://github.com/CHEN-YIZHU/GACL.

cs.LG↗

arXiv2020

3 kW passive-gain-enabled metalized Raman fiber amplifier based on passive gain

Raman fiber lasers (RFLs) are currently promising and versatile light sources for a variety of applications. So far, operations of high power and brightness-enhanced RFLs have absorbed enormous interests along with rapid progress. Nevertheless, the stable Raman lasing at high power levels remains challenged by the thermal effects. In an effort to realize more effective thermal management in high power RFLs, here we demonstrate, for the first time, an all-fiberized RFA employing metal-coated passive fiber enabling high power lasing. By employing aluminum to the cladding of graded-index (GRIN) passive fiber, the thermal abstraction of the laser devices is more sufficient to support low-temperature operation. The maximum output power reaches 3.083 kW at 1130 nm with a conversion efficiency of 78.7%. To the best of our knowledge, this is the first Raman laser generation based on metal-coated passive fiber. Meanwhile, it is also the highest power attained in the fields of all kinds of Raman lasers based on merely nonlinear gain.

physics.optics↗

arXiv2018

Pure passive fiber enabled highly efficient Raman fiber amplifier with record kilowatt power

Kilowatt-level high efficiency all-fiberized Raman fiber amplifier based on pure passive fiber is proposed for the first time in this paper. The laser system is established on master oscillator power amplification configuration while a piece of graded-index passive fiber is utilized as stokes shifting as well as gain medium, which is entirely irrelevant to rare-earth-doped gain mechanism. When the pump power is 1368.8 W, we obtained 1002.3 W continuous-wave laser power at 1060 nm with the corresponding optical-to-optical efficiency of 84%. The beam parameter M2 improves from 9.17 of the pump laser to 5.11 of the signal laser through the amplification process, and the brightness enhancement is about 2.57 at maximum output power as a consequence of the beam clean-up process in the graded-index fiber. To the best of our knowledge, we have demonstrated the first kilowatt-level high efficiency Raman fiber amplifier based on pure passive fiber with brightness enhancement.

physics.optics↗