2023-04-14 更新
Physics-informed radial basis network (PIRBN): A local approximation neural network for solving nonlinear PDEs
Authors:Jinshuai Bai, Gui-Rong Liu, Ashish Gupta, Laith Alzubaidi, Xi-Qiao Feng, YuanTong Gu
Our recent intensive study has found that physics-informed neural networks (PINN) tend to be local approximators after training. This observation leads to this novel physics-informed radial basis network (PIRBN), which can maintain the local property throughout the entire training process. Compared to deep neural networks, a PIRBN comprises of only one hidden layer and a radial basis “activation” function. Under appropriate conditions, we demonstrated that the training of PIRBNs using gradient descendent methods can converge to Gaussian processes. Besides, we studied the training dynamics of PIRBN via the neural tangent kernel (NTK) theory. In addition, comprehensive investigations regarding the initialisation strategies of PIRBN were conducted. Based on numerical examples, PIRBN has been demonstrated to be more effective and efficient than PINN in solving PDEs with high-frequency features and ill-posed computational domains. Moreover, the existing PINN numerical techniques, such as adaptive learning, decomposition and different types of loss functions, are applicable to PIRBN. The programs that can regenerate all numerical results can be found at https://github.com/JinshuaiBai/PIRBN.
PDF 48 pages, 26 figures
点此查看论文截图
Out-of-distribution Few-shot Learning For Edge Devices without Model Fine-tuning
Authors:Xinyun Zhang, Lanqing Hong
Few-shot learning (FSL) via customization of a deep learning network with limited data has emerged as a promising technique to achieve personalized user experiences on edge devices. However, existing FSL methods primarily assume independent and identically distributed (IID) data and utilize either computational backpropagation updates for each task or a common model with task-specific prototypes. Unfortunately, the former solution is infeasible for edge devices that lack on-device backpropagation capabilities, while the latter often struggles with limited generalization ability, especially for out-of-distribution (OOD) data. This paper proposes a lightweight, plug-and-play FSL module called Task-aware Normalization (TANO) that enables efficient and task-aware adaptation of a deep neural network without backpropagation. TANO covers the properties of multiple user groups by coordinating the updates of several groups of the normalization statistics during meta-training and automatically identifies the appropriate normalization group for a downstream few-shot task. Consequently, TANO provides stable but task-specific estimations of the normalization statistics to close the distribution gaps and achieve efficient model adaptation. Results on both intra-domain and out-of-domain generalization experiments demonstrate that TANO outperforms recent methods in terms of accuracy, inference speed, and model size. Moreover, TANO achieves promising results on widely-used FSL benchmarks and data from real applications.
PDF
点此查看论文截图
CoSDA: Continual Source-Free Domain Adaptation
Authors:Haozhe Feng, Zhaorui Yang, Hesun Chen, Tianyu Pang, Chao Du, Minfeng Zhu, Wei Chen, Shuicheng Yan
Without access to the source data, source-free domain adaptation (SFDA) transfers knowledge from a source-domain trained model to target domains. Recently, SFDA has gained popularity due to the need to protect the data privacy of the source domain, but it suffers from catastrophic forgetting on the source domain due to the lack of data. To systematically investigate the mechanism of catastrophic forgetting, we first reimplement previous SFDA approaches within a unified framework and evaluate them on four benchmarks. We observe that there is a trade-off between adaptation gain and forgetting loss, which motivates us to design a consistency regularization to mitigate forgetting. In particular, we propose a continual source-free domain adaptation approach named CoSDA, which employs a dual-speed optimized teacher-student model pair and is equipped with consistency learning capability. Our experiments demonstrate that CoSDA outperforms state-of-the-art approaches in continuous adaptation. Notably, our CoSDA can also be integrated with other SFDA methods to alleviate forgetting.
PDF 15 pages, 6 figures
点此查看论文截图
DiffFit: Unlocking Transferability of Large Diffusion Models via Simple Parameter-Efficient Fine-Tuning
Authors:Enze Xie, Lewei Yao, Han Shi, Zhili Liu, Daquan Zhou, Zhaoqiang Liu, Jiawei Li, Zhenguo Li
Diffusion models have proven to be highly effective in generating high-quality images. However, adapting large pre-trained diffusion models to new domains remains an open challenge, which is critical for real-world applications. This paper proposes DiffFit, a parameter-efficient strategy to fine-tune large pre-trained diffusion models that enable fast adaptation to new domains. DiffFit is embarrassingly simple that only fine-tunes the bias term and newly-added scaling factors in specific layers, yet resulting in significant training speed-up and reduced model storage costs. Compared with full fine-tuning, DiffFit achieves 2$\times$ training speed-up and only needs to store approximately 0.12\% of the total model parameters. Intuitive theoretical analysis has been provided to justify the efficacy of scaling factors on fast adaptation. On 8 downstream datasets, DiffFit achieves superior or competitive performances compared to the full fine-tuning while being more efficient. Remarkably, we show that DiffFit can adapt a pre-trained low-resolution generative model to a high-resolution one by adding minimal cost. Among diffusion-based methods, DiffFit sets a new state-of-the-art FID of 3.02 on ImageNet 512$\times$512 benchmark by fine-tuning only 25 epochs from a public pre-trained ImageNet 256$\times$256 checkpoint while being 30$\times$ more training efficient than the closest competitor.
PDF Tech Report
点此查看论文截图