Denoising diffusion models (DDMs) have recently attracted increasing attention by showing impressive synthesis quality. DDMs are built on a diffusion process that pushes data to the noise distribution and the models learn to denoise. In this paper, we establish the interpretation of DDMs in terms of image restoration (IR). Integrating IR literature allows us to use an alternative objective and diverse forward processes, not confining to the diffusion process. By imposing prior knowledge on the loss function grounded on MAP-based estimation, we eliminate the need for the expensive sampling of DDMs. Also, we propose a multi-scale training, which improves the performance compared to the diffusion process, by taking advantage of the flexibility of the forward process. Experimental results demonstrate that our model improves the quality and efficiency of both training and inference. Furthermore, we show the applicability of our model to inverse problems. We believe that our framework paves the way for designing a new type of flexible general generative model.
Learning underlying dynamics from data is important and challenging in many real-world scenarios. Incorporating differential equations (DEs) to design continuous networks has drawn much attention recently, however, most prior works make specific assumptions on the type of DEs, making the model specialized for particular problems. This work presents a partial differential equation (PDE) based framework which improves the dynamics modeling capability. Building upon the recent Fourier neural operator, we propose a neural operator that can handle time continuously without requiring iterative operations or specific grids of temporal discretization. A theoretical result demonstrating its universality is provided. We also uncover an intrinsic property of neural operators that improves data efficiency and model generalization by ensuring stability. Our model achieves superior accuracy in dealing with time-dependent PDEs compared to existing models. Furthermore, several numerical pieces of evidence validate that our method better represents a wide range of dynamics and outperforms state-of-the-art DE-based models in real-time-series applications. Our framework opens up a new way for a continuous representation of neural networks that can be readily adopted for real-world applications.
A recurrent neural network (RNN) is a widely used deep-learning network for dealing with sequential data. Imitating a dynamical system, an infinite-width RNN can approximate any open dynamical system in a compact domain. In general, deep networks with bounded widths are more effective than wide networks in practice; however, the universal approximation theorem for deep narrow structures has yet to be extensively studied. In this study, we prove the universality of deep narrow RNNs and show that the upper bound of the minimum width for universality can be independent of the length of the data. Specifically, we show that a deep RNN with ReLU activation can approximate any continuous function or $L^p$ function with the widths $d_x+d_y+2$ and $\max\{d_x+1,d_y\}$, respectively, where the target function maps a finite sequence of vectors in $\mathbb{R}^{d_x}$ to a finite sequence of vectors in $\mathbb{R}^{d_y}$. We also compute the additional width required if the activation function is $\tanh$ or more. In addition, we prove the universality of other recurrent networks, such as bidirectional RNNs. Bridging a multi-layer perceptron and an RNN, our theory and proof technique can be an initial step toward further research on deep RNNs.
Universal approximation, whether a set of functions can approximate an arbitrary function in a specific function space, has been actively studied in recent years owing to the significant development of neural networks. However, despite its extensive use, research on the universal properties of the convolutional neural network has been limited due to its complex nature. In this regard, we demonstrate the universal approximation theorem for convolutional neural networks. A convolution with padding outputs the data of the same shape as the input data; therefore, it is necessary to prove whether a convolutional neural network composed of convolutions can approximate such a function. We have shown that convolutional neural networks can approximate continuous functions whose input and output values have the same shape. In addition, the minimum depth of the neural network required for approximation was presented, and we proved that it is the optimal value. We also verified that convolutional neural networks with sufficiently deep layers have universality when the number of channels is limited.
The ideally disentangled latent space in GAN involves the global representation of latent space using semantic attribute coordinates. In other words, in this disentangled space, there exists the global semantic basis as a vector space where each basis component describes one attribute of generated images. In this paper, we propose an unsupervised method for finding this global semantic basis in the intermediate latent space in GANs. This semantic basis represents sample-independent meaningful perturbations that change the same semantic attribute of an image on the entire latent space. The proposed global basis, called Fr\'echet basis, is derived by introducing Fr\'echet mean to the local semantic perturbations in a latent space. Fr\'echet basis is discovered in two stages. First, the global semantic subspace is discovered by the Fr\'echet mean in the Grassmannian manifold of the local semantic subspaces. Second, Fr\'echet basis is found by optimizing a basis of the semantic subspace via the Fr\'echet mean in the Special Orthogonal Group. Experimental results demonstrate that Fr\'echet basis provides better semantic factorization and robustness compared to the previous methods. Moreover, we suggest the basis refinement scheme for the previous methods. The quantitative experiments show that the refined basis achieves better semantic factorization while generating the same semantic subspace as the previous method.
A Fourier neural operator (FNO) is one of the physics-inspired machine learning methods. In particular, it is a neural operator. In recent times, several types of neural operators have been developed, e.g., deep operator networks, GNO, and MWTO. Compared with other models, the FNO is computationally efficient and can learn nonlinear operators between function spaces independent of a certain finite basis. In this study, we investigated the bounding of the Rademacher complexity of the FNO based on specific group norms. Using capacity based on these norms, we bound the generalization error of the FNO model. In addition, we investigated the correlation between the empirical generalization error and the proposed capacity of FNO. Based on this investigation, we gained insight into the impact of the model architecture on the generalization error and estimated the amount of information about FNO models stored in various types of capacities.
To boost the performance, deep neural networks require deeper or wider network structures that involve massive computational and memory costs. To alleviate this issue, the self-knowledge distillation method regularizes the model by distilling the internal knowledge of the model itself. Conventional self-knowledge distillation methods require additional trainable parameters or are dependent on the data. In this paper, we propose a simple and effective self-knowledge distillation method using a dropout (SD-Dropout). SD-Dropout distills the posterior distributions of multiple models through a dropout sampling. Our method does not require any additional trainable modules, does not rely on data, and requires only simple operations. Furthermore, this simple method can be easily combined with various self-knowledge distillation approaches. We provide a theoretical and experimental analysis of the effect of forward and reverse KL-divergences in our work. Extensive experiments on various vision tasks, i.e., image classification, object detection, and distribution shift, demonstrate that the proposed method can effectively improve the generalization of a single network. Further experiments show that the proposed method also improves calibration performance, adversarial robustness, and out-of-distribution detection ability.
The impressive success of style-based GANs (StyleGANs) in high-fidelity image synthesis has motivated research to understand the semantic properties of their latent spaces. Recently, a close relationship was observed between the semantically disentangled local perturbations and the local PCA components in the learned latent space $\mathcal{W}$. However, understanding the number of disentangled perturbations remains challenging. Building upon this observation, we propose a local dimension estimation algorithm for an arbitrary intermediate layer in a pre-trained GAN model. The estimated intrinsic dimension corresponds to the number of disentangled local perturbations. In this perspective, we analyze the intermediate layers of the mapping network in StyleGANs. Our analysis clarifies the success of $\mathcal{W}$-space in StyleGAN and suggests an alternative. Moreover, the intrinsic dimension estimation opens the possibility of unsupervised evaluation of global-basis-compatibility and disentanglement for a latent space. Our proposed metric, called Distortion, measures an inconsistency of intrinsic tangent space on the learned latent space. The metric is purely geometric and does not require any additional attribute information. Nevertheless, the metric shows a high correlation with the global-basis-compatibility and supervised disentanglement score. Our findings pave the way towards an unsupervised selection of globally disentangled latent space among the intermediate latent spaces in a GAN.
Out-of-distribution (OOD) detection is an important task in machine learning systems for ensuring their reliability and safety. Deep probabilistic generative models facilitate OOD detection by estimating the likelihood of a data sample. However, such models frequently assign a suspiciously high likelihood to a specific outlier. Several recent works have addressed this issue by training a neural network with auxiliary outliers, which are generated by perturbing the input data. In this paper, we discover that these approaches fail for certain OOD datasets. Thus, we suggest a new detection metric that operates without outlier exposure. We observe that our metric is robust to diverse variations of an image compared to the previous outlier-exposing methods. Furthermore, our proposed score requires neither auxiliary models nor additional training. Instead, this paper utilizes the likelihood ratio statistic in a new perspective to extract genuine properties from the given single deep probabilistic generative model. We also apply a novel numerical approximation to enable fast implementation. Finally, we demonstrate comprehensive experiments on various probabilistic generative models and show that our method achieves state-of-the-art performance.
In this paper, we propose a method to find local-geometry-aware traversal directions on the intermediate latent space of Generative Adversarial Networks (GANs). These directions are defined as an ordered basis of tangent space at a latent code. Motivated by the intrinsic sparsity of the latent space, the basis is discovered by solving the low-rank approximation problem of the differential of the partial network. Moreover, the local traversal basis leads to a natural iterative traversal on the latent space. Iterative Curve-Traversal shows stable traversal on images, since the trajectory of latent code stays close to the latent space even under the strong perturbations compared to the linear traversal. This stability provides far more diverse variations of the given image. Although the proposed method can be applied to various GAN models, we focus on the W-space of the StyleGAN2, which is renowned for showing the better disentanglement of the latent factors of variation. Our quantitative and qualitative analysis provides evidence showing that the W-space is still globally warped while showing a certain degree of global consistency of interpretable variation. In particular, we introduce some metrics on the Grassmannian manifolds to quantify the global warpage of the W-space and the subspace traversal to test the stability of traversal directions.