Kenji Fukumizu

Takeru Miyato, Masanori Koyama, Kenji Fukumizu

In this study, we present meta-sequential prediction (MSP), an unsupervised framework to learn the symmetry from the time sequence of length at least three. Our method leverages the stationary property (e.g. constant velocity, constant acceleration) of the time sequence to learn the underlying equivariant structure of the dataset by simply training the encoder-decoder model to be able to predict the future observations. We will demonstrate that, with our framework, the hidden disentangled structure of the dataset naturally emerges as a by-product by applying simultaneous block-diagonalization to the transition operators in the latent space, the procedure which is commonly used in representation theory to decompose the feature-space based on the type of response to group actions. We will showcase our method from both empirical and theoretical perspectives. Our result suggests that finding a simple structured relation and learning a model with extrapolation capability are two sides of the same coin. The code is available at https://github.com/takerum/meta_sequential_prediction.

Tam Le, Truyen Nguyen, Kenji Fukumizu

Optimal transport (OT) is a popular and powerful tool for comparing probability measures. However, OT suffers a few drawbacks: (i) input measures required to have the same mass, (ii) a high computational complexity, and (iii) indefiniteness which limits its applications on kernel-dependent algorithmic approaches. To tackle issues (ii)--(iii), Le et al. (2022) recently proposed Sobolev transport for measures on a graph having the same total mass by leveraging the graph structure over supports. In this work, we consider measures that may have different total mass and are supported on a graph metric space. To alleviate the disadvantages (i)--(iii) of OT, we propose a novel and scalable approach to extend Sobolev transport for this unbalanced setting where measures may have different total mass. We show that the proposed unbalanced Sobolev transport (UST) admits a closed-form formula for fast computation, and it is also negative definite. Additionally, we derive geometric structures for the UST and establish relations between our UST and other transport distances. We further exploit the negative definiteness to design positive definite kernels and evaluate them on various simulations to illustrate their fast computation and comparable performances against other transport baselines for unbalanced measures on a graph.

Tam Le, Truyen Nguyen, Kenji Fukumizu

We study optimal transport (OT) problem for probability measures supported on a tree metric space. It is known that such OT problem (i.e., tree-Wasserstein (TW)) admits a closed-form expression, but depends fundamentally on the underlying tree structure over supports of input measures. In practice, the given tree structure may be, however, perturbed due to noisy or adversarial measurements. To mitigate this issue, we follow the max-min robust OT approach which considers the maximal possible distances between two input measures over an uncertainty set of tree metrics. In general, this approach is hard to compute, even for measures supported in one-dimensional space, due to its non-convexity and non-smoothness which hinders its practical applications, especially for large-scale settings. In this work, we propose novel uncertainty sets of tree metrics from the lens of edge deletion/addition which covers a diversity of tree structures in an elegant framework. Consequently, by building upon the proposed uncertainty sets, and leveraging the tree structure over supports, we show that the robust OT also admits a closed-form expression for a fast computation as its counterpart standard OT (i.e., TW). Furthermore, we demonstrate that the robust OT satisfies the metric property and is negative definite. We then exploit its negative definiteness to propose positive definite kernels and test them in several simulations on various real-world datasets on document classification and topological data analysis.

Yuri Kinoshita, Kenta Oono, Kenji Fukumizu et al.

Variational autoencoders (VAEs) are one of the deep generative models that have experienced enormous success over the past decades. However, in practice, they suffer from a problem called posterior collapse, which occurs when the encoder coincides, or collapses, with the prior taking no information from the latent structure of the input data into consideration. In this work, we introduce an inverse Lipschitz neural network into the decoder and, based on this architecture, provide a new method that can control in a simple and clear manner the degree of posterior collapse for a wide range of VAE models equipped with a concrete theoretical guarantee. We also illustrate the effectiveness of our method through several numerical experiments.

Shunya Minami, Kenji Fukumizu, Yoshihiro Hayashi et al.

Supervised transfer learning has received considerable attention due to its potential to boost the predictive power of machine learning in scenarios where data are scarce. Generally, a given set of source models and a dataset from a target domain are used to adapt the pre-trained models to a target domain by statistically learning domain shift and domain-specific factors. While such procedurally and intuitively plausible methods have achieved great success in a wide range of real-world applications, the lack of a theoretical basis hinders further methodological development. This paper presents a general class of transfer learning regression called affine model transfer, following the principle of expected-square loss minimization. It is shown that the affine model transfer broadly encompasses various existing methods, including the most common procedure based on neural feature extractors. Furthermore, the current paper clarifies theoretical properties of the affine model transfer such as generalization error and excess risk. Through several case studies, we demonstrate the practical benefits of modeling and estimating inter-domain commonality and domain-specific factors separately with the affine-type transfer models.

Shoji Toyota, Kenji Fukumizu

Deep Neural Networks often inherit spurious correlations embedded in training data and hence may fail to generalize to unseen domains, which have different distributions from the domain to provide training data. M. Arjovsky et al. (2019) introduced the concept out-of-distribution (o.o.d.) risk, which is the maximum risk among all domains, and formulated the issue caused by spurious correlations as a minimization problem of the o.o.d. risk. Invariant Risk Minimization (IRM) is considered to be a promising approach to minimize the o.o.d. risk: IRM estimates a minimum of the o.o.d. risk by solving a bi-level optimization problem. While IRM has attracted considerable attention with empirical success, it comes with few theoretical guarantees. Especially, a solid theoretical guarantee that the bi-level optimization problem gives the minimum of the o.o.d. risk has not yet been established. Aiming at providing a theoretical justification for IRM, this paper rigorously proves that a solution to the bi-level optimization problem minimizes the o.o.d. risk under certain conditions. The result also provides sufficient conditions on distributions providing training data and on a dimension of feature space for the bi-leveled optimization problem to minimize the o.o.d. risk.

Masanori Koyama, Takeru Miyato, Kenji Fukumizu

Recent years have witnessed the effectiveness of contrastive learning in obtaining the representation of dataset that is useful in interpretation and downstream tasks. However, the mechanism that describes this effectiveness have not been thoroughly analyzed, and many studies have been conducted to investigate the data structures captured by contrastive learning. In particular, the recent study of \citet{content_isolate} has shown that contrastive learning is capable of decomposing the data space into the space that is invariant to all augmentations and its complement. In this paper, we introduce the notion of invariance-adapted latent space that decomposes the data space into the intersections of the invariant spaces of each augmentation and their complements. This decomposition generalizes the one introduced in \citet{content_isolate}, and describes a structure that is analogous to the frequencies in the harmonic analysis of a group. We experimentally show that contrastive learning with lasso-type metric can be used to find an invariance-adapted latent space, thereby suggesting a new potential for the contrastive learning. We also investigate when such a latent space can be identified up to mixings within each component.

Siddharth Vishwanath, Bharath K. Sriperumbudur, Kenji Fukumizu et al.

The distance function to a compact set plays a crucial role in the paradigm of topological data analysis. In particular, the sublevel sets of the distance function are used in the computation of persistent homology -- a backbone of the topological data analysis pipeline. Despite its stability to perturbations in the Hausdorff distance, persistent homology is highly sensitive to outliers. In this work, we develop a framework of statistical inference for persistent homology in the presence of outliers. Drawing inspiration from recent developments in robust statistics, we propose a \textit{median-of-means} variant of the distance function (\textsf{MoM Dist}) and establish its statistical properties. In particular, we show that, even in the presence of outliers, the sublevel filtrations and weighted filtrations induced by \textsf{MoM Dist} are both consistent estimators of the true underlying population counterpart and exhibit near minimax-optimal performance in adversarial settings. Finally, we demonstrate the advantages of the proposed methodology through simulations and applications.

Shoji Toyota, Kenji Fukumizu

Deep Neural Networks inherit spurious correlations embedded in training data and hence may fail to predict desired labels on unseen domains (or environments), which have different distributions from the domain used in training. Invariance Learning (IL) has been developed recently to overcome this shortcoming; using training data in many domains, IL estimates such a predictor that is invariant to a change of domain. However, the requirement of training data in multiple domains is a strong restriction of IL, since it often needs high annotation cost. We propose a novel IL framework to overcome this problem. Assuming the availability of data from multiple domains for a higher level of classification task, for which the labeling cost is low, we estimate an invariant predictor for the target classification task with training data in a single domain. Additionally, we propose two cross-validation methods for selecting hyperparameters of invariance regularization to solve the issue of hyperparameter selection, which has not been handled properly in existing IL methods. The effectiveness of the proposed framework, including the cross-validation, is demonstrated empirically, and the correctness of the hyperparameter selection is proved under some conditions.

Tongtong Fang, Nan Lu, Gang Niu et al.

Importance weighting (IW) is a golden solver for joint distribution shift, where the joint distributions differ between the training and test data. To solve this problem, IW estimates test-to-training density ratios as importance weights and reweights the training losses accordingly. Recent advances in dynamic IW (DIW) integrate weight estimation into model training, enabling scalable IW for deep models and achieving strong performance on large modern datasets. Despite its promise, DIW remains limited in two aspects. First, it incurs substantial computational overhead by solving a kernel mean matching (KMM)-induced optimization problem to convergence in every mini-batch. Second, it relies solely on KMM for weight estimation, whereas the IW literature contains diverse estimation methods based on different divergence measures. In this paper, we propose accelerated DIW (ADIW), a unified and efficient IW framework for deep learning under joint distribution shift. ADIW performs a few lightweight projected gradient descent updates that warm-start from previously updated weights, substantially improving efficiency. Moreover, ADIW generalizes DIW into a unified divergence-minimization framework that supports diverse weight-estimation methods in a plug-and-play manner, including those based on the Kullback-Leibler divergence, squared distance, and Wasserstein-1 distance. We establish convergence guarantees for ADIW under mild conditions, and empirical results demonstrate that ADIW achieves state-of-the-art IW performance while being substantially more efficient.

Ryo Yoshida, Yoshihiro Hayashi, Hidemine Furuya et al.

Developing large-scale foundational datasets is a critical milestone in advancing artificial intelligence (AI)-driven scientific innovation. However, unlike AI-mature fields such as natural language processing, materials science, particularly polymer research, has significantly lagged in developing extensive open datasets. This lag is primarily due to the high costs of polymer synthesis and property measurements, along with the vastness and complexity of the chemical space. This study presents PolyOmics, an omics-scale computational database generated through fully automated molecular dynamics simulation pipelines that provide diverse physical properties for over $10^5$ polymeric materials. The PolyOmics database is collaboratively developed by approximately 260 researchers from 48 institutions to bridge the gap between academia and industry. Machine learning models pretrained on PolyOmics can be efficiently fine-tuned for a wide range of real-world downstream tasks, even when only limited experimental data are available. Notably, the generalisation capability of these simulation-to-real transfer models improve significantly as the size of the PolyOmics database increases, exhibiting power-law scaling. The emergence of scaling laws supports the "more is better" principle, highlighting the significance of ultralarge-scale computational materials data for improving real-world prediction performance. This unprecedented omics-scale database reveals vast unexplored regions of polymer materials, providing a foundation for AI-driven polymer science.

Shunya Minami, Yoshihiro Hayashi, Stephen Wu et al.

To address the challenge of limited experimental materials data, extensive physical property databases are being developed based on high-throughput computational experiments, such as molecular dynamics simulations. Previous studies have shown that fine-tuning a predictor pretrained on a computational database to a real system can result in models with outstanding generalization capabilities compared to learning from scratch. This study demonstrates the scaling law of simulation-to-real (Sim2Real) transfer learning for several machine learning tasks in materials science. Case studies of three prediction tasks for polymers and inorganic materials reveal that the prediction error on real systems decreases according to a power-law as the size of the computational data increases. Observing the scaling behavior offers various insights for database development, such as determining the sample size necessary to achieve a desired performance, identifying equivalent sample sizes for physical and computational experiments, and guiding the design of data production protocols for downstream real-world tasks.

Wei Huang, Andi Han, Mingyuan Bai et al.

Diffusion models generate high-dimensional data with remarkable quality, yet how their training efficiently learns the score function, bypassing the curse of dimensionality when data is supported on low-dimensional manifolds, remains theoretically unexplained. We identify a collapse-and-refine mechanism driven by the geometry of the score function itself: at small noise scales, the diverging singularity of the score drives a rapid dimensional collapse of the induced denoising map onto the data manifold projection; at moderate noise scales, training refines the intrinsic density on the learned manifold. We instantiate this principle as Score-induced Latent Diffusion (SiLD), a two-stage framework in which both manifold learning and density estimation emerge from a single denoising score matching objective, replacing the heuristic KL regularization of VAE-based latent diffusion models. We prove that the resulting sample complexity depends on the intrinsic dimension rather than the ambient dimension. Experiments on Stacked MNIST, CelebA variants, and molecular generation benchmarks show that SiLD matches or outperforms VAE-based LDMs in generation quality and consistently improves reconstruction, validating our theoretical predictions.

Kotaro Ikeda, Masanori Koyama, Jinzhe Zhang et al.

In this paper, we propose a flow-based method for learning all-to-all transfer maps among conditional distributions that approximates pairwise optimal transport. The proposed method addresses the challenge of handling the case of continuous conditions, which often involve a large set of conditions with sparse empirical observations per condition. We introduce a novel cost function that enables simultaneous learning of optimal transports for all pairs of conditional distributions. Our method is supported by a theoretical guarantee that, in the limit, it converges to the pairwise optimal transports among infinite pairs of conditional distributions. The learned transport maps are subsequently used to couple data points in conditional flow matching. We demonstrate the effectiveness of this method on synthetic and benchmark datasets, as well as on chemical datasets in which continuous physical properties are defined as conditions. The code for this project can be found at https://github.com/kotatumuri-room/A2A-FM

Kenji Fukumizu, Wei Huang, Han Bao et al.

We reformulate Optimal Transport Conditional Flow Matching (OT-CFM), a class of dynamical generative models, showing that it admits an exact proximal formulation via an extended Brenier potential, without assuming that the target distribution has a density. In particular, the mapping to recover the target point is exactly given by a proximal operator, which yields an explicit proximal expression of the vector field. We also discuss the convergence of minibatch OT-CFM to the population formulation as the batch size increases. Finally, using second epi-derivatives of convex potentials, we prove that, for manifold-supported targets, OT-CFM is terminally normally hyperbolic: after time rescaling, the dynamics contracts exponentially in directions normal to the data manifold while remaining neutral along tangential directions.

Shunyu Zhao, Yanfeng Yang, Shuai Li et al.

Constraint-based causal discovery methods require a large number of conditional independence (CI) tests, which severely limits their practical applicability due to high computational complexity. Therefore, it is crucial to design an algorithm that accelerates each individual test. To this end, we propose the Flow Matching-based Conditional Independence Test (FMCIT). The proposed test leverages the high computational efficiency of flow matching and requires the model to be trained only once throughout the entire causal discovery procedure, substantially accelerating causal discovery. According to numerical experiments, FMCIT effectively controls type-I error and maintains high testing power under the alternative hypothesis, even in the presence of high-dimensional conditioning sets. In addition, we further integrate FMCIT into a two-stage guided PC skeleton learning framework, termed GPC-FMCIT, which combines fast screening with guided, budgeted refinement using FMCIT. This design yields explicit bounds on the number of CI queries while maintaining high statistical power. Experiments on synthetic and real-world causal discovery tasks demonstrate favorable accuracy-efficiency trade-offs over existing CI testing methods and PC variants.

Noboru Isobe, Masanori Koyama, Jinzhe Zhang et al.

The task of conditional generation is one of the most important applications of generative models, and numerous methods have been developed to date based on the celebrated flow-based models. However, many flow-based models in use today are not built to allow one to introduce an explicit inductive bias to how the conditional distribution to be generated changes with respect to conditions. This can result in unexpected behavior in the task of style transfer, for example. In this research, we introduce extended flow matching (EFM), a direct extension of flow matching that learns a "matrix field" corresponding to the continuous map from the space of conditions to the space of distributions. We show that we can introduce inductive bias to the conditional generation through the matrix field and demonstrate this fact with MMOT-EFM, a version of EFM that aims to minimize the Dirichlet energy or the sensitivity of the distribution with respect to conditions. We will present our theory along with experimental results that support the competitiveness of EFM in conditional generation.

Tam Le, Truyen Nguyen, Kenji Fukumizu

We study the optimal transport (OT) problem for measures supported on a graph metric space. Recently, Le et al. (2022) leverage the graph structure and propose a variant of OT, namely Sobolev transport (ST), which yields a closed-form expression for a fast computation. However, ST is essentially coupled with the $L^p$ geometric structure within its definition which makes it nontrivial to utilize ST for other prior structures. In contrast, the classic OT has the flexibility to adapt to various geometric structures by modifying the underlying cost function. An important instance is the Orlicz-Wasserstein (OW) which moves beyond the $L^p$ structure by leveraging the \emph{Orlicz geometric structure}. Comparing to the usage of standard $p$-order Wasserstein, OW remarkably helps to advance certain machine learning approaches. Nevertheless, OW brings up a new challenge on its computation due to its two-level optimization formulation. In this work, we leverage a specific class of convex functions for Orlicz structure to propose the generalized Sobolev transport (GST). GST encompasses the ST as its special case, and can be utilized for prior structures beyond the $L^p$ geometry. In connection with the OW, we show that one only needs to simply solve a univariate optimization problem to compute the GST, unlike the complex two-level optimization problem in OW. We empirically illustrate that GST is several-order faster than the OW. Moreover, we provide preliminary evidences on the advantages of GST for document classification and for several tasks in topological data analysis.

Manuel Gloeckler, Shoji Toyota, Kenji Fukumizu et al.

Amortized simulation-based inference (SBI) methods train neural networks on simulated data to perform Bayesian inference. While this strategy avoids the need for tractable likelihoods, it often requires a large number of simulations and has been challenging to scale to time series data. Scientific simulators frequently emulate real-world dynamics through thousands of single-state transitions over time. We propose an SBI approach that can exploit such Markovian simulators by locally identifying parameters consistent with individual state transitions. We then compose these local results to obtain a posterior over parameters that align with the entire time series observation. We focus on applying this approach to neural posterior score estimation but also show how it can be applied, e.g., to neural likelihood (ratio) estimation. We demonstrate that our approach is more simulation-efficient than directly estimating the global posterior on several synthetic benchmark tasks and simulators used in ecology and epidemiology. Finally, we validate scalability and simulation efficiency of our approach by applying it to a high-dimensional Kolmogorov flow simulator with around one million data dimensions.

Eiki Shimizu, Kenji Fukumizu, Dino Sejdinovic

Kernel conditional mean embeddings (CMEs) offer a powerful framework for representing conditional distribution, but they often face scalability and expressiveness challenges. In this work, we propose a new method that effectively combines the strengths of deep learning with CMEs in order to address these challenges. Specifically, our approach leverages the end-to-end neural network (NN) optimization framework using a kernel-based objective. This design circumvents the computationally expensive Gram matrix inversion required by current CME methods. To further enhance performance, we provide efficient strategies to optimize the remaining kernel hyperparameters. In conditional density estimation tasks, our NN-CME hybrid achieves competitive performance and often surpasses existing deep learning-based methods. Lastly, we showcase its remarkable versatility by seamlessly integrating it into reinforcement learning (RL) contexts. Building on Q-learning, our approach naturally leads to a new variant of distributional RL methods, which demonstrates consistent effectiveness across different environments.

Tam Le, Truyen Nguyen, Hideitsu Hino et al.

We investigate the Sobolev IPM problem for probability measures supported on a graph metric space. Sobolev IPM is an important instance of integral probability metrics (IPM), and is obtained by constraining a critic function within a unit ball defined by the Sobolev norm. In particular, it has been used to compare probability measures and is crucial for several theoretical works in machine learning. However, to our knowledge, there are no efficient algorithmic approaches to compute Sobolev IPM effectively, which hinders its practical applications. In this work, we establish a relation between Sobolev norm and weighted $L^p$-norm, and leverage it to propose a \emph{novel regularization} for Sobolev IPM. By exploiting the graph structure, we demonstrate that the regularized Sobolev IPM provides a \emph{closed-form} expression for fast computation. This advancement addresses long-standing computational challenges, and paves the way to apply Sobolev IPM for practical applications, even in large-scale settings. Additionally, the regularized Sobolev IPM is negative definite. Utilizing this property, we design positive-definite kernels upon the regularized Sobolev IPM, and provide preliminary evidences of their advantages for comparing probability measures on a given graph for document classification and topological data analysis.

Alex Luedtke, Kenji Fukumizu

Generative models for counterfactual outcomes face two key sources of bias. Confounding bias arises when approaches fail to account for systematic differences between those who receive the intervention and those who do not. Misspecification bias arises when methods attempt to address confounding through estimation of an auxiliary model, but specify it incorrectly. We introduce DoubleGen, a doubly robust framework that modifies generative modeling training objectives to mitigate these biases. The new objectives rely on two auxiliaries -- a propensity and outcome model -- and successfully address confounding bias even if only one of them is correct. We provide finite-sample guarantees for this robustness property. We further establish conditions under which DoubleGen achieves oracle optimality -- matching the convergence rates standard approaches would enjoy if interventional data were available -- and minimax rate optimality. We illustrate DoubleGen with three examples: diffusion models, flow matching, and autoregressive language models.

Tam Le, Truyen Nguyen, Hideitsu Hino et al.

We study the Sobolev IPM problem for measures supported on a graph metric space, where critic function is constrained to lie within the unit ball defined by Sobolev norm. While Le et al. (2025) achieved scalable computation by relating Sobolev norm to weighted $L^p$-norm, the resulting framework remains intrinsically bound to $L^p$ geometric structure, limiting its ability to incorporate alternative structural priors beyond the $L^p$ geometry paradigm. To overcome this limitation, we propose to generalize Sobolev IPM through the lens of \emph{Orlicz geometric structure}, which employs convex functions to capture nuanced geometric relationships, building upon recent advances in optimal transport theory -- particularly Orlicz-Wasserstein (OW) and generalized Sobolev transport -- that have proven instrumental in advancing machine learning methodologies. This generalization encompasses classical Sobolev IPM as a special case while accommodating diverse geometric priors beyond traditional $L^p$ structure. It however brings up significant computational hurdles that compound those already inherent in Sobolev IPM. To address these challenges, we establish a novel theoretical connection between Orlicz-Sobolev norm and Musielak norm which facilitates a novel regularization for the generalized Sobolev IPM (GSI). By further exploiting the underlying graph structure, we show that GSI with Musielak regularization (GSI-M) reduces to a simple \emph{univariate optimization} problem, achieving remarkably computational efficiency. Empirically, GSI-M is several-order faster than the popular OW in computation, and demonstrates its practical advantages in comparing probability measures on a given graph for document classification and several tasks in topological data analysis.

Yanfeng Yang, Siwei Chen, Pingping Hu et al.

Probabilistic forecasting of multivariate time series is challenging due to non-stationarity, inter-variable dependencies, and distribution shifts. While recent diffusion and flow matching models have shown promise, they often ignore informative priors such as conditional means and covariances. In this work, we propose Conditionally Whitened Generative Models (CW-Gen), a framework that incorporates prior information through conditional whitening. Theoretically, we establish sufficient conditions under which replacing the traditional terminal distribution of diffusion models, namely the standard multivariate normal, with a multivariate normal distribution parameterized by estimators of the conditional mean and covariance improves sample quality. Guided by this analysis, we design a novel Joint Mean-Covariance Estimator (JMCE) that simultaneously learns the conditional mean and sliding-window covariance. Building on JMCE, we introduce Conditionally Whitened Diffusion Models (CW-Diff) and extend them to Conditionally Whitened Flow Matching (CW-Flow). Experiments on five real-world datasets with six state-of-the-art generative models demonstrate that CW-Gen consistently enhances predictive performance, capturing non-stationary dynamics and inter-variable correlations more effectively than prior-free approaches. Empirical results further demonstrate that CW-Gen can effectively mitigate the effects of distribution shift.

Yanfeng Yang, Kenji Fukumizu

Creating large-scale datasets for training high-performance generative models is often prohibitively expensive, especially when associated attributes or annotations must be provided. As a result, merging existing datasets has become a common strategy. However, the sets of attributes across datasets are often inconsistent, and their naive concatenation typically leads to block-wise missing conditions. This presents a significant challenge for conditional generative modeling when the multiple attributes are used jointly as conditions, thereby limiting the model's controllability and applicability. To address this issue, we propose a novel generative approach, Diffusion Model with Double Guidance, which enables precise conditional generation even when no training samples contain all conditions simultaneously. Our method maintains rigorous control over multiple conditions without requiring joint annotations. We demonstrate its effectiveness in molecular and image generation tasks, where it outperforms existing baselines both in alignment with target conditional distributions and in controllability under missing condition settings.

Tam Le, Truyen Nguyen, Hideitsu Hino et al.

We investigate optimal transport (OT) for measures on graph metric spaces with different total masses. To mitigate the limitations of traditional $L^p$ geometry, Orlicz-Wasserstein (OW) and generalized Sobolev transport (GST) employ Orlicz geometric structure, leveraging convex functions to capture nuanced geometric relationships and remarkably contribute to advance certain machine learning approaches. However, both OW and GST are restricted to measures with equal total mass, limiting their applicability to real-world scenarios where mass variation is common, and input measures may have noisy supports, or outliers. To address unbalanced measures, OW can either incorporate mass constraints or marginal discrepancy penalization, but this leads to a more complex two-level optimization problem. Additionally, GST provides a scalable yet rigid framework, which poses significant challenges to extend GST to accommodate nonnegative measures. To tackle these challenges, in this work we revisit the entropy partial transport (EPT) problem. By exploiting Caffarelli & McCann (2010)'s insights, we develop a novel variant of EPT endowed with Orlicz geometric structure, called Orlicz-EPT. We establish theoretical background to solve Orlicz-EPT using a binary search algorithmic approach. Especially, by leveraging the dual EPT and the underlying graph structure, we formulate a novel regularization approach that leads to the proposed Orlicz-Sobolev transport (OST). Notably, we demonstrate that OST can be efficiently computed by simply solving a univariate optimization problem, in stark contrast to the intensive computation needed for Orlicz-EPT. Building on this, we derive geometric structures for OST and draw its connections to other transport distances. We empirically illustrate that OST is several-order faster than Orlicz-EPT.

Yuto Tanimoto, Kenji Fukumizu

While many multi-armed bandit algorithms assume that rewards for all arms are constant across rounds, this assumption does not hold in many real-world scenarios. This paper considers the setting of recovering bandits (Pike-Burke & Grunewalder, 2019), where the reward depends on the number of rounds elapsed since the last time an arm was pulled. We propose a new reinforcement learning (RL) algorithm tailored to this setting, named the State-Separate SARSA (SS-SARSA) algorithm, which treats rounds as states. The SS-SARSA algorithm achieves efficient learning by reducing the number of state combinations required for Q-learning/SARSA, which often suffers from combinatorial issues for large-scale RL problems. Additionally, it makes minimal assumptions about the reward structure and offers lower computational complexity. Furthermore, we prove asymptotic convergence to an optimal policy under mild assumptions. Simulation studies demonstrate the superior performance of our algorithm across various settings.

Masanori Koyama, Kenji Fukumizu, Kohei Hayashi et al.

Symmetry learning has proven to be an effective approach for extracting the hidden structure of data, with the concept of equivariance relation playing the central role. However, most of the current studies are built on architectural theory and corresponding assumptions on the form of data. We propose Neural Fourier Transform (NFT), a general framework of learning the latent linear action of the group without assuming explicit knowledge of how the group acts on data. We present the theoretical foundations of NFT and show that the existence of a linear equivariant feature, which has been assumed ubiquitously in equivariance learning, is equivalent to the existence of a group invariant kernel on the dataspace. We also provide experimental results to demonstrate the application of NFT in typical scenarios with varying levels of knowledge about the acting group.

Hironori Murase, Kenji Fukumizu

In many anomaly detection tasks, where anomalous data rarely appear and are difficult to collect, training using only normal data is important. Although it is possible to manually create anomalous data using prior knowledge, they may be subject to user bias. In this paper, we propose an Anomalous Latent variable Generative Adversarial Network (ALGAN) in which the GAN generator produces pseudo-anomalous data as well as fake-normal data, whereas the discriminator is trained to distinguish between normal and pseudo-anomalous data. This differs from the standard GAN discriminator, which specializes in classifying two similar classes. The training dataset contains only normal data; the latent variables are introduced in anomalous states and are input into the generator to produce diverse pseudo-anomalous data. We compared the performance of ALGAN with other existing methods on the MVTec-AD, Magnetic Tile Defects, and COIL-100 datasets. The experimental results showed that ALGAN exhibited an AUROC comparable to those of state-of-the-art methods while achieving a much faster prediction time.

Pengzhou Wu, Kenji Fukumizu

As an important problem in causal inference, we discuss the identification and estimation of treatment effects (TEs) under limited overlap; that is, when subjects with certain features belong to a single treatment group. We use a latent variable to model a prognostic score which is widely used in biostatistics and sufficient for TEs; i.e., we build a generative prognostic model. We prove that the latent variable recovers a prognostic score, and the model identifies individualized treatment effects. The model is then learned as β-Intact-VAE--a new type of variational autoencoder (VAE). We derive the TE error bounds that enable representations balanced for treatment groups conditioned on individualized features. The proposed method is compared with recent methods using (semi-)synthetic datasets.

Pengzhou Wu, Kenji Fukumizu

As an important problem in causal inference, we discuss the estimation of treatment effects (TEs). Representing the confounder as a latent variable, we propose Intact-VAE, a new variant of variational autoencoder (VAE), motivated by the prognostic score that is sufficient for identifying TEs. Our VAE also naturally gives representations balanced for treatment groups, using its prior. Experiments on (semi-)synthetic datasets show state-of-the-art performance under diverse settings, including unobserved confounding. Based on the identifiability of our model, we prove identification of TEs under unconfoundedness, and also discuss (possible) extensions to harder settings.

Hiroaki Mikami, Kenji Fukumizu, Shogo Murai et al.

Synthetic-to-real transfer learning is a framework in which a synthetically generated dataset is used to pre-train a model to improve its performance on real vision tasks. The most significant advantage of using synthetic images is that the ground-truth labels are automatically available, enabling unlimited expansion of the data size without human cost. However, synthetic data may have a huge domain gap, in which case increasing the data size does not improve the performance. How can we know that? In this study, we derive a simple scaling law that predicts the performance from the amount of pre-training data. By estimating the parameters of the law, we can judge whether we should increase the data or change the setting of image synthesis. Further, we analyze the theory of transfer learning by considering learning dynamics and confirm that the derived generalization bound is consistent with our empirical findings. We empirically validated our scaling law on various experimental settings of benchmark tasks, model sizes, and complexities of synthetic images.

Pengzhou Wu, Kenji Fukumizu

NOTE: This preprint has a flawed theoretical formulation. Please avoid it and refer to the ICLR22 publication https://openreview.net/forum?id=q7n2RngwOM. Also, arXiv:2109.15062 contains some new ideas on unobserved Confounding. As an important problem of causal inference, we discuss the identification and estimation of treatment effects under unobserved confounding. Representing the confounder as a latent variable, we propose Intact-VAE, a new variant of variational autoencoder (VAE), motivated by the prognostic score that is sufficient for identifying treatment effects. We theoretically show that, under certain settings, treatment effects are identified by our model, and further, based on the identifiability of our model (i.e., determinacy of representation), our VAE is a consistent estimator with representation balanced for treatment groups. Experiments on (semi-)synthetic datasets show state-of-the-art performance under diverse settings.

Masaaki Imaizumi, Kenji Fukumizu

We develop a minimax rate analysis to describe the reason that deep neural networks (DNNs) perform better than other standard methods. For nonparametric regression problems, it is well known that many standard methods attain the minimax optimal rate of estimation errors for smooth functions, and thus, it is not straightforward to identify the theoretical advantages of DNNs. This study tries to fill this gap by considering the estimation for a class of non-smooth functions that have singularities on hypersurfaces. Our findings are as follows: (i) We derive the generalization error of a DNN estimator and prove that its convergence rate is almost optimal. (ii) We elucidate a phase diagram of estimation problems, which describes the situations where the DNNs outperform a general class of estimators, including kernel methods, Gaussian process methods, and others. We additionally show that DNNs outperform harmonic analysis based estimators. This advantage of DNNs comes from the fact that a shape of singularity can be successfully handled by their multi-layered structure.

Jean-Francois Ton, Dino Sejdinovic, Kenji Fukumizu

The inaccessibility of controlled randomized trials due to inherent constraints in many fields of science has been a fundamental issue in causal inference. In this paper, we focus on distinguishing the cause from effect in the bivariate setting under limited observational data. Based on recent developments in meta learning as well as in causal inference, we introduce a novel generative model that allows distinguishing cause and effect in the small data setting. Using a learnt task variable that contains distributional information of each dataset, we propose an end-to-end algorithm that makes use of similar training datasets at test time. We demonstrate our method on various synthetic as well as real-world data and show that it is able to maintain high accuracy in detecting directions across varying dataset sizes.

Shunya Minami, Song Liu, Stephen Wu et al.

We propose a novel framework that unifies and extends existing methods of transfer learning (TL) for regression. To bridge a pretrained source model to the model on a target task, we introduce a density-ratio reweighting function, which is estimated through the Bayesian framework with a specific prior distribution. By changing two intrinsic hyperparameters and the choice of the density-ratio model, the proposed method can integrate three popular methods of TL: TL based on cross-domain similarity regularization, a probabilistic TL using the density-ratio estimation, and fine-tuning of pretrained neural networks. Moreover, the proposed method can benefit from its simple implementation without any additional cost; the regression model can be fully trained using off-the-shelf libraries for supervised learning in which the original output variable is simply transformed to a new output variable. We demonstrate its simplicity, generality, and applicability using various real data applications.

Siddharth Vishwanath, Kenji Fukumizu, Satoshi Kuriki et al.

Persistent homology has become an important tool for extracting geometric and topological features from data, whose multi-scale features are summarized in a persistence diagram. From a statistical perspective, however, persistence diagrams are very sensitive to perturbations in the input space. In this work, we develop a framework for constructing robust persistence diagrams from superlevel filtrations of robust density estimators constructed using reproducing kernels. Using an analogue of the influence function on the space of persistence diagrams, we establish the proposed framework to be less sensitive to outliers. The robust persistence diagrams are shown to be consistent estimators in bottleneck distance, with the convergence rate controlled by the smoothness of the kernel. This, in turn, allows us to construct uniform confidence bands in the space of persistence diagrams. Finally, we demonstrate the superiority of the proposed approach on benchmark datasets.

Casey Chu, Kentaro Minami, Kenji Fukumizu

We formalize an equivalence between two popular methods for Bayesian inference: Stein variational gradient descent (SVGD) and black-box variational inference (BBVI). In particular, we show that BBVI corresponds precisely to SVGD when the kernel is the neural tangent kernel. Furthermore, we interpret SVGD and BBVI as kernel gradient flows; we do this by leveraging the recent perspective that views SVGD as a gradient flow in the space of probability distributions and showing that BBVI naturally motivates a Riemannian structure on that space. We observe that kernel gradient flow also describes dynamics found in the training of generative adversarial networks (GANs). This work thereby unifies several existing techniques in variational inference and generative modeling and identifies the kernel as a fundamental object governing the behavior of these algorithms, motivating deeper analysis of its properties.

Casey Chu, Kentaro Minami, Kenji Fukumizu

Generative adversarial networks, or GANs, commonly display unstable behavior during training. In this work, we develop a principled theoretical framework for understanding the stability of various types of GANs. In particular, we derive conditions that guarantee eventual stationarity of the generator when it is trained with gradient descent, conditions that must be satisfied by the divergence that is minimized by the GAN and the generator's architecture. We find that existing GAN variants satisfy some, but not all, of these conditions. Using tools from convex analysis, optimal transport, and reproducing kernels, we construct a GAN that fulfills these conditions simultaneously. In the process, we explain and clarify the need for various existing GAN stabilization techniques, including Lipschitz constraints, gradient penalties, and smooth activation functions.

Pengzhou Wu, Kenji Fukumizu

We address the problem of distinguishing cause from effect in bivariate setting. Based on recent developments in nonlinear independent component analysis (ICA), we train nonparametrically general nonlinear causal models that allow non-additive noise. Further, we build an ensemble framework, namely Causal Mosaic, which models a causal pair by a mixture of nonlinear models. We compare this method with other recent methods on artificial and real world benchmark datasets, and our method shows state-of-the-art performance.

Yuki Saito, Takuma Nakamura, Hirotaka Hachiya et al.

Matching two different sets of items, called heterogeneous set-to-set matching problem, has recently received attention as a promising problem. The difficulties are to extract features to match a correct pair of different sets and also preserve two types of exchangeability required for set-to-set matching: the pair of sets, as well as the items in each set, should be exchangeable. In this study, we propose a novel deep learning architecture to address the abovementioned difficulties and also an efficient training framework for set-to-set matching. We evaluate the methods through experiments based on two industrial applications: fashion set recommendation and group re-identification. In these experiments, we show that the proposed method provides significant improvements and results compared with the state-of-the-art methods, thereby validating our architecture for the heterogeneous set matching problem.

Heishiro Kanagawa, Wittawat Jitkrittum, Lester Mackey et al.

We propose a kernel-based nonparametric test of relative goodness of fit, where the goal is to compare two models, both of which may have unobserved latent variables, such that the marginal distribution of the observed variables is intractable. The proposed test generalizes the recently proposed kernel Stein discrepancy (KSD) tests (Liu et al., 2016, Chwialkowski et al., 2016, Yang et al., 2018) to the case of latent variable models, a much more general class than the fully observed models treated previously. The new test, with a properly calibrated threshold, has a well-controlled type-I error. In the case of certain models with low-dimensional latent structure and high-dimensional observations, our test significantly outperforms the relative Maximum Mean Discrepancy test, which is based on samples from the models and does not exploit the latent structure.

Kenji Fukumizu, Shoichiro Yamaguchi, Yoh-ichi Mototake et al.

We theoretically study the landscape of the training error for neural networks in overparameterized cases. We consider three basic methods for embedding a network into a wider one with more hidden units, and discuss whether a minimum point of the narrower network gives a minimum or saddle point of the wider one. Our results show that the networks with smooth and ReLU activation have different partially flat landscapes around the embedded point. We also relate these results to a difference of their generalization abilities in overparameterized realization.

Daniel Andrade, Kenji Fukumizu, Yuzuru Okajima

Clustering, like covariate selection for classification, is an important step to compress and interpret the data. However, clustering of covariates is often performed independently of the classification step, which can lead to undesirable clustering results that harm interpretability and compression rate. Therefore, we propose a method that can cluster covariates while taking into account class label information of samples. We formulate the problem as a convex optimization problem which uses both, a-priori similarity information between covariates, and information from class-labeled samples. Like ordinary convex clustering [Chi and Lange, 2015], the proposed method offers a unique global minima making it insensitive to initialization. In order to solve the convex problem, we propose a specialized alternating direction method of multipliers (ADMM), which scales up to several thousands of variables. Furthermore, in order to circumvent computationally expensive cross-validation, we propose a model selection criterion based on approximating the marginal likelihood. Experiments on synthetic and real data confirm the usefulness of the proposed clustering method and the selection criterion.

Tam Le, Makoto Yamada, Kenji Fukumizu et al.

Optimal transport (\OT) theory defines a powerful set of tools to compare probability distributions. \OT~suffers however from a few drawbacks, computational and statistical, which have encouraged the proposal of several regularized variants of OT in the recent literature, one of the most notable being the \textit{sliced} formulation, which exploits the closed-form formula between univariate distributions by projecting high-dimensional measures onto random lines. We consider in this work a more general family of ground metrics, namely \textit{tree metrics}, which also yield fast closed-form computations and negative definite, and of which the sliced-Wasserstein distance is a particular case (the tree is a chain). We propose the tree-sliced Wasserstein distance, computed by averaging the Wasserstein distance between these measures using random tree metrics, built adaptively in either low or high-dimensional spaces. Exploiting the negative definiteness of that distance, we also propose a positive definite kernel, and test it against other baselines on a few benchmark tasks.

Sho Yokoi, Sosuke Kobayashi, Kenji Fukumizu et al.

In this paper, we propose a new kernel-based co-occurrence measure that can be applied to sparse linguistic expressions (e.g., sentences) with a very short learning time, as an alternative to pointwise mutual information (PMI). As well as deriving PMI from mutual information, we derive this new measure from the Hilbert--Schmidt independence criterion (HSIC); thus, we call the new measure the pointwise HSIC (PHSIC). PHSIC can be interpreted as a smoothed variant of PMI that allows various similarity metrics (e.g., sentence embeddings) to be plugged in as kernels. Moreover, PHSIC can be estimated by simple and fast (linear in the size of the data) matrix calculations regardless of whether we use linear or nonlinear kernels. Empirically, in a dialogue response selection task, PHSIC is learned thousands of times faster than an RNN-based PMI while outperforming PMI in accuracy. In addition, we also demonstrate that PHSIC is beneficial as a criterion of a data selection task for machine translation owing to its ability to give high (low) scores to a consistent (inconsistent) pair with other pairs.

Daniel Andrade, Akiko Takeda, Kenji Fukumizu

Variable clustering is important for explanatory analysis. However, only few dedicated methods for variable clustering with the Gaussian graphical model have been proposed. Even more severe, small insignificant partial correlations due to noise can dramatically change the clustering result when evaluating for example with the Bayesian Information Criteria (BIC). In this work, we try to address this issue by proposing a Bayesian model that accounts for negligible small, but not necessarily zero, partial correlations. Based on our model, we propose to evaluate a variable clustering result using the marginal likelihood. To address the intractable calculation of the marginal likelihood, we propose two solutions: one based on a variational approximation, and another based on MCMC. Experiments on simulated data shows that the proposed method is similarly accurate as BIC in the no noise setting, but considerably more accurate when there are noisy partial correlations. Furthermore, on real data the proposed method provides clustering results that are intuitively sensible, which is not always the case when using BIC or its extensions.

Ho Chung Leon Law, Dino Sejdinovic, Ewan Cameron et al.

While a typical supervised learning framework assumes that the inputs and the outputs are measured at the same levels of granularity, many applications, including global mapping of disease, only have access to outputs at a much coarser level than that of the inputs. Aggregation of outputs makes generalization to new inputs much more difficult. We consider an approach to this problem based on variational learning with a model of output aggregation and Gaussian processes, where aggregation leads to intractability of the standard evidence lower bounds. We propose new bounds and tractable approximations, leading to improved prediction accuracy and scalability to large datasets, while explicitly taking uncertainty into account. We develop a framework which extends to several types of likelihoods, including the Poisson model for aggregated count data. We apply our framework to a challenging and important problem, the fine-scale spatial modelling of malaria incidence, with over 1 million observations.

Takafumi Kajihara, Motonobu Kanagawa, Keisuke Yamazaki et al.

We propose a novel approach to parameter estimation for simulator-based statistical models with intractable likelihood. Our proposed method involves recursive application of kernel ABC and kernel herding to the same observed data. We provide a theoretical explanation regarding why the approach works, showing (for the population setting) that, under a certain assumption, point estimates obtained with this method converge to the true parameter, as recursion proceeds. We have conducted a variety of numerical experiments, including parameter estimation for a real-world pedestrian flow simulator, and show that in most cases our method outperforms existing approaches.

Makoto Yamada, Denny Wu, Yao-Hung Hubert Tsai et al.

Measuring divergence between two distributions is essential in machine learning and statistics and has various applications including binary classification, change point detection, and two-sample test. Furthermore, in the era of big data, designing divergence measure that is interpretable and can handle high-dimensional and complex data becomes extremely important. In the paper, we propose a post selection inference (PSI) framework for divergence measure, which can select a set of statistically significant features that discriminate two distributions. Specifically, we employ an additive variant of maximum mean discrepancy (MMD) for features and introduce a general hypothesis test for PSI. A novel MMD estimator using the incomplete U-statistics, which has an asymptotically Normal distribution (under mild assumptions) and gives high detection power in PSI, is also proposed and analyzed theoretically. Through synthetic and real-world feature selection experiments, we show that the proposed framework can successfully detect statistically significant features. Last, we propose a sample selection framework for analyzing different members in the Generative Adversarial Networks (GANs) family.