Bayesian Nonparametric Global-local Clustering of High-dimensional Functions

arXiv:2412.09792 · stat.ME, math.ST, stat.TH · Submitted 2024-12-13 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Statistics. The day in statistics: methods, theory and the inference under them.

Tom: I'm Tom.

Jane: And I'm Jane, here with Lu, Meng and our in-house large language model, Lalam.

Tom: We're looking at the paper: "Bayesian Nonparametric Global-local Clustering of High-dimensional Functions".

Jane: Here's the gist of the paper. Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering proposes a flexible non-parametric Bayesian approach for multi-resolution clustering of high-dimensional functional data by imposing independent Dirichlet process priors on different subsets of basis coefficients to induce local clustering.

Tom: This method is significant because it addresses the inadequacy of global clustering methods in high-dimensional settings by accounting for heterogeneity in localized features across different resolutions.

Jane: Pruning redundant reasoning tokens The proposed method uses a wavelet basis expansion to represent the underlying random function, which provides a natural partition of the coefficient space based on varying resolution levels conducive to multi-scale local clustering.

Tom: The approach imposes independent Dirichlet process (DP) priors on coefficients corresponding to distinct wavelet resolutions, resulting in a product of DP mixture priors that induce local clustering.

Paper summary: Tom: So wrapping up this discussion on "Bayesian Nonparametric Global-local Clustering of High-dimensional Functions", the authors essentially introduced a flexible non-parametric Bayesian framework that uses wavelet expansions to create resolution-specific Dirichlet process priors to handle multi-scale functional clustering.

Jane: They claim that this approach is significant because it directly addresses the inadequacy of global clustering methods in high dimensions by allowing the model to delineate between clustering patterns driven by high-level features and those driven by finer resolution local features.

Lu: The paper establishes posterior consistency for density estimation under mild conditions, generalizing Bayesian non-parametric multivariate density estimation to functional data analysis, while they also developed an efficient Markov chain Monte Carlo algorithm based on wavelet representations for implementation.

Meng: From a deployment standpoint, the efficiency gains mentioned in Lemma one are compelling because they show a linear increase in complexity instead of quadratic growth with image size when using low-rank decomposition for covariance.

Lalam: The paper's application to spatial transcriptomics data shows that this method can identify interpretable and reproducible gene clusters by leveraging local features, which aligns well with what we need in biological discovery.

Jane: It seems like the main implication is that for complex functional data, you don't have to choose between capturing the big global structure or getting those fine-grained local details; you can get both.

Conclusion: Tom: So we've been looking at this paper, "Bayesian Nonparametric Global-local Clustering of High-dimensional Functions." Basically, they’re taking high-dimensional functional data—think complex curves or time series—and using a specific Bayesian math technique to find clusters in it.

Jane: It sounds really technical. What's the core problem they're trying to fix with this method? Is it just that global clustering methods aren't good enough?

Tom: Exactly. They say existing global clustering methods just force all points inside a cluster to be identical in some way, which doesn't capture the messy, localized differences within those groups.

Lu: The authors propose using wavelet expansions to break down the function into different resolution levels. Then they put separate Dirichlet process priors on coefficients for each of those levels. That’s what creates the local clustering effect we need for high-dimensional data.

Meng: From my side, I'm curious about deployment. If you have a massive dataset, how does this wavelet approach actually scale? Does it become computationally impossible fast?

Jane: That’s a fair question, Meng. The paper mentions that the computational complexity is much better than what you'd expect when using standard methods for covariance estimation. They claim it grows linearly with the size of the input instead of quadratically.

Tom: Linear versus quadratic growth is a big deal in production environments, Jane. It means you can actually use this on larger functional datasets without immediately hitting a wall on computation time.

Lu: Plus, they establish posterior consistency for density estimation under mild conditions, which means if the model is set up right, it actually recovers the true shape of the data distribution asymptotically.

Jane: So, it’s not just finding groups; it’s giving you a robust way to estimate the underlying structure of complex functions even when you don't know exactly what those functions look like beforehand.

Tom: It sounds like they've built a system that balances complexity control with capturing fine detail, which is something we need when dealing with real-world biological or financial data.

Meng: I'm still worried about the practical implementation details of setting up those wavelet bases correctly for production pipelines, though the theoretical scaling sounds promising.

Lu: That’s where the implementation work lies, Meng. But fundamentally, this framework allows for a much richer characterization of local perturbations that aren't actually related to the clustering itself.

Jane: So what does this mean for someone just listening? It means we can move past just seeing big trends and start understanding how those trends are built from the pieces at every level of detail.

Tom: Right. It’s about getting both the big picture and the fine grain, all in one flexible Bayesian structure.

Jane: And that leads us perfectly into how they actually put this to work, moving from theory to application...

Tsung-Hung Yao, Suprateek Kundu

University of Texas MD Anderson Cancer Center

stat.ME, math.ST, stat.TH

Submitted: 2024-12-13

Updated: 2026-11-03

License: http://creativecommons.org/licenses/by-nc-sa/4.0/

Importance score: 68/100

The gist: Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering proposes a flexible non-parametric Bayesian approach for multi-resolution clustering of high-dimensional…

Key concepts

Wavelet Basis Expansion
This technique represents the underlying random function using a set of basis functions derived from wavelets. Wavelets are crucial because they naturally create a partition of the coefficient space based on different resolution levels. This structure is what enables the method to distinguish between high-level (global) and fine-scale (local) features within the data.
Dirichlet Process Priors
These priors are used independently on subsets of coefficients corresponding to distinct wavelet resolutions. By using a product of these Dirichlet process mixtures, the model induces local clustering. This means that different scales or resolutions in the function's representation are clustered separately, capturing localized patterns effectively.
Product of DP Mixture Priors
This is the core prior structure used to define the model's distribution over functional data. It combines independent Dirichlet process priors for each resolution level. This product structure introduces resolution-specific clustering, allowing the model to flexibly delineate where global features cluster versus where finer, local features reside.

Terminology

Summary

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering proposes a flexible non-parametric Bayesian approach for multi-resolution clustering of high-dimensional functional data by imposing independent Dirichlet process priors on different subsets of basis coefficients to induce local clustering. p. 0

Where the paper says it

Focusing on basis expansions for high-dimensional functions, we propose a flexible non-parametric Bayesian approach for multi-resolution clustering. The proposed method imposes independent Dirichlet process (DP) priors on different subsets of basis coefficients that ultimately results in a product of DP mixture priors inducing local clustering.

This method is significant because it addresses the inadequacy of global clustering methods in high-dimensional settings by accounting for heterogeneity in localized features across different resolutions. p. 1
Where the paper says it

We conjecture that a potential reason for inadequate performance of global clustering approaches in high-dimensional settings may stem from their inability to account for: (i) heterogeneity of localized features (Di Iorio & Vantini, 2023); and (ii) scenarios where a small subset of important features separate the clusters (Floriello & Vitelli, 2017). To circumvent difficulties associated with global clustering, a limited number of local clustering methods have been proposed that enable flexible clustering in different resolutions or domains (e.

Pruning redundant reasoning tokens

The proposed method uses a wavelet basis expansion to represent the underlying random function, which provides a natural partition of the coefficient space based on varying resolution levels conducive to multi-scale local clustering. The approach imposes independent Dirichlet process (DP) priors on coefficients corresponding to distinct wavelet resolutions, resulting in a product of DP mixture priors that induce local clustering. p. 2

Where the paper says it

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering A Preprint characterization. To encourage local clustering among the high-dimensional basis coefficients, we apply independent Dirichlet process (DP) priors to coefficients corresponding to distinct wavelet resolutions.

This structure introduces resolution-specific clustering and increases flexibility by delineating the clustering of high-level (or global) features and finer resolution (or local) features. p. 2
Where the paper says it

To encourage local clustering among the high-dimensional basis coefficients, we apply independent Dirichlet process (DP) priors to coefficients corresponding to distinct wavelet resolutions. This structure introduces resolution-specific clustering and increases flexibility by delineating the clustering of high-level (or global) features and finer resolution (or local) features.

How the training signal is built

The model is specified using a product of Dirichlet process priors on the parameter space, denoted as Π∗(βi, Σi) = ÎJ j=0 Pj(βi j) × PΣ (Σi), which ultimately induces a distribution Π on the class of densities on the observed noisy function denoted as fPDPM. This induced density takes the form of a stick-breaking representation, generalizing generic kernel mixture representations to functional data analysis relying on suitable basis expansions. p. 4

Where the paper says it

, 𝐽, and 𝑤ℎ𝑠 is the weight of Σ in the ℎ𝑠th cluster with 𝑤ℎ𝑠 = 𝜈ℎ𝑠 Î 𝑒<ℎ𝑠 (1 − 𝜈𝑒), 𝜈𝑒 ∼ Beta(1, 𝛼Σ) and Σℎ𝑠 ∼ 𝑃 ∗ Σ . Equation (6) generalizes the generic kernel mixture representation in the nonparametric multivariate density estimation literature (Wu & Ghosal, 2008; Kundu & Lukemire, 2024) to the case of functional data analysis relying on suitable basis expansions.

The resulting class of priors is denoted as the product of Dirichlet process mixture on functional data (fPDPM). p. 4
Where the paper says it

This prior ultimately induces a distribution Π on the class of densities on the observed noisy function denoted as F = { 𝑓𝑃 (𝑦𝑖)} via the representation 𝑓𝑃 (𝑦) = ∫ 𝜙Σ 𝑦 − Í𝐽 𝑗=0 Ψ𝑗 𝛽𝑗 𝑑Π ∗ (𝛽, Σ). We denote the resulting class of priors Π on F as the product of Dirichlet process mixture on functional data (fPDPM).

Where the gains come from

The method achieves local clustering by independently partitioning coefficients across distinct resolution levels, which reduces dimensionality and improves computational complexity compared to traditional local clustering methods. This is formalized in Lemma 1, which compares the computational complexity of the proposed model with that of the classical LPP method, showing a linear increase in complexity instead of quadratic growth with image size when using low-rank decomposition for covariance. p. 4

Where the paper says it

Therefore, our specification uses 𝐽 +1 independent priors (instead of 2𝐽+1 in LPP) that results in massive improvements in computational complexity. This result is formalized in the Lemma below that compares the computational complexity of the classical LPP method with that of the proposed local clustering approach using a low rank decomposition to model the residual covariance.

Furthermore, the inclusion of correlated residual noise terms allows for an improved characterization of local perturbations that are unrelated to clustering and may accentuate heterogeneity. p. 2
Where the paper says it

Theoretical justifications based on posterior consistency: Under mild conditions, we establish posterior consistency for density estimation of the random function (Lemma 2 and Theorem 2) that is satisfied under practical prior specifications (Corollary 2). Inclusion of correlated noise: Our framework incorporates correlated residual noise terms (with independent errors as a special case) that allows an improved characterization of local perturbations that are unrelated to clustering and may accentuate heterogeneity.

Posterior consistency and efficiency

The authors establish posterior consistency for density estimation under mild conditions, generalizing rich theoretical results from Bayesian non-parametric multivariate density estimation to the case of Bayesian functional data analysis. They develop an efficient Markov chain Monte Carlo (MCMC) algorithm based on wavelet representations, which facilitates model implementation for high dimensional spatial functions. p. 2

Where the paper says it

We introduce low-rank representations of the error covariance that improves scalability to high-dimensions while retaining desirable theoretical properties. Efficient Gibbs sampler : We implement an efficient Markov chain Monte Carlo (MCMC) algorithm (Section 4) that facilitates model implementation for high dimensional spatial functions.

The Gibbs sampler utilizes a slice sampling technique to harness the number of possible clusters and reduce an infinite sum into a finite sum by introducing auxiliary parameters. p. 7
Where the paper says it

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering A Preprint 4 Posterior Computation and Inference We implement the fPDPM approach via an efficient Gibbs sampler that proceeds via a slice sampling technique. The full details are provided in Supplementary Material Section S4.

Simulation studies illustrate strong advantages of the proposed approach assuming correlated or independent errors compared to global clustering approaches in terms of clustering performance. p. 2
Where the paper says it

We develop an efficient MCMC algorithm for implementation of the proposed approach based on wavelet representations. Simulation studies illustrate strong advantages of the proposed approach assuming correlated or independent errors compared to global clustering approaches, in terms of clustering.

Application and results

The proposed fPDPM method is applied to a spatial transcriptomics data focused on clustering the spatially varying gene expression profiles in a breast cancer study where it infers interpretable and reproducible gene clusters. Simulation results show that fPDPM outperforms competing methods like the joint Dirichlet process model (DPM) and K-means clustering performed on functional principal components (PCA-KM) across various settings, highlighting the advantages of local clustering in high-dimensional spatial functions. p. 8

Where the paper says it

We initialized fPDPM with three different numbers factors of 𝐾 ∈ {1, 3, 10} for the low rank covariance and report results for these choices. We compared the performance with global clustering approaches such as the joint Dirichlet process model (DPM) and a K-means clustering performed on functional principal components (PCA-KM) as proposed in (Happ & Greven, 2018).

The method successfully identifies interpretable gene clusters that corroborate existing biological literature. p. 2
Where the paper says it

Simulation studies illustrate strong advantages of the proposed approach assuming correlated or independent errors compared to global clustering approaches, in terms of clustering. The proposed method is applied to a spatial transcriptomics data focused on clustering the spatially varying gene expression profiles in a breast cancer study where it infers interpretable and reproducible gene clusters.

Future directions

Current work considers cross-sectional functional data under the normality error assumption, with potential extensions to spatio-temporal data and clustering based on images from multiple modalities. The framework is broadly applicable to a wide class of functions that admit suitable basis representations. p. 2

Where the paper says it

The proposed method is applied to a spatial transcriptomics data focused on clustering the spatially varying gene expression profiles in a breast cancer study where it infers interpretable and reproducible gene clusters. While we focus on spatial functions for this article, the proposed approach is broadly applicable to a wide class of functions that admit suitable basis representations.

p. 2
Where the paper says it

The proposed method is applied to a spatial transcriptomics data focused on clustering the spatially varying gene expression profiles in a breast cancer study where it infers interpretable and reproducible gene clusters. While we focus on spatial functions for this article, the proposed approach is broadly applicable to a wide class of functions that admit suitable basis representations.

References

Aghabozorgi, S., Seyed Shirkhorshidi, A. & Ying Wah, T. (2015). Time-series clustering - A decade review. p. 1

Where the paper says it

We refer the readers to Wade (2023) for a more comprehensive review of such approaches. The majority of Bayesian global clustering methods have focused their implementation on one-dimensional functions such as time-series curves or longitudinal trajectories.

p. 1
Where the paper says it

We refer the readers to Wade (2023) for a more comprehensive review of such approaches. The majority of Bayesian global clustering methods have focused their implementation on one-dimensional functions such as time-series curves or longitudinal trajectories.

Andersson, A. p. 9

Where the paper says it

6 Spatial Transcriptomics of Breast Cancer 6. 1 Data Overview and Pre-processing We applied fPDPM to a publicly available spatial transcriptomics data of HER2-positive breast tumor from Andersson et al.

p. 9
Where the paper says it

6 Spatial Transcriptomics of Breast Cancer 6. 1 Data Overview and Pre-processing We applied fPDPM to a publicly available spatial transcriptomics data of HER2-positive breast tumor from Andersson et al.

, Larsson, L., Stenbeck, L., Salmen, F. ´, Ehinger, A., Wu, S. Z., Al-Eryani, G., Roden, D., Swarbrick, A., Borg, ˚A., Frisen, J. ´, Engblom, C. & Lundeberg, J. (2021). Spatial deconvolution of HER2-positive breast cancer delineates tumor-associated cell type interactions. p. 9
Where the paper says it

6 Spatial Transcriptomics of Breast Cancer 6. 1 Data Overview and Pre-processing We applied fPDPM to a publicly available spatial transcriptomics data of HER2-positive breast tumor from Andersson et al.

p. 9
Where the paper says it

6 Spatial Transcriptomics of Breast Cancer 6. 1 Data Overview and Pre-processing We applied fPDPM to a publicly available spatial transcriptomics data of HER2-positive breast tumor from Andersson et al.

Bhattacharya, A. & Dunson, D. p. 1

Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

p. 1
Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

p. 1
Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

p. 1
Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

B. (2011). Sparse Bayesian infinite factor models. p. 0
Where the paper says it

Most Bayesian functional data clustering methods rely on parametric finite mixture models (Bouveyron et al. , 2019; Chamroukhi & Nguyen, 2019) or more flexible non-parametric infinite mixtures (Fr¨uhwirth-Schnatter & Malsiner-Walli, 2019).

p. 0
Where the paper says it

Most Bayesian functional data clustering methods rely on parametric finite mixture models (Bouveyron et al. , 2019; Chamroukhi & Nguyen, 2019) or more flexible non-parametric infinite mixtures (Fr¨uhwirth-Schnatter & Malsiner-Walli, 2019).

Canale, A. & De Blasi, P. p. 2

Where the paper says it

The practical advantages of the proposed approach are complimented by desirable posterior consistency properties for density estimation, under mild assumptions. These results are established via careful sieve constructions and deriving corresponding entropy bounds and builds upon recent developments in non-parametric Bayesian density estimation literature (Canale & De Blasi, 2017; Kundu & Lukemire, 2024) focused on multivariate data.

p. 2
Where the paper says it

The practical advantages of the proposed approach are complimented by desirable posterior consistency properties for density estimation, under mild assumptions. These results are established via careful sieve constructions and deriving corresponding entropy bounds and builds upon recent developments in non-parametric Bayesian density estimation literature (Canale & De Blasi, 2017; Kundu & Lukemire, 2024) focused on multivariate data.

(2017). Posterior asymptotics of nonparametric location-scale mixtures for multivariate density estimation. p. 7
Where the paper says it

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering A Preprint 4 Posterior Computation and Inference We implement the fPDPM approach via an efficient Gibbs sampler that proceeds via a slice sampling technique. The full details are provided in Supplementary Material Section S4.

p. 7
Where the paper says it

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering A Preprint 4 Posterior Computation and Inference We implement the fPDPM approach via an efficient Gibbs sampler that proceeds via a slice sampling technique. The full details are provided in Supplementary Material Section S4.

Dunson, D. B. (2009). Nonparametric Bayes local partition models for random effects. p. 1

Where the paper says it

Beyond these limited findings, posterior consistency guarantees pertaining to density estimation of random functions under Bayesian non-parametric local clustering approaches are limited or essentially absent to our knowledge. Our focus is on developing practical non-parametric Bayesian product mixture models for local clustering of spatial functions that is scalable to high-dimensions and to investigate the theoretical properties such as posterior consistency.

p. 1
Where the paper says it

Beyond these limited findings, posterior consistency guarantees pertaining to density estimation of random functions under Bayesian non-parametric local clustering approaches are limited or essentially absent to our knowledge. Our focus is on developing practical non-parametric Bayesian product mixture models for local clustering of spatial functions that is scalable to high-dimensions and to investigate the theoretical properties such as posterior consistency.

Gelman, A. & Rubin, D. B. (1992). Inference from Iterative Simulation Using Multiple Sequences. p. 2

Where the paper says it

Section 3 shows the theoretical guarantees including posterior consistency, while Section 4 outlines posterior inference strategy. In Section 5, we present results from extensive simulation studies.

p. 2
Where the paper says it

Section 3 shows the theoretical guarantees including posterior consistency, while Section 4 outlines posterior inference strategy. In Section 5, we present results from extensive simulation studies.

Murphy, K., Viroli, C. & Gormley, I. C. (2020). Infinite mixtures of infinite factor analysers. p. 6

Where the paper says it

, 2020) and can vary among different clusters. The above prior construction is borrows from the infinite mixture of factor analyzers considered in Murphy et al.

p. 6
Where the paper says it

, 2020) and can vary among different clusters. The above prior construction is borrows from the infinite mixture of factor analyzers considered in Murphy et al.

Nguyen, X. L. & Gelfand, A. p. 1

Where the paper says it

Related alternative approaches were proposed by Rodr´ıguez et al. (2010), and Nguyen & Gelfand (2011).

p. 1
Where the paper says it

Related alternative approaches were proposed by Rodr´ıguez et al. (2010), and Nguyen & Gelfand (2011).

p. 1
Where the paper says it

Related alternative approaches were proposed by Rodr´ıguez et al. (2010), and Nguyen & Gelfand (2011).

p. 1
Where the paper says it

Related alternative approaches were proposed by Rodr´ıguez et al. (2010), and Nguyen & Gelfand (2011).

p. 1
Where the paper says it

Related alternative approaches were proposed by Rodr´ıguez et al. (2010), and Nguyen & Gelfand (2011).

p. 1
Where the paper says it

Related alternative approaches were proposed by Rodr´ıguez et al. (2010), and Nguyen & Gelfand (2011).

E. (2011). The dirichlet labeling process for clustering functional data. p. 0
Where the paper says it

Our method makes an important contribution by expanding the limited literature on local clustering methods for high-dimensional functions with theoretical guarantees. Keywords Global and local clustering · Product of Dirichlet process mixtures · Posterior consistency · Spatial function · Spatial transcriptomics 1 Introduction Functional data analysis is perhaps one of the most prominent areas of research in statistics and machine learning (Ramsay & Silverman, 2005).

p. 0
Where the paper says it

Our method makes an important contribution by expanding the limited literature on local clustering methods for high-dimensional functions with theoretical guarantees. Keywords Global and local clustering · Product of Dirichlet process mixtures · Posterior consistency · Spatial function · Spatial transcriptomics 1 Introduction Functional data analysis is perhaps one of the most prominent areas of research in statistics and machine learning (Ramsay & Silverman, 2005).

Pan, L., Li, J., Xu, Q., Gao, Z., Yang, M., Wu, X. & Li, X. (2024). HER2/PI3K/AKT pathway in HER2-positive breast cancer: A review. p. 9

Where the paper says it

The second cluster belongs to the pathway of signal transduction, and 42 genes out of 142 genes are part of the pathway. For example, Akt1 is a member of AKT pathway, which is a common mutated pathway related to the drug resistance in HER2 positive breast cancer (Pan et al.

Park, T. & Casella, G. p. 6

Where the paper says it

The proof is a direct result of Corollary 1 of Canale & De Blasi (2017). In this paper, we generalize these priors to Laplace priors (Park & Casella, 2008) on the base measures 𝑃 ∗ (𝛽) that achieve better shrinkage and therefore results in improved model fitting.

p. 6
Where the paper says it

The proof is a direct result of Corollary 1 of Canale & De Blasi (2017). In this paper, we generalize these priors to Laplace priors (Park & Casella, 2008) on the base measures 𝑃 ∗ (𝛽) that achieve better shrinkage and therefore results in improved model fitting.

(2008). The Bayesian Lasso. p. 0
Where the paper says it

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering A Preprint Tsung-Hung Yao Department of Biostatistics The University of Texas MD Anderson Cancer Center Houston, TX 77030 tyao@mdanderson. org Suprateek Kundu∗ Department of Biostatistics The University of Texas MD Anderson Cancer Center Houston, TX 77030 skundu2@mdanderson.

Petrone, S. p. 1

Where the paper says it

To circumvent difficulties associated with global clustering, a limited number of local clustering methods have been proposed that enable flexible clustering in different resolutions or domains (e. Dunson, 2009; Petrone et al.

p. 1
Where the paper says it

To circumvent difficulties associated with global clustering, a limited number of local clustering methods have been proposed that enable flexible clustering in different resolutions or domains (e. Dunson, 2009; Petrone et al.

, Guindani, M. & Gelfand, A. E. (2009). Hybrid dirichlet mixture models for functional data. p. 0
Where the paper says it

, 2015), Bayesian literature on functional data clustering is (somewhat surprisingly) not very well-developed (Zhang & Parnell, 2023). Most Bayesian functional data clustering methods rely on parametric finite mixture models (Bouveyron et al.

Ramsay, J. & Silverman, B. p. 0

Where the paper says it

Our method makes an important contribution by expanding the limited literature on local clustering methods for high-dimensional functions with theoretical guarantees. Keywords Global and local clustering · Product of Dirichlet process mixtures · Posterior consistency · Spatial function · Spatial transcriptomics 1 Introduction Functional data analysis is perhaps one of the most prominent areas of research in statistics and machine learning (Ramsay & Silverman, 2005).

p. 0
Where the paper says it

Our method makes an important contribution by expanding the limited literature on local clustering methods for high-dimensional functions with theoretical guarantees. Keywords Global and local clustering · Product of Dirichlet process mixtures · Posterior consistency · Spatial function · Spatial transcriptomics 1 Introduction Functional data analysis is perhaps one of the most prominent areas of research in statistics and machine learning (Ramsay & Silverman, 2005).

(2005). Functional Data Analysis. p. 0
Where the paper says it

Our method makes an important contribution by expanding the limited literature on local clustering methods for high-dimensional functions with theoretical guarantees. Keywords Global and local clustering · Product of Dirichlet process mixtures · Posterior consistency · Spatial function · Spatial transcriptomics 1 Introduction Functional data analysis is perhaps one of the most prominent areas of research in statistics and machine learning (Ramsay & Silverman, 2005).

Ray, S. & Mallick, B. p. 1

Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

p. 1
Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

(2006). Functional clustering by bayesian wavelet methods. p. 1
Where the paper says it

Moreover, the theoretical implications of Bayesian local clustering methods are not particularly wellunderstood, to our knowledge. Using wavelet expansions, Suarez & Ghosal (2016) established asymptotic consistency in recovering the true (finite dimensional) model parameters as well as the true partition structure in the ideal case when the observed measurement error associated with the function approaches zero.

Reiss, P. T., Huo, L., Zhao, Y., Kelly, C. & Ogden, R. T. (2015). Wavelet-domain regression and predictive inference in psychiatric neuroimaging. p. 2

Where the paper says it

We develop the proposed methodology for noisy functional data in Section 2. Section 3 shows the theoretical guarantees including posterior consistency, while Section 4 outlines posterior inference strategy.

Rodriguez, A. p. 1

Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

p. 1
Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

& Dunson, D. B. (2014). Functional clustering in nested designs: Modeling variability in reproductive epidemiology studies. p. 0
Where the paper says it

org Suprateek Kundu∗ Department of Biostatistics The University of Texas MD Anderson Cancer Center Houston, TX 77030 skundu2@mdanderson. org December 16, 2024 Abstract There is a rich literature on clustering functional data with applications to time-series modeling, trajectory data, and even spatio-temporal applications.

Rodr´ıguez, A. p. 1

Where the paper says it

Dunson, 2009; Petrone et al. , 2009; Rodr´ıguez et al.

p. 1
Where the paper says it

Dunson, 2009; Petrone et al. , 2009; Rodr´ıguez et al.

, Dunson, D. p. 1
Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

p. 1
Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

B. & Gelfand, A. E. (2010). Latent stick-breaking processes. p. 4
Where the paper says it

This method generalizes previous work on product mixture priors that was proposed for multivariate time-series data (Kundu & Lukemire, 2024) to the case of spatial functions. By using the stick-breaking representation (Sethuraman, 1994), the induced density takes the form: 𝑓𝑃 (𝑦) = ∫ .

Rousseeuw, P. p. 8

Where the paper says it

For the global clustering approaches, the ARI from the different resolutions are obtained from comparing the global membership to the true local membership at different resolutions. For PCA-KM, we select smallest number of principal components that explain at least 95% variance and decide the number of clusters by maximizing the average silhouette width (Rousseeuw, 1987).

p. 8
Where the paper says it

For the global clustering approaches, the ARI from the different resolutions are obtained from comparing the global membership to the true local membership at different resolutions. For PCA-KM, we select smallest number of principal components that explain at least 95% variance and decide the number of clusters by maximizing the average silhouette width (Rousseeuw, 1987).

J. (1987). Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. p. 9
Where the paper says it

We used expression data of patient that consists of three consecutive cryosection samples (H1, H2, and H3, henceforth), and compared the clustering reliability of the genes across the tissues, along with the biological interpretation of the clusters. We present the results for cluster memberships for the H1 tissue sample in the main manuscript, with the results for the remaining two samples in Supplementary Material Section S6.

p. 9
Where the paper says it

We used expression data of patient that consists of three consecutive cryosection samples (H1, H2, and H3, henceforth), and compared the clustering reliability of the genes across the tissues, along with the biological interpretation of the clusters. We present the results for cluster memberships for the H1 tissue sample in the main manuscript, with the results for the remaining two samples in Supplementary Material Section S6.

Schwartz, L. p. 5

Where the paper says it

Under an orthonormal basis (resulting in a finite determinant and Frobenius norm of Ψ𝑗), the proof mainly follows from the results in Wu & Ghosal (2008). Following the argument of Schwartz (1965), a positive prior support for arbitrarily small KL neighborhoods of 𝑓0 from Lemma 2 provides the weak consistency of fPDPM.

p. 5
Where the paper says it

Under an orthonormal basis (resulting in a finite determinant and Frobenius norm of Ψ𝑗), the proof mainly follows from the results in Wu & Ghosal (2008). Following the argument of Schwartz (1965), a positive prior support for arbitrarily small KL neighborhoods of 𝑓0 from Lemma 2 provides the weak consistency of fPDPM.

(1965). On Bayes procedures.

Sethuraman, J. p. 4

Where the paper says it

This method generalizes previous work on product mixture priors that was proposed for multivariate time-series data (Kundu & Lukemire, 2024) to the case of spatial functions. By using the stick-breaking representation (Sethuraman, 1994), the induced density takes the form: 𝑓𝑃 (𝑦) = ∫ .

p. 4
Where the paper says it

This method generalizes previous work on product mixture priors that was proposed for multivariate time-series data (Kundu & Lukemire, 2024) to the case of spatial functions. By using the stick-breaking representation (Sethuraman, 1994), the induced density takes the form: 𝑓𝑃 (𝑦) = ∫ .

(1994). A constructive definition of Dirichlet priors. p. 0
Where the paper says it

Focusing on basis expansions for high-dimensional functions, we propose a flexible non-parametric Bayesian approach for multi-resolution clustering. The proposed method imposes independent Dirichlet process (DP) priors on different subsets of basis coefficients that ultimately results in a product of DP mixture priors inducing local clustering.

p. 0
Where the paper says it

Focusing on basis expansions for high-dimensional functions, we propose a flexible non-parametric Bayesian approach for multi-resolution clustering. The proposed method imposes independent Dirichlet process (DP) priors on different subsets of basis coefficients that ultimately results in a product of DP mixture priors inducing local clustering.

Shang, L. & Zhou, X.(2022). Spatially aware dimension reduction for spatial transcriptomics. p. 2

Where the paper says it

This strategy is particularly useful when local features introduce additional noise and heterogeneity, or when the clusters are separated by a small subset of global features only. The proposed approach facilitates dimension reduction by dividing the basis coefficients into smaller subsets, which enhances model parsimony, reduces computational complexity, and improves scalability.

p. 2
Where the paper says it

This strategy is particularly useful when local features introduce additional noise and heterogeneity, or when the clusters are separated by a small subset of global features only. The proposed approach facilitates dimension reduction by dividing the basis coefficients into smaller subsets, which enhances model parsimony, reduces computational complexity, and improves scalability.

Song, P. N., Mansur, A., Dugger, K. J., Davis, T. R., Howard, G., Yankeelov, T. E. & Sorace, A. G.(2020). CD4 t-cell immune stimulation of HER2 + breast cancer cells alters response to trastuzumab in vitro. p. 9

Where the paper says it

The third cluster is characterized by immune system with 50 out of 124 genes belong to the immune system pathway. For example, CD4 is a receptor on the surface of many immune cells such as T cells, and the stimulated CD4 improves the treatment efficacy of HER2 positive breast cancer (Song et al.

p. 9
Where the paper says it

The third cluster is characterized by immune system with 50 out of 124 genes belong to the immune system pathway. For example, CD4 is a receptor on the surface of many immune cells such as T cells, and the stimulated CD4 improves the treatment efficacy of HER2 positive breast cancer (Song et al.

Suarez, A. J. & Ghosal, S. p. 1

Where the paper says it

Dunson (2009) proposed local partition processes (LPP) that enabled independent clustering across basis coefficients, while also specifying an additional shared mixture component for global clustering. Suarez & Ghosal (2016) proposed a spike and slab prior independently for wavelet basis coefficients, and proposed a post-MCMC decision-theoretic criteria to consolidate the local clusters to inform global clustering.

p. 1
Where the paper says it

Dunson (2009) proposed local partition processes (LPP) that enabled independent clustering across basis coefficients, while also specifying an additional shared mixture component for global clustering. Suarez & Ghosal (2016) proposed a spike and slab prior independently for wavelet basis coefficients, and proposed a post-MCMC decision-theoretic criteria to consolidate the local clusters to inform global clustering.

p. 1
Where the paper says it

Dunson (2009) proposed local partition processes (LPP) that enabled independent clustering across basis coefficients, while also specifying an additional shared mixture component for global clustering. Suarez & Ghosal (2016) proposed a spike and slab prior independently for wavelet basis coefficients, and proposed a post-MCMC decision-theoretic criteria to consolidate the local clusters to inform global clustering.

p. 1
Where the paper says it

Dunson (2009) proposed local partition processes (LPP) that enabled independent clustering across basis coefficients, while also specifying an additional shared mixture component for global clustering. Suarez & Ghosal (2016) proposed a spike and slab prior independently for wavelet basis coefficients, and proposed a post-MCMC decision-theoretic criteria to consolidate the local clusters to inform global clustering.

(2016). Bayesian clustering of functional data using local features. p. 2
Where the paper says it

In summary, our approach based on non-parametric Bayesian product mixture models for high-dimensional functional data introduces four main contributions: a. Local clustering to address the curse of dimensionality resulting in improved computation complexity: Using independent DPs on basis coefficients corresponding to distinct wavelet resolutions, the proposed approach partitions the parameter space and results in local clustering across resolutions and global clustering within a given resolution.

p. 2
Where the paper says it

In summary, our approach based on non-parametric Bayesian product mixture models for high-dimensional functional data introduces four main contributions: a. Local clustering to address the curse of dimensionality resulting in improved computation complexity: Using independent DPs on basis coefficients corresponding to distinct wavelet resolutions, the proposed approach partitions the parameter space and results in local clustering across resolutions and global clustering within a given resolution.

Wade, S. (2023). Bayesian cluster analysis Subject Areas:. p. 0

Where the paper says it

Our method makes an important contribution by expanding the limited literature on local clustering methods for high-dimensional functions with theoretical guarantees. Keywords Global and local clustering · Product of Dirichlet process mixtures · Posterior consistency · Spatial function · Spatial transcriptomics 1 Introduction Functional data analysis is perhaps one of the most prominent areas of research in statistics and machine learning (Ramsay & Silverman, 2005).

p. 0
Where the paper says it

Our method makes an important contribution by expanding the limited literature on local clustering methods for high-dimensional functions with theoretical guarantees. Keywords Global and local clustering · Product of Dirichlet process mixtures · Posterior consistency · Spatial function · Spatial transcriptomics 1 Introduction Functional data analysis is perhaps one of the most prominent areas of research in statistics and machine learning (Ramsay & Silverman, 2005).

Wang, J.-L., Chiou, J.-M. & Muller, H.-G.(2015). Review of functional data analysis. p. 0

Where the paper says it

Our method makes an important contribution by expanding the limited literature on local clustering methods for high-dimensional functions with theoretical guarantees. Keywords Global and local clustering · Product of Dirichlet process mixtures · Posterior consistency · Spatial function · Spatial transcriptomics 1 Introduction Functional data analysis is perhaps one of the most prominent areas of research in statistics and machine learning (Ramsay & Silverman, 2005).

Wu, Y. & Ghosal, S.(2008). Kullback leibler property of kernel mixture priors in bayesian density estimation. p. 4

Where the paper says it

, 𝐽, and 𝑤ℎ𝑠 is the weight of Σ in the ℎ𝑠th cluster with 𝑤ℎ𝑠 = 𝜈ℎ𝑠 Î 𝑒<ℎ𝑠 (1 − 𝜈𝑒), 𝜈𝑒 ∼ Beta(1, 𝛼Σ) and Σℎ𝑠 ∼ 𝑃 ∗ Σ . Equation (6) generalizes the generic kernel mixture representation in the nonparametric multivariate density estimation literature (Wu & Ghosal, 2008; Kundu & Lukemire, 2024) to the case of functional data analysis relying on suitable basis expansions.

p. 4
Where the paper says it

, 𝐽, and 𝑤ℎ𝑠 is the weight of Σ in the ℎ𝑠th cluster with 𝑤ℎ𝑠 = 𝜈ℎ𝑠 Î 𝑒<ℎ𝑠 (1 − 𝜈𝑒), 𝜈𝑒 ∼ Beta(1, 𝛼Σ) and Σℎ𝑠 ∼ 𝑃 ∗ Σ . Equation (6) generalizes the generic kernel mixture representation in the nonparametric multivariate density estimation literature (Wu & Ghosal, 2008; Kundu & Lukemire, 2024) to the case of functional data analysis relying on suitable basis expansions.

Ye, Y., Xu, C., Chen, F., Liu, Q. & Cheng, N.(2412.09792pg13>.

Zhang, M. & Parnell, A. p. 0

Where the paper says it

In contrast to a rich literature for analyzing functional datasets (Wang et al. , 2015), Bayesian literature on functional data clustering is (somewhat surprisingly) not very well-developed (Zhang & Parnell, 2023).

p. 0
Where the paper says it

In contrast to a rich literature for analyzing functional datasets (Wang et al. , 2015), Bayesian literature on functional data clustering is (somewhat surprisingly) not very well-developed (Zhang & Parnell, 2023).

(2412.09792pg13>.

--- Page 1 ---

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering A Preprint

Tsung-Hung Yao Department of Biostatistics The University of Texas MD Anderson Cancer Center Houston, TX 77030 tyao@mdanderson.org Suprateek Kundu∗ Department of Biostatistics The University of Texas MD Anderson Cancer Center Houston, TX 77030 skundu2@mdanderson. p. 0

Where the paper says it

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering A Preprint Tsung-Hung Yao Department of Biostatistics The University of Texas MD Anderson Cancer Center Houston, TX 77030 tyao@mdanderson. org Suprateek Kundu∗ Department of Biostatistics The University of Texas MD Anderson Cancer Center Houston, TX 77030 skundu2@mdanderson.

org December 16, 2024. p. 0
Where the paper says it

org Suprateek Kundu∗ Department of Biostatistics The University of Texas MD Anderson Cancer Center Houston, TX 77030 skundu2@mdanderson. org December 16, 2024 Abstract There is a rich literature on clustering functional data with applications to time-series modeling, trajectory data, and even spatio-temporal applications.

Abstract There is a rich literature on clustering functional data with applications to time-series modeling, trajectory data, and even spatio-temporal applications. p. 0

Where the paper says it

org Suprateek Kundu∗ Department of Biostatistics The University of Texas MD Anderson Cancer Center Houston, TX 77030 skundu2@mdanderson. org December 16, 2024 Abstract There is a rich literature on clustering functional data with applications to time-series modeling, trajectory data, and even spatio-temporal applications.

However, existing methods routinely perform global clustering that enforces identical atom values within the same cluster. p. 0
Where the paper says it

org December 16, 2024 Abstract There is a rich literature on clustering functional data with applications to time-series modeling, trajectory data, and even spatio-temporal applications. However, existing methods routinely perform global clustering that enforces identical atom values within the same cluster.

--- Page 5 ---

Bayesian functional clustering approaches typically assign a global membership, creating subgroups in which functions share the same cluster membership within each subgroup. These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. p. 1

Where the paper says it

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering A Preprint Bayesian functional clustering approaches typically assign a global membership, creating subgroups in which functions share the same cluster membership within each subgroup. These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster.

p. 1
Where the paper says it

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering A Preprint Bayesian functional clustering approaches typically assign a global membership, creating subgroups in which functions share the same cluster membership within each subgroup. These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster.

p. 1
Where the paper says it

Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering A Preprint Bayesian functional clustering approaches typically assign a global membership, creating subgroups in which functions share the same cluster membership within each subgroup. These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster.

Bayesian functional clustering approaches typically assign a global membership, creating subgroups in which functions share the same cluster membership within each subgroup. These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster.

Bayesian functional clustering approaches typically assign a global membership, creating subgroups in which functions share the same cluster membership within each subgroup. These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster.

To circumvent difficulties associated with global clustering, a limited number of local clustering methods have been proposed that enable flexible clustering in different resolutions or domains (e.g. Dunson, 2009; Petrone et al. p. 1

Where the paper says it

To circumvent difficulties associated with global clustering, a limited number of local clustering methods have been proposed that enable flexible clustering in different resolutions or domains (e. Dunson, 2009; Petrone et al.

, 2009; Rodriguez et al. p. 1
Where the paper says it

These global clustering methods produce distinct groups of functions defined by overarching patterns; however, they may not capture the heterogeneity in localized features within the functions lying in the same cluster. Some examples of global clustering methods include (Ray & Mallick, 2006; Crandell & Dunson, 2011; Bigelow & Dunson, 2009; Rodriguez & Dunson, 2014).

, 2010). These approaches produce distinct clustering allocations across various subsets of model parameters characterizing the function and thereby relax the restrictions inherent in global clustering. p. 1
Where the paper says it

, 2009; Rodr´ıguez et al. These approaches produce distinct clustering allocations across various subsets of model parameters characterizing the function and thereby relax the restrictions inherent in global clustering.

The authors show posterior consistency properties under the local clustering approach that asymptotically recovers the true density of random functions. The method incorporates correlated residual noise terms, which allows for an improved characterization of local perturbations that are unrelated to clustering and may accentuate heterogeneity. p. 2

Where the paper says it

Theoretical justifications based on posterior consistency: Under mild conditions, we establish posterior consistency for density estimation of the random function (Lemma 2 and Theorem 2) that is satisfied under practical prior specifications (Corollary 2). Inclusion of correlated noise: Our framework incorporates correlated residual noise terms (with independent errors as a special case) that allows an improved characterization of local perturbations that are unrelated to clustering and may accentuate heterogeneity.

They implement an efficient Markov chain Monte Carlo (MCMC) algorithm based on wavelet representations. p. 2
Where the paper says it

We introduce low-rank representations of the error covariance that improves scalability to high-dimensions while retaining desirable theoretical properties. Efficient Gibbs sampler : We implement an efficient Markov chain Monte Carlo (MCMC) algorithm (Section 4) that facilitates model implementation for high dimensional spatial functions.

Simulation studies demonstrate that fPDPM outperforms existing methods that focus on global clustering under a variety of settings, with the proposed method showing better accuracy for local clustering scenarios. The approach is applied to a spatial transcriptomics data where it infers interpretable and reproducible gene clusters. p. 2

Where the paper says it

Simulation studies illustrate strong advantages of the proposed approach assuming correlated or independent errors compared to global clustering approaches, in terms of clustering. The proposed method is applied to a spatial transcriptomics data focused on clustering the spatially varying gene expression profiles in a breast cancer study where it infers interpretable and reproducible gene clusters.

The results identify replicable gene clusters comprising genes that corroborate existing biological literature. p. 9
Where the paper says it

We used expression data of patient that consists of three consecutive cryosection samples (H1, H2, and H3, henceforth), and compared the clustering reliability of the genes across the tissues, along with the biological interpretation of the clusters. We present the results for cluster memberships for the H1 tissue sample in the main manuscript, with the results for the remaining two samples in Supplementary Material Section S6.

The authors consider generalizations to spatio-temporal data and clustering based on images from multiple modalities as potential future extensions. The framework is broadly applicable to a wide class of functions that admit suitable basis representations.

Aghabozorgi, S., Seyed Shirkhorshidi, A. & Ying Wah, T. (2015). Time-series clustering - A decade review.

Andersson, A., Larsson, L., Stenbeck, L., Salmen, F. ´, Ehinger, A., Wu, S. Z., Al-Eryani, G., Roden, D., Swarbrick, A., Borg, ˚A., Frisen, J. ´, Engblom, C. & Lundeberg, J. (2021). Spatial deconvolution of HER2-positive breast cancer delineates tumor-associated cell type interactions.

Bhattacharya, A. & Dunson, D. B. (2011). Sparse Bayesian infinite factor models.

Canale, A. & De Blasi, P. (2017). Posterior asymptotics of nonparametric location-scale mixtures for multivariate density estimation.

Dunson, D. B. (2009). Nonparametric Bayes local partition models for random effects.

Gelman, A. & Rubin, D. B. (1992). Inference from Iterative Simulation Using Multiple Sequences.

Murphy, K., Viroli, C. & Gormley, I. C. (2020). Infinite mixtures of infinite factor analysers.

Nguyen, X. L. & Gelfand, A. E.(2011). The dirichlet labeling process for clustering functional data.

Pan, L., Li, J., Xu, Q., Gao, Z., Yang, M., Wu, X. & Li, X.(2412.09792pg13>.

Park, T. & Casella, G.(2412.09792pg38>.

Petrone, S., Guindani, M. & Gelfand, A. E.(2412.09792pg38>.

Ramsay, J. & Silverman, B.(2412.09792pg13>.

Ray, S. & Mallick, B.(2412.09792pg13>.

Reiss, P. T., Huo, L., Zhao, Y., Kelly, C. & Ogden, R. T.(2412.09792pg13>.

Rodriguez, A. & Dunson, D. B.(2412.09792pg38>.

Rodr´ıguez, A., Dunson, D. B.& Gelfand, A. E.(2412.09792pg38>.

Rousseeuw, P. J.(1987). Silhouettes: A graphical aid to the interpretation and validation of cluster analysis.

Schwartz, L.(1965). On Bayes procedures.

Sethuraman, J.(1994). A constructive definition of Dirichlet priors.

Shang, L. & Zhou, X.(2022). Spatially aware dimension reduction for spatial transcriptomics.

Song, P. N., Mansur, A., Dugger, K. J., Davis, T. R., Howard, G., Yankeelov, T. E. & Sorace, A. G.(2020). CD4 t-cell immune stimulation of HER2 + breast cancer cells alters response to trastuzumab in vitro.

Suarez, A. J. & Ghosal, S.(2016). Bayesian clustering of functional data using local features.

Wade, S.(2023). Bayesian cluster analysis Subject Areas:.

Wang, J.-L., Chiou, J.-M. & Muller, H.-G.(2412.09792pg13>.

Wu, Y. & Ghosal, S.(2008). Kullback leibler property of kernel mixture priors in bayesian density estimation.

Ye, Y., Xu, C., Chen, F., Liu, Q. & Cheng, N.(2412.09792pg13>.

Zhang, M. & Parnell, A.(2412.09792pg13>.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Flexible Bayesian Nonparametric Product Mixtures for Multi-scale Functional Clustering (fPDPM) can be used to perform local clustering of high-dimensional functional data across multiple resolution levels by imposing independent Dirichlet process priors on basis coefficients corresponding to distinct wavelet resolutions.

  2. The method incorporates spatially correlated error terms, specifically by using low-rank decompositions of the error covariance matrix, which allows for a more flexible characterization and improved model fitting compared to methods that ignore correlated errors.

  3. The system can achieve improved computational complexity by partitioning the basis coefficients into smaller subsets and employing independent DP priors on these subsets, leading to complexity proportional to O(H J K2L) plus terms related to the low-rank covariance, which is better than LPP's O(HL2).

  4. The AI system can be used for robust posterior inference by implementing an efficient Gibbs sampler with a slice sampling technique that harnesses the number of possible clusters and reduces an infinite sum into a finite sum through auxiliary parameters.

  5. The proposed fPDPM framework provides theoretical guarantees of strong posterior consistency for density estimation under mild conditions, ensuring that the posterior distribution concentrates in arbitrarily small neighborhoods of the true density as sample size increases.

Abstract

There is a rich literature on clustering functional data with applications to time-series modeling, trajectory data, and even spatiotemporal applications. Existing methods are classified into two categories of global clustering that assigns an identical membership over the whole domain, and local clustering that allows multiple memberships based on the sub-domains. Classical literature has typically focused on global clustering that may struggle with degenerate clustering or produce numerous spurious clusters for high-dimensional functions. In contrast, a limited number of local clustering approaches have been proposed that are not scalable to high-dimensional functions and have limited theoretical guarantees. Focusing on basis expansions for high-dimensional functions, we propose a novel global-local framework that partitions coefficients into exclusive subsets and imposes independent Dirichlet process priors on each subset. The proposed framework reduces the total number of possible clustering configurations compared to local clustering and therefore provides greater scalability, while simultaneously providing enhanced flexibility compared to global clustering by relaxing the assumption of identical clustering for all elements in the high-dimensional function. Moreover, the proposed method incorporates spatially correlated errors to provide better model fitting and shows posterior consistency properties that asymptotically recover the true density of random functions. An efficient Markov chain Monte Carlo algorithm is developed for implementation. Extensive simulations involving images illustrate the improved scalability, clustering, and function estimation under the proposed method compared to classical approaches. We apply the proposed approach to a spatial transcriptomics application where the goal is to infer clusters of genes with distinct spatial patterns of expressions.

Related papers