Aligning Embeddings and Geometric Random Graphs: Informational Results and Computational Approaches for the Procrustes-Wasserstein Problem

The Procrustes-Wasserstein problem consists in matching two high-dimensional point clouds in an unsupervised setting, and has many applications in natural language processing and computer vision. We consider a planted model with two datasets $X,Y$ that consist of $n$ datapoints in $\mathbb{R}^d$, where $Y$ is a noisy version of $X$, up to an orthogonal transformation and a relabeling of the data points. This setting is related to the graph alignment problem in geometric models. In this work, we focus on the euclidean transport cost between the point clouds as a measure of performance for the alignment. We first establish information-theoretic results, in the high ($d \gg \log n$) and low ($d \ll \log n$) dimensional regimes. We then study computational aspects and propose the Ping-Pong algorithm, alternatively estimating the orthogonal transformation and the relabeling, initialized via a Franke-Wolfe convex relaxation. We give sufficient conditions for the method to retrieve the planted signal after one single step. We provide experimental results to compare the proposed approach with the state-of-the-art method of Grave et al. (2019).

Paper

References (27)

Scroll for more · 15 remaining

Similar papers

Peer review

Reviewer dVzJ5/10 · confidence 3/52024-07-04

Summary

The authors consider the alignment problem between a point cloud generated from a Gaussian distribution and its noisy version under orthogonal transformation, formulated as the Procrustes-Wasserstein problem. The authors derive information-theoretic results for both high and low dimensional regime (i.e., bounds on the maximum likelihood estimators to the optimal one). Additionally, the authors also propose an alternating algorithm, namely Ping-Pong algorithm, which is a variant of the algorithm in Grave et al. (2019). Empirically, the authors show advantages of the proposed algorithm over other baselines.

Strengths

+ The authors derive the upper bound (or lower bound) of the maximum likelihood estimator to the optimal one for the Procrustes-Wasserstein (PW) problem in both high and low dimensional regime. + The authors propose a variant alternating algorithm, namely Ping-Pong for the PW problem, and illustrate its advantages over other baselines.

Weaknesses

+ The title seems misleading. It is not clear about its relation with random graphs. + The motivation of the considered problem seems weak. It is better to elaborate the problem. (e.g., why it is interesting to consider point clouds generated from a Gaussian distribution and its noisy version under orthogonal transformation) + It is not clear about the advantage of the proposed Ping-Pong algorithm over the alternating algorithm considered in Grave et al. (2019). It is better to deepen analysis and empirical evidences.

Questions

It is interesting to generalize the problem considered in Kunisky and Niles-Weed (2022) into the considered Procrustes-Wasserstein problem. I have some following concerns: + The title seems misleading. Could the authors comment on the relation between the considered problems in line 29-39 with random graphs in the title? + Could the authors elaborate why it is interesting to consider Procrustes-Wasserstein problem between a point cloud generated from Gaussian distribution and its noisy version under orthogonal transformation? + For the high dimensional regime, why do we need lower-bound on \pi, but upper bound on Q? How’s about the upper-bound on \pi, and lower-bound on Q – do they have some roles in the considered problem setting? + For the Algorithm 1, why is it possible to fix T, and K? Does the algorithm converge? Is it possible to have some reasonable stopping conditions? + Could the authors clarify why the algorithm 1 is superior to the algorithm in Grave et al. (2019)? It is better to deepen the analysis and illustrate with rigorous empirical evidence? E.g., with large number of data points (e.g., in Figure 1, it shows for n <= 200). + In line 104-108, the authors emphasize the difference with other related works. Additionally, as in Figure 1, the authors also consider the case where d=2 and d=5, it is not clear why these approaches are not considered as baselines (as claimed in line 102-103). Could the authors clarify it? + When Q*= I, could the authors compare the presented results with those in Kunisky and Niles-Weed (2022)? + The Procrustes-Wasserstein problem is a non-convex problem. It is better to consider its initialization problem as well. It is not clear about the importance of the one-step analysis for Ping-pong algorithm in the context of nonconvex optimization? Could the authors clarify it? --------- The rebuttal has addressed some of my concerns, I increase the score to 5.

Rating

5

Confidence

3

Soundness

2

Presentation

2

Contribution

2

Limitations

It seems there is no discussion on the limitation.

Reviewer JfWm6/10 · confidence 3/52024-07-13

Summary

This paper studies the Procrustes-Wasserstein problem that aims to match two high-dimensional point clouds where one is a noisy version of the other up to an orthogonal transformation. The authors establish information-theoretic results in the high ($d \gg \log n$) and low ($d \ll \log n$) dimensional regimes. Further, the authors also propose a "Ping-Pong algorithm" that alternatively estimates the orthogonal transformation and the matching. Sufficient conditions for the method to recover the planted signal after one step is provided. The theoretical finds are also supported by numerical experiments.

Strengths

1. This paper defines a planted model for the Procrustes-Wasserstein problem that extends the work of Kunisky and Niles-Weed [2022] and Wang et al. [2022]. 2. Focusing on the $L_2$ transport cost between the point clouds, in contrast to previous works that mostly consider the overlap, the authors established information-theoretic limits in the low-dimensional regime ($d \ll \log n$), which substantially differ from those of Wang et al. [2022], and in the high-dimensional regime ($d \gg \log n$), which was not explored before. 3. A "Ping-Pong Algorithm", first initialized by a Franke-Wolfe convex relaxation, then alternatively estimating the orthogonal transformation and the relabeling, is proposed. Statistical guarantees for *one single step* of the algorithm is analyzed.

Weaknesses

1. Due to technical challenges, only guarantees for one step of the proposed algorithm is analyzed. 2. The recovery guarantees are only provided in the overlap metric rather than the $c^2$ loss which is claimed to behave very differently when $d$ is small. 3. The dependency on the noise parameter $\sigma$ in the statistical rates seems to be different from that of the ML estimators. The tightness of the results in terms of $\sigma$ is unknown.

Questions

1. How do the statistical guarantees for the proposed algorithm compare to the information-theoretic results? It would be nice if the authors could remark on this. 2. Are there any negative information-theoretic results, i.e., lower bounds on the costs, for the Procrustes-Wasserstein Problem of the planted model? 3. Can there possibly be computational-statistical gaps for the problem?

Rating

6

Confidence

3

Soundness

3

Presentation

3

Contribution

3

Limitations

The authors addressed most limitations of the work. Some additional ones are highlighted in Weaknesses and Questions sections.

Reviewer ERNU8/10 · confidence 3/52024-07-13

Summary

This submission is concerned with the problem of aligning planted graphs. To this end, both a permutation $\pi$ and an orthogonal matrix $Q$ must be estimated from observations of $X$ and $Y$. To evaluate the performance of this approach, it is proposed to measure to measure the error induced by $Q$ in terms of the squared Frobenius norm to the true underlying matrix $Q^{\star}$ (denoted $\ell^2(Q,Q^{\star})$) whereas the error induced by $\pi$ is given by $c^2(\pi,\pi^{\star})=\frac 1 n\sum_{i=1}^n||x_{\pi(i)}-x_{\pi^{\star}(i)}||^2$. The first contribution of this work is to provide information-theoretic results for this problem. Namely, the simple maximum likelihood estimators for $\hat \pi$ (which can be efficiently computed) and $\hat Q$ (which has a closed form solution) recover $\pi^{\star}$ and $Q^{\star}$ almost exactly in the limit of vanishing noise in the high-dimensional regime ($d\geq 2\log n$). In the low-dimensional regime $d\ll \log n$, it is shown that if the noise is of the order $o(d^{-1/2})$, then estimators $\hat Q,\hat \pi$ can be formulated for which $c^2( hat \pi,\pi^{\star})=o(d)$ and $\ell^2(\hat Q,Q^{\star})=o(d)$. It is shown, moreover that the good performance of one of the two estimators can be transferred to the estimator in a straightforward (and computationally tractable) manner. Next, as the PW problem is equivalent to the quadratic assignment problem (QAP) which is known to be NP-hard in general, a convex relaxation of the QAP is considered which consists of maximizing the objective over all bistochastic matrices. The proposed algorithm for approximating the QAP consists of first solving the relaxed problem via the Frank-Wolfe algorithm then using this solution to hot start an alternating minimization procedure for the permutation matrix and the orthogonal matrix. Guarantees for one step of this algorithm are then provided. The paper concludes with some numerical experiments which shows that the proposed Ping-Pong algorithm generally outperforms direct resolution of the relaxed QAP and the method of Grave et al.

Strengths

The paper is well-written and places itself well within the existing literature on this problem. To my understanding, the high dimensional results are first of their kind whereas the low dimensional results improve on previously known results. The proposed algorithm is also seen to perform best out of the other considered.

Weaknesses

The analysis of the proposed ping-pong algorithm is quite limited. The one-step result provided in Proposition 1 is already an interesting step in the analysis, but it would be interesting to see a more refined analysis. Given the fact that the QAP problem is NP hard it is clear that we cannot expect particularly strong results in general (e.g. convergence to global minimizer), but it may be possible to discern some properties of the matrices to which the algorithm converges or to provide a bound on the number of iterations required in certain cases. At the very least it would be useful to recall the per iteration complexity.

Questions

The authors are kindly requested to answer to the above point. I noted the following typos: 1. line 139: is its probability -> if its probability. 2. line 279: envelop -> envelope.

Rating

8

Confidence

3

Soundness

3

Presentation

4

Contribution

3

Limitations

The assumptions are clearly stated in each relevant result.

Reviewer 6nbD7/10 · confidence 3/52024-07-23

Summary

This paper studies the theoretical limits of the Procruste-Wasserstein problem which consists in finding an optimal assigment of cloud of points in Euclidean space up to a global rotation. The paper focuses on the model of a cloud of point perturbed by a global rotation and some gaussian noise. The authors establish theoretical results on the recovery of the optimal rotation and the optimal assigment in two cases of interest: high-dimension/low number of samples and low dimensions/high number of samples. They also propose a Frank-Wolfe type of algorithm with an initial step based on convex relaxation: this algorithm is shown to perform better than other state of the art algorithm.

Strengths

The paper is extremely well-written, it is easy to read. In my opinion, the questions addressed in the paper are of interest to the community interested in geometric alignment. The improvement upon Grave's algorithm seems interesting as well. The estimator in the low-dimensional case introduced the idea of slicing along one-dimensional subspace and it is remarkable that this idea which was used in other contexts, works well in this low-dimensional problem.

Weaknesses

The theoretical results are interesting for a relatively narrow audience of NeurIPS.

Questions

Why is the conical alignment loss not used in practical experiments? Is it because it gives poor practical performances?

Rating

7

Confidence

3

Soundness

3

Presentation

4

Contribution

3

Limitations

not applicable.

Reviewer ERNU2024-08-11

I thank the authors for clarifying these points. I concur that a more in-depth study of the ping-pong algorithm is of a great interest, but is likely complicated and would be deserving of a separate paper. The addition of the overall complexity is useful in my opinion.

Reviewer dVzJ2024-08-11

I thank the authors for your explanation in the rebuttal. It addresses some of my concerns, and I increase the score to 5. I have some quick questions as follows: **(1) For GGA (geometric graph alignment)** - Could you elaborate with further details about applying an orthogonal map $Q^*$ to all the vectors leaves the scalar products invariant? I agree about the role of $P^*$, but it is still unclear about $Q^{*}$ in the GGA? - Do we only try to recover the node correspondence? how's about the role of edge weights in the GGA? **(2) For the proposed Ping-Pong algorithm** - In Grave et al. (2019), they also consider the convex relaxation and leverage the Frank-Wolfe algorithm as in the proposed algorithm. Could you comment about it? It is better in case you could give some analysis/discussion and/or empirical illustration for them? I will further adjust the score accordingly.

Authorsrebuttal2024-08-12

Response and clarifications

Thank you for your questions. **(1)** In GGA, whatever the orthogonal transformation applied, the scalar product is unchanged: $\langle Qx_i, Qx_j\rangle = \langle x_i,x_j\rangle$. We thus only seek to retrieve the permutation, since the GGA problem is invariant by orthogonal transformation. As for the role of edge weights, it is unclear for us what the question is: our point above is that the edge weights observed are invariant under orthogonal transformations. **(2)** Grave et al indeed use the same initialization. Our paper distinguishes itself from their work in two ways. The first is in the methodology: the *amplification* step is more greedy and thus more efficient, while having the same computational cost. The second is that we propose a 1-step analysis, paving the way for a more general analysis.

Reviewer dVzJ2024-08-12

Thank you for your answer. I have some follow-up questions **(1) about Lemma 1** - The PW problem is non-convex for both (P, Q). However, in the proof of Lemma 1, line 119, it mention about the (sub)optimal $\pi*$ (which lead to the optimal $Q^*$). It is still unclear how to approximate such $\pi^*$ since the optimization problem involves both $(\pi, Q)$. Could you comment about it? **(2) about GGA** - Is it right that in GGA, we only care about the node corresponding $P^*$. Do we care about $Q^*$ in GGA? I confused about the role of $Q^*$ in GGA, do we really need to use $G$ in GGA?

Authorsrebuttal2024-08-13

Answers to follow-up

Thank you again for your additional questions which we hope to clarify in the following. Let's start by recalling the two problems (GGA) and (PW). In both cases, we consider two point clouds written into matrices, namely $X = (x_1 | x_2 | .... | x_n)$ and $Y = (y_1 | y_2 | .... | y_n)$ with $x_i, y_i \in \mathbb{R}^d$. In this setting, (GGA) amounts to solving the minimization problem $ P^*_{GGA} \in \mathrm{argmin}_{P \in \mathcal{S}_n} \Vert X^\top X P - P Y^\top Y \Vert_F^2 $ while (PW) amounts to solving $ (P^*_{PW}, Q^*_{PW}) \in \mathrm{argmin}_{(P,Q) \in \mathcal{S}_n \times \mathcal{O}_d} \Vert QX - YP \Vert_F^2 $ Lemma 3 tells us that in (PW), it suffices to retrieve either $P^*_{PW}$ or $Q^*_{PW}$ to retrieve a very good guess of its respective counterpart. If we understand you correctly, the key to both of your questions lies in the following "soft reformulation" of the optimization problem which was performed by (Grave, Joulin, Berthet, 2018) in their subsection **Convex relaxation**. The idea is to consider the following problem which is equivalent to (PW): $ \text{max}_{P \in \mathcal{S}_n} \max_Q \text{tr}(Q X P^T Y^T) $ One has $ \max_Q \text{tr}(Q X P^T Y^T) = \Vert X P^T Y^T \Vert_\star $ where $\Vert A \Vert_\star$ denotes the trace norm (or nuclear norm) of a matrix $A$. If we put the singular values $\sigma_1, ..., \sigma_m$ of $A$ into a vector $v := (\sigma_1, ..., \sigma_m)$, then $\Vert A \Vert_\star= \sum_i \sigma_i = \Vert v \Vert_1$. Since $\Vert A \Vert_F = \sqrt{\sum_i \sigma_i^2} = \Vert v \Vert_2$, one can argue, using norm equivalence in $ \mathbb{R}^m$, that the problems $ \max_P \Vert X P^T Y^T \Vert_\star $ and $ \max_P \Vert X P^T Y^T \Vert_F^2 $ are very similar. [Note: We don't know a result to quantify how similar the solutions are, this is why this whole "soft reformulation" is more about giving intuition than providing rigorous results.] Developing the square in $ \max_P \Vert X P^T Y^T \Vert_F^2 $ and omiting constant terms, this problem is further equivalent to $ \max_P \mathrm{tr}( Y P X^T X P^T Y^T ) = \text{max}_P \mathrm{tr}( X^T X P^T Y^T Y P) $ Rewriting this as the minimum of the squared norm of a matrix distance, one obtains exactly (GGA). This whole reformulation process can also be reversed to move back, from (GGA) to (PW). Now, let's move to answering your concrete questions: **(1) about Lemma 1** It is true that in order to infer $Q^*$, one needs a good estimate of $P^*$. To obtain this estimate, we use the above "soft reformulation" and first solve (GGA) which is agnostic to $Q^*$. This is done via convex relaxation of $\mathcal{S}_n$ to the set of doubly stochastic matrices, followed by Frank-Wolve optimisation. This procedure is what Grave et al. call the initialisation to their algorithm; it is also the initialisation to our Ping-Pong steps. The doubly stochastic matrix obtained at convergence of Frank-Wolfe if our initial guess of $P^*$. Our one-step analysis of the algorithm lets us hope that this is indeed a good guess in general. **(2) about GGA** You are right that there is no retrieval of $Q^*$ in GGA. Our hope is that the "soft reformulation" can give a good idea where this $Q^*$ is hidden in the process. Since the geometric graph in GGA only depends on the scalar products between the vectors in $X$ and $Y$, our proof of lemma 1 explains in detail how an orthogonal transformation between the point clouds emerges when doing the problem reformulation (GGA) $\to$ (PW).

Reviewer JfWm2024-08-12

Reply to rebuttal

I thank the authors for the detailed responses. I maintain my postive rating of the paper.

Program Chairsdecision2024-09-25

Decision

Accept (poster)

© 2026 NYSGPT2525 LLC