- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 1
- Soundness
- 2
Read review
Paper summary
The paper generalises Siegel-space neural networks to a larger family of complex matrix spaces and builds neural network layers on one such space, termed HB_n. It evaluates these networks on radar clutter classification, action recognition, graph reconstruction, and node classification.
Critical strengths
- The proposed models outperform Siegel and SPD baselines on most benchmarks across several different tasks (Tables 1-4).
- The ablations (Table 5, Fig 1) suggest that the new geometry contributes to the gains, which is the paper’s main claim.
Critical weaknesses
- The theoretical sections are very dense and give little intuition for a general ML reader, which makes the contribution hard to assess.
- The method is slower to train than the Siegel baseline (about 1.5× per batch - see Table 8), which the authors already acknowledge as a limitation in Appendix D. A brief discussion of possible ways to reduce this cost would be helpful.
Questions for authors
- Could the authors give some intuition as to why HB_n is a good fit for these tasks? This would help readers who are less familiar with the underlying mathematics.
- HBNetBN has more parameters than SiegelNetBN (see Table 7). Have the authors compared the two at a similar parameter count? This would help me judge how much of the gain comes from the HB_n geometry itself.
3 comments
Q2: HBNetBN has more parameters than SiegelNetBN (see Table 7). Have the authors compared the two at a similar parameter count? This would help me judge how much of the gain comes from the HB_n geometry itself.
We thank the reviewer for raising this question. We agree that this is important to evaluate the real geometric benefits of the proposed domain.
For the radar clutter classification and human action recognition experiments, our model consists of a BN layer and a classification layer, which is a minimal architecture and cannot be further simplified. This is also the case of the Siegel models (i.e., SiegelNetFC and SiegelNetBN). For the SPD models (i.e., SPDNetBN and LieBN), we used the best architectures in the original papers. Thus, matching the number of trainable parameters is not necessarily a meaningful way to ensure a fair comparison because it would require artificially constraining our model or the baselines without obtaining a better interpretation of the final results. Note that for these experiments, we cannot change the dimensions of input matrices for the models (so that the models are matched in terms of representation dimension as in the graph reconstruction and node classification experiments) because these dimensions are determined by the dimensions of input time series and of video sequences.
For the node classification experiments, our model has fewer learnable parameters than the SPD model (see Appendix A.4.3). We performed additional experiments where our model and the SPD model are matched in terms of representation dimension. Specifically, we learn embeddings on and . The real dimensions of and are 128 and 136, respectively. The results are provided in Table E. We can observe that increasing the dimension of the SPD embeddings has little impact on mean accuracy of the SPD model.
| Model | Airport | Pubmed | Cora | Disease |
|---|---|---|---|---|
| SPD-GCN | 82.8$\pm$0.9 | 78.5$\pm$0.4 | 82.1$\pm$0.6 | 96.6$\pm$2.5 |
| HB-GCN | 91.1$\pm$0.7 | 79.2$\pm$0.4 | 82.6$\pm$0.5 | 97.2$\pm$1.4 |
Table E: Node classification results (mean accuracy standard deviation) across ten runs.
We now compare our model, the Siegel models, and the SPD model on the graph reconstruction task where they are matched in terms of representation dimension. Specifically, we learn embeddings on , , , and . The real dimensions of , , , and are 72, 72, 72, and 78, respectively. The results are provided in Table F.
| Space | Grid | Tree | Tree Grid | Grid Tree | Tree Grid | Grid Tree |
|---|---|---|---|---|---|---|
| SPD$^R_{12}$ | 14.48$\pm$0.03 | 5.81$\pm$0.04 | 8.43$\pm$0.02 | 8.56$\pm$0.02 | 1.84$\pm$0.03 | 1.59$\pm$0.03 |
| S$^{F_1}_{8}$ | 0.01$\pm$0.0 | 1.02$\pm$0.08 | 0.44$\pm$0.05 | 0.43$\pm$0.04 | 0.52$\pm$0.06 | 0.48$\pm$0.05 |
| B$^{F_1}_{8}$ | 0.01$\pm$0.0 | 0.82$\pm$0.04 | 0.41$\pm$0.03 | 0.42$\pm$0.03 | 0.40$\pm$0.03 | 0.43$\pm$0.03 |
| 0.01$\pm$0.0 | 0.80$\pm$0.04 | 0.50$\pm$0.03 | 0.48$\pm$0.02 | 0.37$\pm$0.02 | 0.40$\pm$0.04 |
Table F: Graph reconstruction results (mean distortion standard deviation) across five runs.
In terms of mean distortion, our model and the Siegel models significantly outperform the SPD model, and our model achieves mean accuracy comparable to or better than the best Siegel model (i.e., B$^{F_1}_{8}$) on 4 of the 6 datasets.
Finally, the training times of the competing models on the graph reconstruction task are reported in Table G. For comparable representation dimensions, our model is computationally more efficient during the training phase than the competing models. It is important to note that the training time of the best Siegel model is about 4$\times$ that of our model.
| Space | Grid | Tree | Tree Grid | Grid Tree | Tree Grid | Grid Tree |
|---|---|---|---|---|---|---|
| SPD$^R_{12}$ | 1.08 | 0.03 | 25.53 | 25.47 | 8.28 | 20.96 |
| S$^{F_1}_{8}$ | 1.95 | 0.05 | 48.73 | 49.36 | 15.80 | 43.15 |
| B$^{F_1}_{8}$ | 2.92 | 0.07 | 83.32 | 83.44 | 24.56 | 69.94 |
| 0.94 | 0.02 | 20.47 | 20.36 | 6.94 | 17.68 |
Table G: Training times (seconds) per epoch on the graph reconstruction task.
Thank you for the detailed response and additional experiments. The intuition that you have given for each task is helpful, and the dimension-matched results in Tables E and F address my concern for the graph tasks. I also appreciate the training-time comparison in Table G. I will maintain my score of 3 (weak acceptance), noting that I am not an expert in the underlying mathematics.
We thank the reviewer for their insightful questions, which provides us an opportunity to clarify and strengthen our work.
Q1: Could the authors give some intuition as to why HB_n is a good fit for these tasks? This would help readers who are less familiar with the underlying mathematics.
For the radar clutter classification task, by assumption, the time series are stationary centered autoregressive (AR) Gaussian time series. Consequently, they can be represented in a product space of HBn and Symn+ (Cabanes, 2022). Specifically, each time series can be represented in product space Symn+×SDnr−1, where r (r>1) is the order of the AR model. Since SDn is a subdomain of HBn, both these spaces are appropriate representation spaces for this task.
For the human action recognition task, we have shown that meaningful mappings between the proposed bounded domain and spaces of matrices with positive-definiteness constraints (i.e., SPD and HPD manifolds) can be built (see Proposition 3.6 and Definition 5.2). Since SPD manifolds have been extensively used as representation spaces for human action recognition, HBn could potentially be effective for this task. This was the main motivation behind the use of HBn for human action recognition experiments.
Regarding the graph reconstruction and node classification tasks, our work was mainly motivated by the results in López et al. (2021) where the authors have shown that Siegel models when equipped with vector-valued distances outperform SPD models on these tasks. Since SDn is a subdomain of HBn, the latter can provide geometric inductive biases that the former may fail to capture. This has been demonstrated by our experimental results provided in Tab. 3.