The reviewed record of science sign in
Pith

arxiv: 2605.27144 · v1 · pith:ZSLLK4PJ · submitted 2026-05-26 · cs.CV · cs.LG

Is an Image Also Worth 16x16=256 Superpixels? A Framework for Attentional Image Classification

Pedro Henrique da Costa Avelar , Anderson R. Tavares , Lu\'is C. Lamb This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-06-29 17:46 UTCgrok-4.3pith:ZSLLK4PJrecord.jsonopen to challenge →

classification cs.CV cs.LG
keywords superpixel transformersvision transformersimage classificationgraph attention networkssuperpixel segmentationpositional encoding
0
0 comments X

The pith

Superpixel Transformers process images as graphs of 256 superpixels to match Vision Transformer accuracy while beating prior GNN methods.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces Superpixel Transformers (SPT) as a framework that combines superpixel image representations with transformer-style self-attention. It generalizes earlier graph attention models on superpixels and standard Vision Transformers to handle arbitrary superpixel chunking, connectivity graphs, and positional encodings. Refinements include multidimensional sine-cosine positional encodings and an enriched patch structure that carries superpixel shape and color details. Tests on CIFAR10, FashionMNIST, and Imagenette across multiple superpixel generators and graph strategies show SPT surpassing previous superpixel GNN approaches and remaining competitive with ViTs. The work argues that this unification reduces information loss from pixel aggregation and demonstrates how graph constraints can improve transformer performance.

Core claim

SPT generalizes the SICGAT model and ViT to arbitrary superpixel-based chunking strategies, connectivity graphs, and positional encodings; the multidimensional sine-cosine encoding and enriched patch data structure that incorporates superpixel shape and color allow SPT to outperform prior superpixel GNN methods while remaining competitive with ViTs on CIFAR10, FashionMNIST, and Imagenette.

What carries the argument

Superpixel Transformer (SPT) architecture that applies self-attention to superpixel patches via enriched data structures and multidimensional sine-cosine positional encodings.

If this is right

  • SPT supports arbitrary superpixel generation methods and graph connectivities without architectural changes.
  • Constrained graph connectivity within SPT can improve performance over unconstrained ViT patch attention.
  • The enriched patch representation reduces information loss compared to simple pixel aggregation in SICGAT.
  • SPT opens hybrid attentional models that mix superpixel irregularity with transformer self-attention.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same enriched superpixel structure might reduce token count and compute in high-resolution images by using fewer, larger superpixels.
  • SPT-style models could transfer more naturally to medical or satellite imagery where superpixel boundaries align with anatomical or land-use edges.
  • Testing SPT with learned superpixel generators instead of fixed algorithms would clarify whether the performance gains depend on the segmentation method.

Load-bearing premise

The multidimensional sine-cosine positional encoding and enriched superpixel patch structure integrate into the transformer without introducing unaddressed information loss or needing dataset-specific tuning beyond the tested cases.

What would settle it

SPT underperforming both prior superpixel GNN methods and standard ViTs by a clear margin on a new dataset such as ImageNet with the same superpixel and connectivity settings.

Figures

Figures reproduced from arXiv: 2605.27144 by Anderson R. Tavares, Lu\'is C. Lamb, Pedro Henrique da Costa Avelar.

Figure 1
Figure 1. Figure 1: The ViT architecture [16] (left) and the SICGAT architecture [5] (right). Note that, apart from [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The Transformer Encoder [38] used on the [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The information loss that happens when converting an image to a superpixel representation and [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Our proposed Superpixel Transformer (SPT) architecture. Differently from SICGAT [5] we use [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Plots highlighting the Validation Accuracy throughout the epochs for all the configurations we [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Plots highlighting the Validation Accuracy throughout the epochs for all the configurations we [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
read the original abstract

Superpixel-based image classification has traditionally leveraged graph neural networks (GNNs) for processing irregular image representations. Recent advances in computer vision, driven by Vision Transformers (ViTs), have introduced new paradigms in self-attentional models, surpassing convolutional neural networks (CNNs) in various tasks. However, a synergistic connection between GNNs, superpixels, and transformers remains unexplored. In this work, we propose Superpixel Transformers (SPT), a novel framework that unifies superpixel-based image classification and ViTs. SPT generalizes the Superpixel Image Classification with Graph Attention Networks (SICGAT) model and ViT to support arbitrary superpixel-based chunking strategies, connectivity graphs, and positional encodings. We introduce refinements including a multidimensional sine-cosine positional encoding and an enriched patch data structure that fully incorporates superpixel shape and color information. By testing SPT across datasets such as CIFAR10, FashionMNIST, and Imagenette, with various superpixel generation and graph connectivity strategies, we demonstrate that SPT achieves superior performance compared to previous superpixel-based GNN methods and remains competitive with ViTs. Notably, our approach addresses the limitations of SICGAT, such as information loss during pixel aggregation, and shows how constrained graph connectivity can enhance ViT performance. SPT bridges the gap between superpixel-based and transformer models, opening avenues for cross-domain generalization and future innovations in hybrid attentional frameworks, and showing that an image can also be worth $16\times16$ superpixels.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes Superpixel Transformers (SPT), a framework unifying superpixel-based image classification (via GNNs) with Vision Transformers. SPT generalizes SICGAT and ViT to arbitrary superpixel chunking, connectivity graphs, and positional encodings; it introduces a multidimensional sine-cosine positional encoding and an enriched patch representation that incorporates superpixel shape and color. Experiments on CIFAR-10, FashionMNIST, and Imagenette with multiple superpixel generators and graph connectivities are claimed to show SPT outperforming prior superpixel GNN methods while remaining competitive with ViTs, while mitigating information loss during aggregation.

Significance. If the reported performance gains are robustly supported, SPT provides a concrete bridge between irregular superpixel graphs and self-attention, with potential utility for hybrid models that combine the efficiency of superpixels with transformer expressivity. The generality across chunking strategies and the constrained-connectivity enhancement to ViTs are notable strengths.

major comments (2)
  1. [Abstract] Abstract: the central claim of superior performance over prior GNN methods and competitiveness with ViTs is asserted without any quantitative metrics, error bars, baseline details, or statistical tests, which is load-bearing for assessing the empirical contribution.
  2. [§3] §3 (Architecture): the multidimensional sine-cosine positional encoding and enriched patch data structure (shape + color) are presented as the mechanisms that recover lost information and generalize SICGAT/ViT, but the exact formulas, dimensionality, and integration into the transformer attention layers are not specified, preventing verification that no unaddressed information loss or dataset-specific tuning is introduced.
minor comments (2)
  1. [§4] The experimental protocol (training details, number of runs, hyperparameter matching across baselines) should be stated explicitly in §4 to support reproducibility of the cross-method comparisons.
  2. [§4] Notation for graph connectivity strategies and superpixel generators could be standardized in a single table for clarity when comparing results across datasets.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments and the recommendation of minor revision. We address each major comment below and will update the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim of superior performance over prior GNN methods and competitiveness with ViTs is asserted without any quantitative metrics, error bars, baseline details, or statistical tests, which is load-bearing for assessing the empirical contribution.

    Authors: We agree that the abstract would benefit from quantitative support. In the revised version we will insert specific accuracy figures (with standard deviations from repeated runs) comparing SPT to the cited GNN baselines and to ViT on each dataset, along with a brief note on the evaluation protocol. revision: yes

  2. Referee: [§3] §3 (Architecture): the multidimensional sine-cosine positional encoding and enriched patch data structure (shape + color) are presented as the mechanisms that recover lost information and generalize SICGAT/ViT, but the exact formulas, dimensionality, and integration into the transformer attention layers are not specified, preventing verification that no unaddressed information loss or dataset-specific tuning is introduced.

    Authors: We acknowledge that the current presentation of these components could be more explicit. The revised §3 will include the precise mathematical definitions of the multidimensional sine-cosine encoding (including its dimensionality and concatenation with patch embeddings), the exact feature vector for the enriched superpixel patches, and the manner in which both are supplied to the multi-head attention layers. These additions will also clarify that the encoding is dataset-agnostic and does not introduce hidden per-dataset tuning. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The manuscript presents SPT as an empirical framework that generalizes SICGAT and ViT via new positional encodings and enriched superpixel patches, then reports test accuracies on CIFAR-10, FashionMNIST and Imagenette under multiple generators and connectivities. No equations, derivations or load-bearing self-citations appear that reduce the claimed performance gains to fitted inputs or prior author results by construction; the evidence consists of direct experimental outcomes.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Framework rests on standard CV domain assumptions about superpixel utility and transformer compatibility; no free parameters or invented entities explicitly introduced in abstract.

axioms (1)
  • domain assumption Arbitrary superpixel-based chunking strategies, connectivity graphs, and positional encodings can be supported without breaking the attentional mechanism
    Invoked in the generalization of SICGAT and ViT to SPT.

pith-pipeline@v0.9.1-grok · 5847 in / 1223 out tokens · 50760 ms · 2026-06-29T17:46:45.659858+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

48 extracted references · 5 canonical work pages · 3 internal anchors

  1. [1]

    Achanta, A

    R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua, and Sabine S¨ usstrunk. SLIC Superpixels Compared to State-of-the-Art Superpixel Meth- ods.IEEE Transactions on Pattern Analysis and Machine Intelligence, 34(11):2274–2282, Novem- ber 2012

    2012

  2. [2]

    SLIC Superpixels: SLICO, June 2010

    Radhakrishna Achanta, Appu Shaji, Kevin Smith, Aurelien Lucchi, Pascal Fua, and Sabine S¨ usstrunk. SLIC Superpixels: SLICO, June 2010

    2010

  3. [3]

    Superpixels and Polygons Using Simple Non- iterative Clustering

    Radhakrishna Achanta and Sabine Susstrunk. Superpixels and Polygons Using Simple Non- iterative Clustering. In2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 4895–4904, Honolulu, HI, July

  4. [4]

    Multitask Learning on Graph Neural Networks: Learning Multiple Graph Cen- trality Measures with a Unified Network

    Pedro Avelar, Henrique Lemos, Marcelo Prates, and Luis Lamb. Multitask Learning on Graph Neural Networks: Learning Multiple Graph Cen- trality Measures with a Unified Network. In Igor V. Tetko, Vˇ era K˚ urkov´ a, Pavel Karpov, and Fabian Theis, editors,Artificial Neural Net- works and Machine Learning – ICANN 2019: Workshop and Special Sessions, volume 1...

    2019

  5. [5]

    Pedro H. C. Avelar, Anderson R. Tavares, Thi- ago L. T. da Silveira, Cl´ audio R. Jung, and Luis C. Lamb. Superpixel Image Classification with Graph Attention Networks. In2020 33rd SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI), pages 203–209, Re- cife/Porto de Galinhas, Brazil, November 2020. IEEE

    2020

  6. [6]

    Neural Machine Translation by Jointly Learning to Align and Translate

    Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. Neural Machine Translation by Jointly Learning to Align and Translate. In Yoshua Bengio and Yann LeCun, editors,3rd International Conference on Learning Represen- tations, ICLR 2015, Conference Track Proceed- ings, San Diego, CA, USA,, May 2015

    2015

  7. [7]

    Relational inductive biases, deep learning, and graph networks

    Peter W. Battaglia, Jessica B. Hamrick, Victor Bapst, Alvaro Sanchez-Gonzalez, Vinicius Zam- baldi, Mateusz Malinowski, Andrea Tacchetti, David Raposo, Adam Santoro, Ryan Faulkner, Caglar Gulcehre, Francis Song, Andrew Ballard, Justin Gilmer, George Dahl, Ashish Vaswani, Kelsey Allen, Charles Nash, Victoria Langston, Chris Dyer, Nicolas Heess, Daan Wierst...

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  8. [8]

    Bordes, F., Garrido, Q., Kao, J., Williams, A., Rabbat, M., and Dupoux, E

    Lucas Beyer, Xiaohua Zhai, and Alexander Kolesnikov. Better plain ViT baselines for ImageNet-1k, May 2022. arXiv:2205.01580 [cs]

    work page arXiv 2022

  9. [9]

    Bianchini, M

    M. Bianchini, M. Maggini, L. Sarti, and F. Scarselli. Recursive neural networks learn to localize faces.Pattern Recognition Letters, 26(12):1885–1895, September 2005

    2005

  10. [10]

    How Attentive are Graph Attention Networks? In The Tenth International Conference on Learn- ing Representations, ICLR 2022, Virtual Event, April 2022

    Shaked Brody, Uri Alon, and Eran Yahav. How Attentive are Graph Attention Networks? In The Tenth International Conference on Learn- ing Representations, ICLR 2022, Virtual Event, April 2022. OpenReview.net

    2022

  11. [11]

    Jiawei Chen, Hao Xu, Wanyu Tao, Zhaoxiong Chen, Yuxuan Zhao, and Jing-Dong J. Han. Transformer for one stop interpretable cell type annotation.Nature Communications, 14(1):223, January 2023. 14

    2023

  12. [12]

    Remote Sensing Image Scene Classification: Benchmark and State of the Art.Proceedings of the IEEE, 105(10):1865–1883, October 2017

    Gong Cheng, Junwei Han, and Xiaoqiang Lu. Remote Sensing Image Scene Classification: Benchmark and State of the Art.Proceedings of the IEEE, 105(10):1865–1883, October 2017

    2017

  13. [13]

    scGPT: To- wards Building a Foundation Model for Single- Cell Multi-omics Using Generative AI

    Haotian Cui, Chloe Wang, Hassaan Maan, Kuan Pang, Fengning Luo, and Bo Wang. scGPT: To- wards Building a Foundation Model for Single- Cell Multi-omics Using Generative AI. preprint, Bioinformatics, May 2023

    2023

  14. [14]

    ImageNet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. ImageNet: A large-scale hierarchical image database. In2009 IEEE Con- ference on Computer Vision and Pattern Recog- nition, pages 248–255, Miami, FL, June 2009. IEEE

    2009

  15. [15]

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Jill Burstein, Christy Doran, and Thamar Solorio, editors,Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguis- tics: Human Language Technologies, NA...

    2019

  16. [16]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

    Alexey Dosovitskiy, Lucas Beyer, Alexan- der Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In9th International Conference on Learning Representations, ICLR 2021,...

    2021

  17. [17]

    A generalization of transformer networks to graphs,

    Vijay Prakash Dwivedi and Xavier Bresson. A Generalization of Transformer Networks to Graphs, January 2021. arXiv:2012.09699 [cs]

    work page arXiv 2021

  18. [18]

    SplineCNN: Fast Geometric Deep Learning with Continuous B- Spline Kernels

    Matthias Fey, Jan Eric Lenssen, Frank We- ichert, and Heinrich Muller. SplineCNN: Fast Geometric Deep Learning with Continuous B- Spline Kernels. In2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 869–877, Salt Lake City, UT, June 2018. IEEE

    2018

  19. [19]

    Schoenholz, Patrick F

    Justin Gilmer, Samuel S. Schoenholz, Patrick F. Riley, Oriol Vinyals, and George E. Dahl. Neu- ral Message Passing for Quantum Chemistry. In Doina Precup and Yee Whye Teh, editors, Proceedings of the 34th International Conference on Machine Learning, ICML 2017, volume 70 ofProceedings of Machine Learning Research, pages 1263–1272, Sydney, NSW, Australia, ...

    2017

  20. [20]

    Bidirectional LSTM Networks for Improved Phoneme Classification and Recogni- tion

    Alex Graves, Santiago Fern´ andez, and J¨ urgen Schmidhuber. Bidirectional LSTM Networks for Improved Phoneme Classification and Recogni- tion. In Wlodzislaw Duch, Janusz Kacprzyk, Erkki Oja, and Slawomir Zadrozny, editors,Ar- tificial Neural Networks: Formal Models and Their Applications - ICANN 2005, 15th Inter- national Conference, Warsaw, Poland, Sept...

    2005

  21. [21]

    Shion Honda, Shoi Shi, and Hiroki R. Ueda. SMILES Transformer: Pre-trained Molecular Fingerprint for Low Data Drug Discovery. 2019. Publisher: arXiv Version Number: 1

    2019

  22. [22]

    fastai/imagenette, November 2024

    Jeremy Howard and Imagenette Contributors. fastai/imagenette, November 2024. original- date: 2019-03-06T01:58:45Z

    2024

  23. [23]

    VGG model construction and Imagenette image recognition based on PyTorch, September 2024

    Kasen. VGG model construction and Imagenette image recognition based on PyTorch, September 2024

    2024

  24. [24]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A Method for Stochastic Optimization. In Yoshua Bengio and Yann LeCun, editors,3rd Interna- tional Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, 2015

    2015

  25. [25]

    Kipf and Max Welling

    Thomas N. Kipf and Max Welling. Semi- Supervised Classification with Graph Convolu- tional Networks. In5th International Conference on Learning Representations, ICLR 2017, Con- ference Track Proceedings, Toulon, France, April

    2017

  26. [26]

    Learning multiple layers of fea- tures from tiny images

    Alex Krizhevsky. Learning multiple layers of fea- tures from tiny images. Master’s thesis, Univer- sity of Toronto, Toronto, Canada, April 2009. 15

    2009

  27. [27]

    VGG16 on MNIST and Fash- ion mnist, November 2023

    Shreyasi Mandal. VGG16 on MNIST and Fash- ion mnist, November 2023

    2023

  28. [28]

    Bronstein

    Federico Monti, Davide Boscaini, Jonathan Masci, Emanuele Rodola, Jan Svoboda, and Michael M. Bronstein. Geometric Deep Learning on Graphs and Manifolds Using Mixture Model CNNs. In2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 5425–5434, Honolulu, HI, July 2017. IEEE

    2017

  29. [29]

    OpenAI, Josh Achiam, Steven Adler, Sand- hini Agarwal, Lama Ahmad, Ilge Akkaya, Flo- rencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anad- kat, Red Avila, Igor Babuschkin, Suchir Balaji, Valerie Balcom, Paul Baltescu, Haiming Bao, Mo Bavarian, Jeff Belgum, Irwan Bello, Jake Berdine, Gabriel Bernadett-Shapiro, Christo- pher Bern...

  30. [30]

    arXiv:2303.08774 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv

  31. [31]

    Graph Convolutional Networks for Image Classifica- tion: Comparing Approaches for Building Graphs from Images

    J´ ulia Rodrigues and Joel Carbonera. Graph Convolutional Networks for Image Classifica- tion: Comparing Approaches for Building Graphs from Images. InProceedings of the 26th International Conference on Enterprise In- formation Systems, volume 1, pages 437–446, Angers, France, 2024. SCITEPRESS - Science and Technology Publications

    2024

  32. [32]

    Scarselli, M

    F. Scarselli, M. Gori, Ah Chung Tsoi, M. Hagen- buchner, and G. Monfardini. The Graph Neural Network Model.IEEE Transactions on Neural Networks, 20(1):61–80, January 2009

    2009

  33. [33]

    Schuster and K.K

    M. Schuster and K.K. Paliwal. Bidirectional re- current neural networks.IEEE Transactions on Signal Processing, 45(11):2673–2681, November 1997

    1997

  34. [34]

    Daniel Selsam, Matthew Lamm, Benedikt B¨ unz, Percy Liang, Leonardo de Moura, and David L. Dill. Learning a SAT Solver from Single-Bit Su- pervision. In7th International Conference on Learning Representations, ICLR 2019, New Or- leans, LA, USA, May 6-9, 2019, New Orleans, LA, USA, May 2019. OpenReview.net

    2019

  35. [35]

    How to train your ViT? Data, Augmentation, and Regularization in Vi- sion Transformers.Transactions on Machine Learning Research, 2022

    Andreas Peter Steiner, Alexander Kolesnikov, Xiaohua Zhai, Ross Wightman, Jakob Uszkor- eit, and Lucas Beyer. How to train your ViT? Data, Augmentation, and Regularization in Vi- sion Transformers.Transactions on Machine Learning Research, 2022

    2022

  36. [36]

    Superpixels: An evaluation of the state- of-the-art.Computer Vision and Image Under- standing, 166:1–27, January 2018

    David Stutz, Alexander Hermans, and Bastian Leibe. Superpixels: An evaluation of the state- of-the-art.Computer Vision and Image Under- standing, 166:1–27, January 2018

    2018

  37. [37]

    Image Classifica- tion Based on Deep Graph Convolutional Net- works

    Tinglong Tang, Xiaowang Chen, Yirong Wu, Shuifa Sun, and Mei Yu. Image Classifica- tion Based on Deep Graph Convolutional Net- works. In2022 IEEE 9th International Confer- ence on Data Science and Advanced Analytics (DSAA), pages 1–6, Shenzhen, China, October

  38. [38]

    SEEDS: Superpixels Extracted via Energy- Driven Sampling

    Michael Van Den Bergh, Xavier Boix, Gemma Roig, Benjamin De Capitani, and Luc Van Gool. SEEDS: Superpixels Extracted via Energy- Driven Sampling. In David Hutchison, Takeo Kanade, Josef Kittler, Jon M. Kleinberg, Friede- mann Mattern, John C. Mitchell, Moni Naor, Oscar Nierstrasz, C. Pandu Rangan, Bernhard Steffen, Madhu Sudan, Demetri Terzopoulos, Doug T...

    2012

  39. [39]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Atten- tion is All you Need. In Isabelle Guyon, Ul- rike von Luxburg, Samy Bengio, Hanna M. Wal- lach, Rob Fergus, S. V. N. Vishwanathan, and Roman Garnett, editors,Advances in Neural In- formation Processing Systems 30: Annual Con...

    2017

  40. [40]

    Graph Attention Networks

    Petar Velickovic, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Li` o, and Yoshua Bengio. Graph Attention Networks. In6th International Conference on Learning Representations, ICLR 2018, Vancouver, BC, Canada, April 30 - May 3, 2018, Conference Track Proceedings, Vancouver, BC, Canada, May 2018. OpenReview.net

    2018

  41. [41]

    Pointer Networks

    Oriol Vinyals, Meire Fortunato, and Navdeep Jaitly. Pointer Networks. In Corinna Cortes, Neil D. Lawrence, Daniel D. Lee, Masashi Sugiyama, and Roman Garnett, editors,Ad- vances in Neural Information Processing Sys- tems 28: Annual Conference on Neural Informa- tion Processing Systems 2015, December 7-12, 2015, Montreal, Quebec, Canada, pages 2692– 2700, 2015

    2015

  42. [42]

    SMILES- BERT: Large Scale Unsupervised Pre-Training for Molecular Property Prediction

    Sheng Wang, Yuzhi Guo, Yuhong Wang, Hong- mao Sun, and Junzhou Huang. SMILES- BERT: Large Scale Unsupervised Pre-Training for Molecular Property Prediction. InProceed- ings of the 10th ACM International Conference on Bioinformatics, Computational Biology and Health Informatics, pages 429–436, Niagara Falls NY USA, September 2019. ACM. 17

    2019

  43. [43]

    Super- pixel Hierarchy.IEEE Transactions on Image Processing, 27(10):4838–4849, October 2018

    Xing Wei, Qingxiong Yang, Yihong Gong, Narendra Ahuja, and Ming-Hsuan Yang. Super- pixel Hierarchy.IEEE Transactions on Image Processing, 27(10):4838–4849, October 2018

    2018

  44. [44]

    Fang Wu, Dragomir Radev, and Stan Z. Li. Mol- former: Motif-Based Transformer on 3D Het- erogeneous Molecular Graphs.Proceedings of the AAAI Conference on Artificial Intelligence, 37(4):5312–5320, June 2023

    2023

  45. [45]

    Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms

    Han Xiao, Kashif Rasul, and Roland Vollgraf. Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms, September 2017. arXiv:1708.07747 [cs, stat]

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  46. [46]

    Girshick

    Tete Xiao, Mannat Singh, Eric Mintun, Trevor Darrell, Piotr Doll´ ar, and Ross B. Girshick. Early Convolutions Help Transformers See Bet- ter. In Marc’Aurelio Ranzato, Alina Beygelz- imer, Yann N. Dauphin, Percy Liang, and Jen- nifer Wortman Vaughan, editors,Advances in Neural Information Processing Systems 34: An- nual Conference on Neural Information Pr...

    2021

  47. [47]

    How Powerful are Graph Neural Networks? In7th International Conference on Learning Representations, ICLR 2019, New Or- leans, LA, USA, May 6-9, 2019, New Orleans, LA, USA, May 2019

    Keyulu Xu, Weihua Hu, Jure Leskovec, and Ste- fanie Jegelka. How Powerful are Graph Neural Networks? In7th International Conference on Learning Representations, ICLR 2019, New Or- leans, LA, USA, May 6-9, 2019, New Orleans, LA, USA, May 2019. OpenReview.net

    2019

  48. [48]

    Real-time coarse-to-fine topo- logically preserving segmentation

    Jian Yao, Marko Boben, Sanja Fidler, and Raquel Urtasun. Real-time coarse-to-fine topo- logically preserving segmentation. In2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 2947–2955, Boston, MA, USA, June 2015. IEEE. 18 Appendix A Algorithm A1Square Grid Superpixel Require:image∈R C×h×w, patch size∈Z Ensure:segments∈Z h×w En...

    2015