Pith. sign in

REVIEW 5 major objections 5 minor 22 references

Exploring Semantic Clustering and Similarity Search for Heterogeneous Traffic Scenario Graph

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A self-supervised graph embedding of heterogeneous traffic scenarios captures semantic similarity, allowing variable-length driving scenes to be condensed into fixed vectors for clustering and nearest-neighbor retrieval without manual…

desk verdict A useful but under-evidenced adaptation of BGRL to heterogeneous traffic scenario graphs; the evaluation needs baselines and a label-free hyperparameter choice before the semantic-embedding claim is credible. read the letter →

arxiv 2507.05086 v1 pith:D2G4NNXP submitted 2025-07-07 cs.LG

Ferdinand Mütsch , Maximilian Zipfl , Nikolai Polley , J. Marius Zöllner This is my paper
classification cs.LG
keywords trafficscenariographsheterogeneousgraphneuralnetworksself-supervisedlearningclusteringsimilaritysearchautonomousvehicletestingembeddings
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes representing a traffic scenario as a heterogeneous spatio-temporal graph with obstacle and road nodes plus typed edges for spatial relations and time-links, and learning one fixed-size embedding per scenario with a self-supervised graph neural network. The goal is to make distances in embedding space reflect semantic similarity, so that variable-length recordings condense into single vectors that can be clustered and searched by nearest neighbor. The authors argue this works without manual labels and without committing to one objective such as criticality, which matters because AV validation needs a finite but representative set of test scenarios covering the long tail of rare situations. On held-out scenarios, a simple classifier reading only the embeddings separates scenario labels usefully (an AUPRC of 0.757 in the home city, lower but still nonzero in unseen cities), and density-based clusters correspond to maneuvers like left turns and stopping at a traffic light.

What carries the argument

The load-bearing object is the heterogeneous spatio-temporal scenario graph: obstacle nodes carry state features such as position, velocity, acceleration, yaw, and traffic-light state; road-segment nodes carry centerline, width, and type; four edge families encode road topology, obstacle relations, obstacle-road relations, and temporal links between obstacle states. Coordinates are expressed relative to the median obstacle position and edge attributes carry pairwise relative geometry, giving translation and rotation invariance. Time enters through temporal unrolling with a configurable reach (default four steps) plus sinusoidal positional encoding. The argument rides on the encoder that pools per-edge-type graph convolutions into a fixed-size graph vector, and on the self-supervised bootstrapping loss that makes augmented views of the same graph land close together while no negative samples are needed.

What would settle it

Take a scenario and produce a transformed twin that preserves the same maneuver under translation, rotation, and small perturbations; if nearest-neighbor search over many unrelated scenarios does not rank the twin first among the learned embeddings, the claimed semantic similarity is not actually captured.

Watch

Extended reading notes

Core claim

The paper's central claim is that a graph-level, self-supervised learning objective alone can organize traffic scenarios by meaning. Variable-length scenarios are unrolled in time, encoded as heterogeneous graphs, and mapped to a 128-dimensional vector by a GNN with min/max/mean pooling; a bootstrapping objective, where an online predictor is trained to match a slowly updated target encoder on augmented views, shapes the space, and a contrastive variant is compared. The downstream evidence is that a multi-label classifier fed only these vectors reaches an AUPRC of 0.757 on held-out scenarios from the same city, and that density-based clusters visibly align with distinct driving maneuvers. The authors conclude that the embeddings capture scenario semantics well enough for clustering and similarity search, while acknowledging that cluster quality still leaves room for improvement before real-world use.

Load-bearing premise

The load-bearing premise is that the graph representation preserves all information that defines a scenario type, even though temporal order beyond four steps is only partially encoded, road segments are optional, and semantic rule context is not fully represented.

Editorial extensions

If this is right

  • Scenario corpora can be organized automatically into maneuver-like groups, allowing test engineers to pick representative scenarios per cluster without hand-curated catalogs.
  • Variable-length recordings reduce to fixed vectors, so nearest-neighbor retrieval and density-based clustering scale to large scenario databases.
  • The learned space transfers partially across cities, so clusters found in one operational domain can seed test selection for another.
  • Because training is label-free and not biased toward criticality, the same embeddings can later be searched or reweighted for rare long-tail cases.
  • The fixed-size embedding provides a common index key for scenario databases, enabling queries like 'find scenarios similar to this one' across recordings.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The label-based evaluation is only a rough proxy: a stricter test would check whether embeddings separate scenarios that differ only by traffic-rule context or lane topology, exactly the information the graph model encodes partially or optionally.
  • The large share of unclustered scenarios at practical settings (roughly seven in ten) suggests that a deployment should treat unassigned points as a 'normal driving' reservoir and sample from it deliberately rather than discarding it.
  • The same embedding space could serve as conditioning for generative scenario synthesis, an extension the authors mention as future work but do not demonstrate.
  • The 'at least one label matches' accuracy can overstate cluster purity; re-scoring with exact label-set agreement would give a stricter upper bound on semantic separation.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper proposes a heterogeneous spatio-temporal graph representation for traffic scenarios and two self-supervised graph-level embedding methods, one based on bootstrapping (BGRL) and one on contrastive learning (GraphCL). The embeddings are evaluated on nuPlan data through a downstream multi-label classifier and HDBSCAN clustering, with qualitative inspection of clusters. The authors report a multi-label classification AUPRC of 0.757 on Boston, a clustering accuracy of 0.442 with 74 clusters at min_cluster_size=25, and generalization to Pittsburgh and Singapore, and conclude that the self-supervised embeddings capture scenario semantics without manual labels.

Significance. If substantiated, the contribution would be useful for scenario-based testing: an unsupervised, scalable way to organize the scenario space into meaningful categories and retrieve similar scenarios. The paper combines a flexible graph model, a graph-level adaptation of BGRL, and an application-oriented evaluation on nuPlan, including cross-city generalization and qualitative cluster inspection; these are concrete strengths. However, the current evaluation does not establish that the learned embeddings add value over non-learned features or random embeddings, and the clustering evaluation is affected by label-guided hyperparameter selection and an uncalibrated metric. The central claim therefore needs additional evidence before it can be accepted.

major comments (5)
  1. [Sec. IV-C and Table II] The downstream classification results are reported without any baseline. To support the claim that the self-supervised embeddings capture semantics, the authors should compare against random embeddings, non-learned feature vectors constructed from the raw scenario graph (e.g., scenario length, obstacle counts, speed and acceleration statistics, road-edge histograms), and PCA or autoencoders on those features. Without such comparisons, the AUPRC values (0.757 on Boston, 0.563 on Pittsburgh, 0.488 on Singapore) cannot be interpreted, because coarse labels such as high_magnitude_speed may be predictable from simple statistics. In addition, report the chance-level AUPRC and per-label prevalence.
  2. [Sec. IV-D and Table III] The main clustering hyperparameter min_cluster_size was set to 25 because it yields roughly 69 clusters, explicitly 'guided by the number of 69 ground-truth classes,' and the same ground-truth labels are then used to compute clustering accuracy. This introduces label information into model selection and biases the reported accuracy of 0.442. Please either select min_cluster_size using a label-free criterion (e.g., stability or silhouette on a validation split), or report the full min_cluster_size sweep with accuracy and justify the chosen operating point without using the evaluation labels.
  3. [Sec. IV-D, Eq. (3)] The clustering accuracy metric counts a sample as correct if the cluster's most frequent label is contained in the sample's label set, with ten imbalanced multi-label classes. This 'at-least-one' rule can yield high accuracy from a small number of frequent labels, and no chance level is reported. Moreover, the metric is computed only on clustered points, which at min_cluster_size=25 is only 28.6% of the data (71.4% unclustered). Please report the expected accuracy under random cluster assignment, per-label support, and metrics on all points (e.g., treating unclustered points as errors or as a separate cluster), together with an analysis of why the unclustered points are 'normal driving' rather than a failure of the representation.
  4. [Sec. IV-B/C and Sec. III-B3] All quantitative results are from single runs, and the paper states that hyperparameters were 'determined empirically' with details omitted. Given the number of free hyperparameters (augmentation probability, EMA interval, hidden dimensions, embedding dimension, temporal reach, HDBSCAN parameters), single-run results do not establish reliability. Please provide multiple seeds with mean and standard deviation for Tables II and III, and report the hyperparameter values in the paper or a supplement so that the experiments are reproducible.
  5. [Sec. III-A] The paper does not independently validate that the graph representation preserves the semantics needed to distinguish scenario types. The default temporal reach of 4 and the optional road segments mean that temporal dependencies longer than four steps and lane-topology context can be absent; if so, no embedding method can separate those scenario types. Please add an ablation or a probe task (e.g., predicting scenario labels or attributes from the raw graph features, or varying temporal reach and road-segment inclusion) to demonstrate that the representation itself is sufficient for the semantic distinctions claimed.
minor comments (5)
  1. [Sec. III-B3 and Sec. V] There are typos: 'paramteres' should be 'parameters' in Sec. III-B3, and 'heterogeneouos' should be 'heterogeneous' in Sec. V.
  2. [Sec. IV-A] The definition of multi-label classification accuracy is unclear: the text says the predicted label set must be a superset of the ground-truth labels, but Table II labels the column 'Accuracy (contain)'. Please give the exact formula for this metric, as is done for Eq. (3).
  3. [Sec. IV-C] The reported 99.99% compliance with Eq. (2) does not state the number of test samples or whether this is computed over all augmented pairs; please clarify the exact evaluation protocol.
  4. [Sec. IV-B] The paper says the downstream classification is 'cross-validated' on Pittsburgh and Singapore, but these are independent holdout sets rather than cross-validation folds; please use more precise terminology.
  5. [Figure 5] A single random subset of seven clusters is shown; the qualitative claim would be stronger with a quantitative analysis of cluster-label agreement over all clusters, including the proportion of clusters with a clear dominant label and the distribution of cluster sizes.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial evaluation circularity: the clustering hyperparameter min_cluster_size is chosen using the ground-truth class count and the same ground-truth labels are then used to score cluster-label accuracy; the unsupervised embedding derivation itself is not circular.

  1. fitted input called prediction [Section IV-D (Clustering), hyperparameter search for HDBSCAN min_cluster_size]
    "Guided by the number of 69 ground-truth classes, we instead choose a setting that leads to an approximately similar number of clusters at mcs = 25 and obtain a multi-label accuracy of 0.44 and a silhouette score of 0.38."

    The HDBSCAN min_cluster_size is selected so that the number of discovered clusters approximates the number of ground-truth label sets (69). The same nuPlan labels are then used in Eq. (3) to compute the cluster-label accuracy, meaning the reported cluster-quality figure is conditioned on a hyperparameter chosen from the evaluation labels. This makes the observed cluster-label correspondence partly an artifact of the fitting choice rather than an independent confirmation that embeddings encode scenario semantics. The unsupervised embedding training itself does not use labels, so this circularity is confined to the clustering evaluation rather than the core representation-learning derivation.

full rationale

The core derivation—self-supervised BGRL/GraphCL training on heterogeneous traffic-scenario graphs without labels—is not circular. The graph representation and the contrastive/bootstrapping objectives are not defined in terms of the nuPlan labels, and the downstream classifier uses labels only as an external proxy; the paper itself says the labels serve as a rough proxy. No load-bearing uniqueness theorem or same-author justification is invoked; [4] is an external method and [15] is related/future work. The main circular element is the clustering evaluation: mcs=25 is chosen 'guided by the number of 69 ground-truth classes,' and the same ground-truth labels then define the multi-label accuracy in Eq. (3). This is a genuine but partial fitted-input-called-prediction effect, lowering the evidentiary value of the reported cluster accuracy. The absence of baselines or chance calibration for AUPRC and Eq. (3) is a correctness/completeness concern, not a circularity reduction, so it does not further raise the circularity score.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the completeness of the graph representation and on the self-supervised objective producing a similarity-preserving space. These are domain assumptions, not mathematical axioms, and they are not independently validated beyond the label-based proxy.

free parameters (6)
  • HDBSCAN min_cluster_size (mcs) = 25
    Set to produce a number of clusters comparable to the 69 ground-truth label sets; the same labels are later used to compute cluster accuracy, so this choice tunes the evaluation target.
  • temporal reach = 4
    Default number of hops for temporal edges to the past; set by hand as a configurable parameter (Section III-A).
  • augmentation probability p = 0.1 to 0.2
    Binomial probability for edge dropping, attribute dropping, and Gaussian attribute perturbation; adopted from BGRL (Section III-B1).
  • EMA update interval k = 10
    Target network updated every k-th step with reverse-decaying momentum; chosen to deviate from original BGRL (Section III-B1).
  • GNN hidden dimensions = 32/64/128 (obstacle), 64/128/256 (road)
    Chosen empirically; the authors state all hyperparameters were determined empirically and details are omitted (Section III-B3).
  • embedding dimension = 128
    Graph-level output dimension after pooling and linear layer; set by hand (Section III-B3).
assumptions (4)
  • domain assumption The heterogeneous graph model preserves the semantics of a traffic scenario.
    The representation uses relative coordinates, temporal edges with reach 4, and optional road segments; if it discards defining information (e.g., traffic rules, lane topology, time beyond 4 steps), embeddings cannot separate scenario types. Invoked throughout Section III-A.
  • domain assumption The bootstrapping objective (BGRL) learns a graph-level embedding space where L2 distance reflects semantic similarity.
    The paper extends BGRL from node-level to graph-level without theoretical justification; it observes the objective is satisfied (Equation 2) but does not independently verify that the distance metric matches human notions of scenario similarity. Section III-B1.
  • domain assumption The chosen augmentations preserve scenario identity.
    Edge/attribute dropping and Gaussian perturbation are assumed to create valid positive pairs; if augmentations alter the scenario's core semantics, the learned invariance will discard relevant information. Section III-B1.
  • domain assumption nuPlan ground-truth labels are a rough proxy for semantic scenario similarity.
    The paper explicitly notes labels are fine-grained, partially ambiguous, and ego-centric while scenarios are scene-centric; nonetheless, all quantitative evaluations rely on them. Section IV-A.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Exploring Semantic Clustering and Similarity Search for Heterogeneous Traffic Scenario Graph." pith.science (2026). https://pith.science/paper/D2G4NNXP

@misc{pith2026250705086,
  author       = {Pith},
  title        = {Pith review of: Exploring Semantic Clustering and Similarity Search for Heterogeneous Traffic Scenario Graph},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D2G4NNXP}},
  note         = {Machine review of arXiv:2507.05086}
}
read the original abstract

Scenario-based testing is an indispensable instrument for the comprehensive validation and verification of automated vehicles (AVs). However, finding a manageable and finite, yet representative subset of scenarios in a scalable, possibly unsupervised manner is notoriously challenging. Our work is meant to constitute a cornerstone to facilitate sample-efficient testing, while still capturing the diversity of relevant operational design domains (ODDs) and accounting for the "long tail" phenomenon in particular. To this end, we first propose an expressive and flexible heterogeneous, spatio-temporal graph model for representing traffic scenarios. Leveraging recent advances of graph neural networks (GNNs), we then propose a self-supervised method to learn a universal embedding space for scenario graphs that enables clustering and similarity search. In particular, we implement contrastive learning alongside a bootstrapping-based approach and evaluate their suitability for partitioning the scenario space. Experiments on the nuPlan dataset confirm the model's ability to capture semantics and thus group related scenarios in a meaningful way despite the absence of discrete class labels. Different scenario types materialize as distinct clusters. Our results demonstrate how variable-length traffic scenarios can be condensed into single vector representations that enable nearest-neighbor retrieval of representative candidates for distinct scenario categories. Notably, this is achieved without manual labeling or bias towards an explicit objective such as criticality. Ultimately, our approach can serve as a basis for scalable selection of scenarios to further enhance the efficiency and robustness of testing AVs in simulation.

Figures

Figures reproduced from arXiv: 2507.05086 by the authors.

Figure 1
Figure 1. Heterogeneous scenario graph schema Besides pure extraction, it is important to group scenarios into distinct classes and thus reduce the scenario space down to a manageable, finite, yet representative test set. Sampling from these clusters enables for more efficient testing, whereby “the more complete a set [...] is, the more convincingly testing can ensure correct system behavior” [3]. A prerequisite for effective… view at source ↗
Figure 2
Figure 2. Heterogeneous scenario graph data model T1 T2 Online Encoder Target Encoder Predictor diff EMA [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. BGRL architecture schema, inspired by [4] [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Heterogeneous graph encoder architecture [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: Hyperparameter search for min_cluster_size. Percentage of unassigned points increases for increasingly large minimum cluster size. scenarios exhibit striking features, the majority of them is rather a “blend” of multiple types. E. Qualitative results Given the ultimate…
Figure 7
Figure 7. Figure 7: Exemplary scenarios drawn from three different [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

22 extracted references · 19 canonical work pages

  1. [1]

    Did We Test All Scenarios for Automated and Autonomous Driving Systems?

    F. Hauer, T. Schmidt, B. Holzm ¨uller, and A. Pretschner, “Did We Test All Scenarios for Automated and Autonomous Driving Systems?” in IEEE ITSC, 2019

    work page 2019

  2. [2]

    Toward Unsupervised Test Scenario Extraction for Automated Driving Systems from Urban Natu- ralistic Road Traffic Data,

    N. Weber, C. Thiem, and U. Konigorski, “Toward Unsupervised Test Scenario Extraction for Automated Driving Systems from Urban Natu- ralistic Road Traffic Data,” SAE Intl. J CA V , no. 3, 2023

    work page 2023

  3. [3]

    Clustering Traffic Scenarios Using Mental Models as Little as Possible,

    F. Hauer, I. Gerostathopoulos, T. Schmidt, and A. Pretschner, “Clustering Traffic Scenarios Using Mental Models as Little as Possible,” in IEEE IV, 2020

    work page 2020

  4. [4]

    Large-Scale Representation Learning on Graphs via Bootstrapping,

    S. Thakoor, C. Tallec, M. G. Azar, M. Azabou, E. L. Dyer, R. Munos, P. Veliˇckovi´c, and M. Valko, “Large-Scale Representation Learning on Graphs via Bootstrapping,” 2023, arXiv:2102.06514 [cs]

    arXiv 2023

  5. [5]

    BEVFormer: Learning Bird’s-Eye-View Representation From LiDAR- Camera via Spatiotemporal Transformers,

    Z. Li, W. Wang, H. Li, E. Xie, C. Sima, T. Lu, Q. Yu, and J. Dai, “BEVFormer: Learning Bird’s-Eye-View Representation From LiDAR- Camera via Spatiotemporal Transformers,” IEEE PAMI, 2025

    work page 2025

  6. [6]

    DriveGAN: Towards a Controllable High-Quality Neural Simulation,

    S. W. Kim, J. Philion, A. Torralba, and S. Fidler, “DriveGAN: Towards a Controllable High-Quality Neural Simulation,” 2021

    work page 2021

  7. [7]

    Geometric Deep Learning for Autonomous Driving: Unlocking the Power of Graph Neural Networks With CommonRoad-Geometric,

    E. Meyer, M. Brenner, B. Zhang, M. Schickert, B. Musani, and M. Al- thoff, “Geometric Deep Learning for Autonomous Driving: Unlocking the Power of Graph Neural Networks With CommonRoad-Geometric,” in 2023 IEEE Intelligent V ehicles Symposium (IV) , 2023

    work page 2023

  8. [8]

    HDGT: Heterogeneous Driving Graph Transformer for Multi-Agent Trajectory Prediction via Scene Encoding,

    X. Jia, P. Wu, L. Chen, Y . Liu, H. Li, and J. Yan, “HDGT: Heterogeneous Driving Graph Transformer for Multi-Agent Trajectory Prediction via Scene Encoding,” IEEE PAMI, no. 11, 2023

    work page 2023

Show all 22 references
  1. [9]

    Heterogeneous Graph-based Trajectory Prediction using Local Map Context and Social Interactions,

    D. Grimm, M. Zipfl, F. Hertlein, A. Naumann, J. Luettin, S. Thoma, S. Schmid, L. Halilaj, A. Rettinger, and J. M. Z ¨ollner, “Heterogeneous Graph-based Trajectory Prediction using Local Map Context and Social Interactions,” in IEEE ITSC , 2023

    2023

  2. [10]

    L. Wu, P. Cui, J. Pei, and L. Zhao, Eds., Graph Neural Networks: F oundations, Frontiers, and Applications. Springer Nature, 2022

    2022

  3. [11]

    Towards Traffic Scene Description: The Semantic Scene Graph,

    M. Zipfl and J. M. Z ¨ollner, “Towards Traffic Scene Description: The Semantic Scene Graph,” in IEEE ITSC , 2022

    2022

  4. [12]

    roadscene2vec: A tool for extracting and embedding road scene-graphs,

    A. V . Malawade, S.-Y . Yu, B. Hsu, H. Kaeley, A. Karra, and M. A. Al Faruque, “roadscene2vec: A tool for extracting and embedding road scene-graphs,” Knowledge-Based Systems , 2022

    2022

  5. [13]

    VistaScenario: Interaction Scenario Engineering for Vehicles with In- telligent Systems for Transport Automation,

    C. Chang, J. Zhang, J. Ge, Z. Zhang, J. Wei, L. Li, and F.-Y . Wang, “VistaScenario: Interaction Scenario Engineering for Vehicles with In- telligent Systems for Transport Automation,” IEEE IV , 2024

    2024

  6. [14]

    An Unsupervised Random Forest Clustering Technique for Automatic Traffic Scenario Categorization,

    F. Kruber, J. Wurst, and M. Botsch, “An Unsupervised Random Forest Clustering Technique for Automatic Traffic Scenario Categorization,” in IEEE ITSC , 2018, iSSN: 2153-0017

    2018

  7. [15]

    Traffic Scene Similarity: a Graph-based Contrastive Learning Approach,

    M. Zipfl, M. Jarosch, and J. M. Z ¨ollner, “Traffic Scene Similarity: a Graph-based Contrastive Learning Approach,” in IEEE SSCI , 2023

    2023

  8. [16]

    Graph Matching Networks for Learning the Similarity of Graph Structured Objects,

    Y . Li, C. Gu, T. Dullien, O. Vinyals, and P. Kohli, “Graph Matching Networks for Learning the Similarity of Graph Structured Objects,” in Proceedings of the 36th ICML . PMLR, 2019

    2019

  9. [17]

    graph2vec: Learning Distributed Representations of Graphs,

    A. Narayanan, M. Chandramohan, R. Venkatesan, L. Chen, Y . Liu, and S. Jaiswal, “graph2vec: Learning Distributed Representations of Graphs,” 2017, arXiv:1707.05005 [cs]

    2017 arXiv

  10. [18]

    Permutation-Invariant Variational Autoencoder for Graph-Level Representation Learning,

    R. Winter, F. Noe, and D.-A. Clevert, “Permutation-Invariant Variational Autoencoder for Graph-Level Representation Learning,” in NeurIPS. Curran Associates, Inc., 2021

    2021

  11. [19]

    Graph Contrastive Learning with Augmentations,

    Y . You, T. Chen, Y . Sui, T. Chen, Z. Wang, and Y . Shen, “Graph Contrastive Learning with Augmentations,” in NeurIPS, 2020

    2020

  12. [20]

    Deep Graph Contrastive Representation Learning,

    Y . Zhu, Y . Xu, F. Yu, Q. Liu, S. Wu, and L. Wang, “Deep Graph Contrastive Representation Learning,” 2020, arXiv:2006.04131

    2020 arXiv

  13. [21]

    Generalizing Pooling Functions in Convolutional Neural Networks: Mixed, Gated, and Tree,

    C.-Y . Lee, P. W. Gallagher, and Z. Tu, “Generalizing Pooling Functions in Convolutional Neural Networks: Mixed, Gated, and Tree,” in Pro- ceedings of the 19th AISTATS . PMLR, 2016

    2016

  14. [22]

    ScenarioNet: Open-Source Platform for Large-Scale Traffic Scenario Simulation and Modeling,

    Q. Li, Z. M. Peng, L. Feng, Z. Liu, C. Duan, W. Mo, and B. Zhou, “ScenarioNet: Open-Source Platform for Large-Scale Traffic Scenario Simulation and Modeling,” NeurIPS, 2023

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.