Pith. sign in

REVIEW 3 major objections 6 minor 52 references

FAP-CD: Fairness-Driven Age-Friendly Community Planning via Conditional Diffusion Generation

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read A conditional graph diffusion model can generate age-friendly community plans that beat existing planning baselines by an average of 41% on efficiency and equity metrics.

desk verdict A legitimate new application of graph diffusion to age-friendly community planning, but the headline 41% gain is inflated because the model invents walking edges that the metrics then count as real accessibility. read the letter →

arxiv 2412.16699 v1 pith:GXG6IYXA submitted 2024-12-21 cs.AI

classification cs.AI
keywords age-friendlycommunitiesconditionaldiffusionmodelsgraphgenerationurbanplanningspatialequity15-minutecityfairnesswalkability
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

This paper claims that age-friendly community planning can be treated as a conditional graph-generation problem: given a city region's demographics, existing facilities, and street network, a diffusion model can produce new layouts of elderly-care stations, meal services, and other amenities that are both responsive to local demand and fair across neighborhoods. It introduces FAP-CD, a conditioned graph denoising diffusion model that learns the joint distribution of facility types and their spatial relationships, with a pre-training module that pushes region representations toward max-min fairness. The 15-minute walkability graph is used as a discrete condition during denoising, so generated facilities are placed within walking reach of residential areas. On Beijing data, the method reports an average 41% improvement over competitive baselines across efficiency, diversity, accessibility, and equity metrics. If the result holds in practice, planners would have a fast, evidence-based tool for urban renewal that targets aging populations.

What carries the argument

The load-bearing object is a conditioned graph denoising diffusion model over $G=(X,A)$, where $X$ holds one-hot facility category features and $A$ is an adjacency matrix with $A_{i,j}=1$ for pairs reachable on foot within 15 minutes. A fair-demand module produces the conditioning embedding $C$ by combining urban attributes, grid features, and a four-level demand state through attention and a max-min entropy objective, so the condition itself encodes both need and equity. The reverse denoising process $p_\theta(G_{t-1}|G_t,C,\bar A)$ is guided by the discrete walkability graph $\bar A$, and a residual hybrid network combines a graph transformer block for global information with a message-passing block for local neighborhood aggregation, augmented by $m$-step random walk statistics. This machinery carries the paper's argument because it is what turns noisy graphs into facility layouts that respect walkable access while pushing for fair regional coverage.

What would settle it

Take any FAP-CD output for a Beijing grid and attempt to site every proposed facility node on the actual parcel map: count how many fall on existing buildings, protected land, or privately held land with no acquisition pathway, and check whether the 'reconfigured roadways' such as subways and overpasses are credible within a real budget. If most placements cannot be realized, the 41% average improvement is a gain on abstract graph scores rather than a planning solution.

Watch

Extended reading notes

Core claim

The paper's central claim is that a graph diffusion model conditioned on fair, demand-aware region embeddings can generate age-friendly community planning schemes that outperform both traditional resource-allocation algorithms and deep generative baselines. Specifically, FAP-CD treats each region as a graph whose nodes are facility categories and whose edges encode whether two facilities are connected by a 15-minute walk; the model learns to denoise noisy graphs into optimized facility placements. A fair-demand module, trained with a max-min entropy objective and attention over urban attributes, grid features, and demand states, supplies the conditioning signal, while a discrete walkability graph from the road network guides sampling. The reported results put FAP-CD ahead of all baselines on average, with accessibility up 29% and the Gini coefficient, a measure of inequality, down 28% relative to the best baseline's scores.

Load-bearing premise

The load-bearing premise is that the generated graph—facility nodes connected by 15-minute walking paths—can actually be built into real urban space, because the model treats facility placement and road reconfiguration as free design choices without representing land ownership, zoning, budgets, or construction feasibility.

Editorial extensions

If this is right

  • Urban planners could generate candidate age-friendly layouts at a fine grid scale in minutes rather than through prolonged manual planning cycles.
  • Equity can be optimized during generation itself, meaning renewed communities would not concentrate elderly services in already well-served districts.
  • The conditional framework could be re-run as elderly demographics shift, producing updated layouts that track changing demand over time.
  • The same graph-diffusion formulation could be extended to other public-service planning tasks, such as schools, clinics, or parks, where fair walking access is the goal.
  • The discrete walkability graph makes the outputs legible as 15-minute-city plans, connecting the model directly to an established urban-design standard.

Reading between the lines

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

  • The paper's own results are scores computed on generated graphs; a natural next step the authors do not take is to test whether the proposed facility placements can actually be built on real parcels, given land ownership, zoning, and budget constraints.
  • The max-min fairness condition could be compared against other fairness definitions, such as Rawlsian lexicographic equity or per-capita equality, to see which one best matches what residents value.
  • The 41% improvement depends on the five chosen metrics; a different weighting that penalizes construction cost or disruption might change which planning method looks best, and that trade-off is left implicit.
  • The Beijing-specific facility categories and per-thousand-resident standards would need re-parameterization before the model transfers to cities with different service norms or data availability.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes FAP-CD, a conditional graph diffusion framework for generating age-friendly community facility layouts. The method models a city region as a facility graph, learns region-level fair-demand embeddings through an attention-based min-max pre-training module, and uses a hybrid graph transformer / message-passing denoising network conditioned on those embeddings and on a discrete 15-minute walking graph. The authors evaluate on real Beijing data (POIs, senior care facilities, residential areas, road networks) with five metrics—life-service efficiency, elderly-care efficiency, diversity, accessibility, and Gini—and report an average improvement of 41% over baselines such as DDPM, CondGEN, and DRF. A GitHub repository is provided.

Significance. If the evaluation is sound, the paper would be a useful first demonstration of conditional diffusion for age-friendly planning, combining real urban data with a generative model and addressing both service coverage and equity. Strengths include the use of real Beijing data, a public code repository, an ablation study, and the fact that the headline metrics are computed post-generation rather than optimized directly in the diffusion loss, which reduces circularity concerns. The significance is, however, contingent on two issues: the evaluation metrics must be precisely specified, and the generated graph must be shown to correspond to buildable or walkable urban configurations rather than to abstract graph scores.

major comments (3)
  1. [Appendix D, Eqs. (16)-(19) and Table 1] The evaluation metrics are under-specified, which undermines the reproducibility of the headline 41% improvement. In Eq. (17), the denominator N-1 cannot refer to the total number of grids N as defined in the text, since Diversity is summed over grids and would otherwise be a constant; it presumably should be the number of facility nodes or the number of facility types minus one. Similarly, N is used for nodes in Eq. (16), for grids in Eq. (19), and for an unspecified count in Eq. (17). The Average metric in Table 1 is presented as a raw mean of five metrics with different scales, and stating that the Gini metric is 'negated' does not specify normalization. Please define every symbol in every equation and describe how Average is computed; otherwise the reported 41% average improvement cannot be verified.
  2. [Methodology, Eq. (8) and Experiments, Figure 6] The planning-feasibility claim is not supported because the model generates the adjacency matrix A_t without constraining the final edges to the existing pedestrian road network. The reverse process in Eq. (8) denoises both X_t and A_t, and the conditioning on the discrete graph A-bar is not a hard constraint on which edges may appear in the output. The Accessibility and Efficiency metrics in Eqs. (16) and (18) then credit any generated edge as real walking access. Figure 6 explicitly attributes improvements to 'reconfigured roadways, including major urban renewal projects like subways and overpasses,' but the model has no representation of budget, land ownership, zoning, or construction feasibility. Please report the fraction of generated edges that are present in the walking graph A-bar, and also evaluate the metrics when the generated edges are restricted to existing walkable connections. Without such an analysis, the 41% improvement may be an artifact of invented edges rather than a planning solution.
  3. [Fair-demand Module, Eq. (6)] The max-min entropy loss in Eq. (6) is not well-defined as written. The expression p-bar_i = Pn_{i=1} p_ij uses the same symbol n for the batch size and the summation index, the condition p_ij = 0 if j in R_i is not consistent with the softmax output over facilities, and the objective L = 1 / min(-p-bar_i * log(p-bar_i)) + 1 is ambiguous regarding whether the minimum is over i and whether the +1 is in the denominator. Since the fair-demand embedding C_i is the main conditioning variable for the diffusion process, this objective must be stated precisely for the method to be reproducible.
minor comments (6)
  1. [Problem Statement] The demand representation D is written as D in {0,1,2,3}^{N x N}, but each grid presumably has a demand vector over facility types rather than an N x N matrix; please correct the dimensions.
  2. [Methodology, Eq. (5)] The projection matrix in Eq. (5) is written as W_H, while the text refers to an output projection matrix W_O; please make the notation consistent.
  3. [Appendix B] There is a typo in the implementation details: '3st-order DPM-Solver' should read '3rd-order DPM-Solver.'
  4. [Baselines, Section on ACA and GA] The descriptions of ACA and GA appear to be swapped: ACA is described as using crossover and mutation, while GA is described as using pheromone updates; please verify and correct these descriptions.
  5. [Experiments, Figure 4] The ablation results in Figure 4 are reported without error bars, even though Table 1 includes standard deviations for the main comparison; please add repeated-run variability to the ablation figure.
  6. [Appendix D, Eq. (19)] The Gini coefficient formula in Eq. (19) is written inline in a way that is hard to parse; please typeset it with explicit summation limits and define X_(i) as the sorted composite metric.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the headline 41% gain is an evaluated output, not a term in the objective, and no fitted parameter is renamed as a prediction.

full rationale

The derivation chain is self-contained against the circularity patterns checked. The score-matching objective in Eq. 9 trains the denoiser to predict added Gaussian noise conditioned on the fair-demand embedding C and the discrete walking graph A-bar; the reported metrics (Efficiency, Diversity, Accessibility, Gini, Average) are computed only after generation and do not appear in the diffusion loss or the fair-demand loss. The claimed 41% average improvement over baselines is therefore an evaluated outcome rather than a self-fulfilling optimization target. The demand-fairness condition C is derived from the same grid-demand categories that inform the problem setup, but conditioning on a data-derived representation is not equivalent to predicting the evaluation metric by construction; the model still must produce a spatial graph whose scores emerge from the generated nodes and edges. The paper does cite prior work by its own group (Yong and Zhou 2024; Xu and Zhou 2024), but only in a passing sentence about urban multimodal data; the citations are not load-bearing and no uniqueness claim or fitted ansatz is imported from them. The strongest limitation noted in the manuscript is that generated adjacency edges are scored as real walkable access even when the accompanying discussion invokes major urban renewal projects; that is a realism/feasibility validity concern, not a circularity in which the prediction reduces to its input.

Assumptions & free parameters 7 free parameters · 5 assumptions · 1 invented entities

The model's central claim rests on several domain assumptions about how to represent a city and measure fairness. The fairness loss and evaluation metrics are introduced without external validation, and the grid size and 15-minute threshold are hand-chosen. These are not free parameters in a physical derivation, but they shape the result.

free parameters (7)
  • grid_size_d = 2 km
    The city is partitioned into 2 km by 2 km grid regions; this discretization controls the scale at which facilities are generated and is chosen by the authors.
  • walking_time_threshold = 15 minutes
    Edges in the adjacency matrix are set to 1 when walking time between facilities is within 15 minutes. This is a policy-inspired threshold, not fitted.
  • diffusion_steps = 200
    Appendix B sets 200 reverse steps; this hyperparameter affects sample quality and is chosen by hand.
  • hidden_dim = 128
    Appendix B; hidden embedding dimension of the graph denoising network, selected via the hyperparameter experiments in Figure 5.
  • num_residual_layers = 4
    Appendix B; number of residual hybrid layers, selected via experiments shown in Figure 5.
  • random_walk_steps_m = 20
    Appendix B; number of random walk steps for node and edge augmentation, chosen by hand.
  • beta_schedule_extremes = beta_min=0.1, beta_max=20
    Appendix C; VPSDE noise schedule parameters, set without tuning.
assumptions (5)
  • domain assumption A city region can be represented as a graph of facility nodes with one-hot categories and a binary adjacency matrix based on 15-minute walking time.
    Problem statement and the graph formulation; the entire generation task is defined on this graph representation.
  • domain assumption Elderly demand can be reduced to four ordinal states (no supply, under supplied, appropriately supplied, oversupplied) computed from Beijing's per-thousand-resident facility standards.
    Problem Statement, Grid region demand; this discretization is an input to the model and is not externally validated.
  • ad hoc to paper Max-min entropy in Eq. (6) is an appropriate proxy for spatial equity in facility allocation.
    Fair-demand Module; the paper asserts this without a formal connection to the Gini or accessibility metrics used in evaluation.
  • standard math The learned denoising score network approximates the reverse SDE sufficiently well that samples correspond to realistic facility layouts.
    Methodology, Eqs. 8-9; relies on standard diffusion theory and the DPM-Solver, with no convergence certificate.
  • domain assumption The four reported metrics (Efficiency, Diversity, Accessibility, Gini) measure age-friendliness and fairness of a plan.
    Experiments, Evaluation Metrics; no external validation or user study connects these scores to resident outcomes.
invented entities (1)
  • Fair-demand embedding C_i
    purpose: A learned condition vector that combines region demand, urban attributes, and grid features to guide the graph denoising process.
    It is a model-internal latent constructed by attention and max-min pretraining; no falsifiable prediction outside the model is attached to it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FAP-CD: Fairness-Driven Age-Friendly Community Planning via Conditional Diffusion Generation." pith.science (2026). https://pith.science/paper/GXG6IYXA

@misc{pith2026241216699,
  author       = {Pith},
  title        = {Pith review of: FAP-CD: Fairness-Driven Age-Friendly Community Planning via Conditional Diffusion Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GXG6IYXA}},
  note         = {Machine review of arXiv:2412.16699}
}
read the original abstract

As global populations age rapidly, incorporating age-specific considerations into urban planning has become essential to addressing the urgent demand for age-friendly built environments and ensuring sustainable urban development. However, current practices often overlook these considerations, resulting in inadequate and unevenly distributed elderly services in cities. There is a pressing need for equitable and optimized urban renewal strategies to support effective age-friendly planning. To address this challenge, we propose a novel framework, Fairness-driven Age-friendly community Planning via Conditional Diffusion generation (FAP-CD). FAP-CD leverages a conditioned graph denoising diffusion probabilistic model to learn the joint probability distribution of aging facilities and their spatial relationships at a fine-grained regional level. Our framework generates optimized facility distributions by iteratively refining noisy graphs, conditioned on the needs of the elderly during the diffusion process. Key innovations include a demand-fairness pre-training module that integrates community demand features and facility characteristics using an attention mechanism and min-max optimization, ensuring equitable service distribution across regions. Additionally, a discrete graph structure captures walkable accessibility within regional road networks, guiding model sampling. To enhance information integration, we design a graph denoising network with an attribute augmentation module and a hybrid graph message aggregation module, combining local and global node and edge information. Empirical results across multiple metrics demonstrate the effectiveness of FAP-CD in balancing age-friendly needs with regional equity, achieving an average improvement of 41% over competitive baseline models.

Figures

Figures reproduced from arXiv: 2412.16699 by the authors.

Figure 1
Figure 1. Local Moran’s Index analysis of the distribution [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Framework overview of the proposed FAP-CD. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The architecture of (a) graph denoising network, (b) graph transfomer block, and (c) fair-demand module. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Performance Comparison in the Ablation Study. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Analysis of Hyperparameters. Ablation Study. We conduct an ablation study to validate the efficacy of the Residual Hybrid Layer. Results in Fig￾ure 4 indicate that incorporating the residual hybrid mod￾ule (w/o RHL) within the denoising model enhances AFCs generation p…
Figure 6
Figure 6. Figure 6: Comparison of Generated and Original AFCs. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Details of datasets in Beijing. 25 VGAE and GraphRNN, and a MLP decoder with inner prod￾uct computation for edge existence to reconstruct the graph structure. We provide a detailed description of the experimental setup in the baseline EDGE, which refers to the experi￾m…
Figure 8
Figure 8. Figure 8: Visual comparison pairs of generated AFCs. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 31 canonical work pages

  1. [1]

    Abas, A.; Arifin, K.; Ali, M. A. M.; and Khairil, M. 2023. A systematic literature review on public participation in decision-making for local authority planning: A decade of progress and challenges. Environmental Development, 46: 100853

  2. [2]

    Chen, X.; He, J.; Han, X.; and Liu, L.-P. 2023. Efficient and degree-guided graph generation via discrete diffusion modeling. arXiv preprint arXiv:2305.04111

  3. [3]

    Dhariwal, P.; and Nichol, A. 2021. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34: 8780--8794

  4. [4]

    P.; and Bresson, X

    Dwivedi, V. P.; and Bresson, X. 2020. A generalization of transformer networks to graphs. arXiv preprint arXiv:2012.09699

  5. [5]

    Eskandar, G.; Farag, Y.; Yenamandra, T.; Cremers, D.; Guirguis, K.; and Yang, B. 2023. Urban-StyleGAN: Learning to Generate and Manipulate Images of Urban Scenes. In 2023 IEEE Intelligent Vehicles Symposium (IV), 1--8. IEEE

  6. [6]

    Ghodsi, A.; Zaharia, M.; Hindman, B.; Konwinski, A.; Shenker, S.; and Stoica, I. 2011. Dominant resource fairness: Fair allocation of multiple resource types. In 8th USENIX symposium on networked systems design and implementation (NSDI 11)

  7. [7]

    Goodfellow, I.; Pouget-Abadie, J.; Mirza, M.; Xu, B.; Warde-Farley, D.; Ozair, S.; Courville, A.; and Bengio, Y. 2020. Generative adversarial networks. Communications of the ACM, 63(11): 139--144

  8. [8]

    K.; Martinkus, K.; Perraudin, N.; and Wattenhofer, R

    Haefeli, K. K.; Martinkus, K.; Perraudin, N.; and Wattenhofer, R. 2022. Diffusion models for graphs benefit from discrete state spaces. arXiv preprint arXiv:2210.01549

Show all 52 references
  1. [9]

    Han, S.; and Sung, Y. 2021. A max-min entropy framework for reinforcement learning. Advances in Neural Information Processing Systems, 34: 25732--25745

  2. [10]

    Ho, J.; Jain, A.; and Abbeel, P. 2020. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33: 6840--6851

  3. [11]

    Huang, H.; Sun, L.; Du, B.; and Lv, W. 2023. Conditional diffusion based on discrete graph structures for molecular graph generation. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, 4302--4311

  4. [12]

    Jang, S.; An, Y.; Yi, C.; and Lee, S. 2017. Assessing the spatial equity of Seoul’s public transportation using the Gini coefficient based on its accessibility. International Journal of Urban Sciences, 21(1): 91--107

  5. [13]

    Jensen, J. H. 2019. A graph-based genetic algorithm and generative model/Monte Carlo tree search for the exploration of chemical space. Chemical science, 10(12): 3567--3572

  6. [14]

    Jo, J.; Lee, S.; and Hwang, S. J. 2022. Score-based generative modeling of graphs via the system of stochastic differential equations. In International conference on machine learning, 10362--10383. PMLR

  7. [15]

    P.; and Welling, M

    Kingma, D. P.; and Welling, M. 2013. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114

  8. [16]

    N.; and Welling, M

    Kipf, T. N.; and Welling, M. 2016. Variational graph auto-encoders. arXiv preprint arXiv:1611.07308

  9. [17]

    S.; and Hashimoto, T

    Li, X.; Thickstun, J.; Gulrajani, I.; Liang, P. S.; and Hashimoto, T. B. 2022. Diffusion-lm improves controllable text generation. Advances in Neural Information Processing Systems, 35: 4328--4343

  10. [18]

    Liu, Y.-F.; Dai, Y.-H.; and Luo, Z.-Q. 2013. Max-min fairness linear transceiver design for a multi-user MIMO interference channel. IEEE Transactions on Signal Processing, 61(9): 2413--2423

  11. [19]

    Lu, C.; Zhou, Y.; Bao, F.; Chen, J.; Li, C.; and Zhu, J. 2022. Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps. Advances in Neural Information Processing Systems, 35: 5775--5787

  12. [20]

    H.; Means, R.; Keating, N.; Parkhurst, G.; and Eales, J

    Menec, V. H.; Means, R.; Keating, N.; Parkhurst, G.; and Eales, J. 2011. Conceptualizing age-friendly communities. Canadian Journal on Aging/La revue canadienne du vieillissement, 30(3): 479--493

  13. [21]

    15-Minute City

    Moreno, C.; Allam, Z.; Chabaud, D.; Gall, C.; and Pratlong, F. 2021. Introducing the “15-Minute City”: Sustainability, resilience and place identity in future post-pandemic cities. Smart cities, 4(1): 93--111

  14. [22]

    Q.; and Dhariwal, P

    Nichol, A. Q.; and Dhariwal, P. 2021. Improved denoising diffusion probabilistic models. In International conference on machine learning, 8162--8171. PMLR

  15. [23]

    Niu, C.; Song, Y.; Song, J.; Zhao, S.; Grover, A.; and Ermon, S. 2020. Permutation invariant graph generation via score-based generative modeling. In International Conference on Artificial Intelligence and Statistics, 4474--4484. PMLR

  16. [24]

    Organization, W. H. 2007. Global age-friendly cities: A guide. World Health Organization

  17. [25]

    M.; Migdalas, A.; and Pitsoulis, L

    Pardalos, P. M.; Migdalas, A.; and Pitsoulis, L. 2008. Pareto optimality, game theory and equilibria, volume 17. Springer Science & Business Media

  18. [26]

    Razaviyayn, M.; Hong, M.; and Luo, Z.-Q. 2013. Linear transceiver design for a MIMO interfering broadcast channel achieving max--min fairness. Signal Processing, 93(12): 3327--3340

  19. [27]

    Rudnicka, E.; Napiera a, P.; Podfigurna, A.; M e czekalski, B.; Smolarczyk, R.; and Grymowicz, M. 2020. The World Health Organization (WHO) approach to healthy ageing. Maturitas, 139: 6--11

  20. [28]

    Song, J.; Meng, C.; and Ermon, S. 2020. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502

  21. [29]

    Song, L.; Kong, X.; and Cheng, P. 2024. Supply-demand matching assessment of the public service facilities in 15-minute community life circle based on residents' behaviors. Cities, 144: 104637

  22. [30]

    P.; Kumar, A.; Ermon, S.; and Poole, B

    Song, Y.; Sohl-Dickstein, J.; Kingma, D. P.; Kumar, A.; Ermon, S.; and Poole, B. 2020. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456

  23. [31]

    Sun, W.; and Yang, W. 2023. The Application of Rehabilitation Garden Design in Elderly Care Communities in China Against of the Background of Aging. Journal of Innovation and Development, 3(3): 17--21

  24. [32]

    Talen, E.; and Anselin, L. 1998. Assessing spatial equity: an evaluation of measures of accessibility to public playgrounds. Environment and planning A, 30(4): 595--613

  25. [33]

    Vignac, C.; Krawczuk, I.; Siraudin, A.; Wang, B.; Cevher, V.; and Frossard, P. 2022. Digress: Discrete denoising diffusion for graph generation. arXiv preprint arXiv:2209.14734

  26. [34]

    Vincent, P. 2011. A connection between score matching and denoising autoencoders. Neural computation, 23(7): 1661--1674

  27. [35]

    Wai, H.-T.; Hong, M.; Yang, Z.; Wang, Z.; and Tang, K. 2019. Variance reduced policy evaluation with smooth function approximation. Advances in Neural Information Processing Systems, 32

  28. [36]

    Wang, D.; Fu, Y.; Wang, P.; Huang, B.; and Lu, C.-T. 2020. Reimagining city configuration: Automated urban planning via adversarial learning. In Proceedings of the 28th international conference on advances in geographic information systems, 497--506

  29. [37]

    Wang, D.; Wu, L.; Zhang, D.; Zhou, J.; Sun, L.; and Fu, Y. 2023. Human-instructed deep hierarchical generative learning for automated urban planning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, 4660--4667

  30. [38]

    Wang, Y.; Gonzales, E.; and Morrow-Howell, N. 2017. Applying WHO’s age-friendly communities framework to a national survey in China. Journal of Gerontological Social Work, 60(3): 215--231

  31. [39]

    Wen, H.; Lin, Y.; Xia, Y.; Wan, H.; Wen, Q.; Zimmermann, R.; and Liang, Y. 2023. Diffstg: Probabilistic spatio-temporal graph forecasting with denoising diffusion models. In Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems, 1--12

  32. [40]

    Xie, L. 2018. Age-friendly communities and life satisfaction among the elderly in urban China. Research on aging, 40(9): 883--905

  33. [41]

    Xu, Z.; and Zhou, X. 2024. CGAP: Urban Region Representation Learning with Coarsened Graph Attention Pooling. arXiv preprint arXiv:2407.02074

  34. [42]

    Yang, C.; Zhuang, P.; Shi, W.; Luu, A.; and Li, P. 2019. Conditional structure generation through graph variational generative adversarial nets. Advances in neural information processing systems, 32

  35. [43]

    Yang, Z.; Yu, B.; and Cheng, C. 2007. A parallel ant colony algorithm for bus network optimization. Computer-Aided Civil and Infrastructure Engineering, 22(1): 44--55

  36. [44]

    Ye, X.; Du, J.; and Ye, Y. 2022. MasterplanGAN: Facilitating the smart rendering of urban master plans via generative adversarial networks. Environment and Planning B: Urban Analytics and City Science, 49(3): 794--814

  37. [45]

    Yong, X.; and Zhou, X. 2024. MuseCL: Predicting Urban Socioeconomic Indicators via Multi-Semantic Contrastive Learning. arXiv preprint arXiv:2407.09523

  38. [46]

    You, J.; Ying, R.; Ren, X.; Hamilton, W.; and Leskovec, J. 2018. Graphrnn: Generating realistic graphs with deep auto-regressive models. In International conference on machine learning, 5708--5717. PMLR

  39. [47]

    H.; Conejos, S.; and Chan, E

    Yung, E. H.; Conejos, S.; and Chan, E. H. 2016. Social needs of the elderly and active aging in public open spaces in urban renewal. Cities, 52: 114--122

  40. [48]

    H.; Lemoine, B.; and Mitchell, M

    Zhang, B. H.; Lemoine, B.; and Mitchell, M. 2018. Mitigating unwanted biases with adversarial learning. In Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society, 335--340

  41. [49]

    Zheng, Y.; Lin, Y.; Zhao, L.; Wu, T.; Jin, D.; and Li, Y. 2023. Spatial planning of urban communities via deep reinforcement learning. Nature Computational Science, 3(9): 748--762

  42. [50]

    Zhou, Z.; Ding, J.; Liu, Y.; Jin, D.; and Li, Y. 2023. Towards generative modeling of urban flow through knowledge-enhanced denoising diffusion. In Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems, 1--12

  43. [51]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  44. [52]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.