pith. sign in

arxiv: 2606.22241 · v1 · pith:WYW57I2Unew · submitted 2026-06-20 · 💻 cs.DS

Efficient Algorithms for Influence Maximization in General Models and Observed Cascades

Pith reviewed 2026-06-26 10:37 UTC · model grok-4.3

classification 💻 cs.DS
keywords influence maximizationobserved cascadesindependent cascade modelsketchinglow-adaptivity optimizationblack-box sample accesssubmodular maximization
0
0 comments X

The pith

A low-adaptivity optimization framework combined with sketching yields the first nearly-linear time algorithm with provable guarantees for influence maximization on observed cascades.

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

The paper develops algorithms for influence maximization under general stochastic models that provide only black-box sample access, as well as under the observed-cascades model and the independent-cascade model. It presents a low-adaptivity optimization framework that reduces sample complexity and running time relative to earlier methods, together with a variance-guided adaptive variant. When this framework is paired with sketching, it produces an algorithm for observed cascades whose running time is nearly linear in the input size and that carries provable guarantees, optimal up to logarithmic factors. For the independent-cascade model the work supplies a new tail bound that replaces an n-factor with a τ-factor in the sample complexity whenever the diffusion depth τ is small. Experiments on real and synthetic instances confirm that these theoretical improvements translate into substantial practical speed-ups while preserving solution quality.

Core claim

We introduce a low-adaptivity optimization framework for general stochastic models with black-box sample access that improves sample complexity and running time over prior work. Combining the framework with sketching produces the first algorithm with provable guarantees and nearly-linear running time for influence maximization on observed cascades, optimal up to logarithmic factors. For the independent-cascade model we prove a novel tail bound that replaces a factor of n with τ (the number of diffusion steps) in the sample complexity whenever τ is small.

What carries the argument

Low-adaptivity optimization framework that uses adaptive sampling guided by empirical variance, combined with sketching to obtain nearly-linear running time.

If this is right

  • The framework improves sample complexity and running time over Sadeh et al. (2020) for any stochastic model with black-box access.
  • Observed-cascade influence maximization now admits a nearly-linear-time algorithm with guarantees, optimal up to log factors.
  • In the independent-cascade model the sample complexity improves from an n-factor to a τ-factor whenever diffusion depth is small.
  • An empirical-variance adaptive variant avoids the pessimistic worst-case sample bounds of earlier methods.

Where Pith is reading between the lines

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

  • The same low-adaptivity plus sketching pattern may apply to other submodular maximization tasks that currently rely on black-box oracles.
  • If τ grows with network size, the IC-model improvement disappears and a different tail bound would be needed.
  • The variance-guided adaptivity could be combined with existing sketching libraries to obtain practical implementations that beat the theoretical nearly-linear bound on real data.
  • Extending the framework to models without efficient black-box sampling would require an alternative way to generate unbiased estimates.

Load-bearing premise

The stochastic models admit efficient black-box sample access and, in the independent-cascade case, the diffusion depth τ remains small.

What would settle it

A runtime measurement on a sequence of observed-cascade instances whose size doubles each time, checking whether wall-clock time grows linearly rather than quadratically or worse.

Figures

Figures reproduced from arXiv: 2606.22241 by Alina Ene, Fabian Spaeh, Huy L. Nguyen, Themistoklis Haris.

Figure 1
Figure 1. Figure 1: Influence maximization for a general threshold model on Facebook for [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Influence maximization for τ = 5 steps on ℓ = 100 observed cascades from the Twitter network. 2500 5000 7500 budget k 50000 100000 150000 200000 250000 objective value Our Algorithm 2500 5000 7500 budget k 4.7 × 102 4.8 × 102 4.9 × 102 5 × 102 5.1 × 102 5.2 × 102 running time [s] [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Influence maximization for τ = n steps on ℓ = 100 observed cascades from the Pokec network. setting where this holds is when edge probabilities scale with degrees and τ is constant. We note that our IC algorithm is not designed to compete with highly optimized implementations of reverse influence sam￾pling (RIS) such as [Tang et al., 2015, Nguyen et al., 2016, Tang et al., 2018] but rather offers improved … view at source ↗
Figure 4
Figure 4. Figure 4: Influence maximization for a general threshold model on the Facebook network for [PITH_FULL_IMAGE:figures/full_fig_p049_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Influence maximization in the observed cascades model for [PITH_FULL_IMAGE:figures/full_fig_p049_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Influence maximization for 100 observed cascades with τ ∈ {n, 5, 10} on the Twitter network. 250 500 750 1000 5000 10000 15000 20000 objective value Our Algorithm 250 500 750 1000 budget k 2.3 × 102 2.35 × 102 2.4 × 102 2.45 × 102 2.5 × 102 running time [s] 250 500 750 1000 55000 60000 65000 70000 75000 80000 objective value Our Algorithm 250 500 750 1000 budget k 1.525 × 102 1.55 × 102 1.575 × 102 1.6 × 1… view at source ↗
Figure 7
Figure 7. Figure 7: Influence maximization for 100 observed cascades with τ ∈ {n, 5, 10} on the Google-Plus network. 50 [PITH_FULL_IMAGE:figures/full_fig_p050_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Influence maximization in the IC model on the Facebook and Twitter networks for [PITH_FULL_IMAGE:figures/full_fig_p051_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Influence maximization in the independent cascade model on the Facebook network for a budget of [PITH_FULL_IMAGE:figures/full_fig_p052_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Influence maximization for a general threshold model on the Facebook network with a budget of [PITH_FULL_IMAGE:figures/full_fig_p052_10.png] view at source ↗
read the original abstract

We study influence maximization in general stochastic models, the observed cascades model, and the independent cascade (IC) model. For general stochastic models with only black-box sample access, we introduce a low-adaptivity optimization framework that improves sample complexity and running time over Sadeh et al. (2020) and is instrumental to all our results. We further introduce an adaptive algorithm guided by empirical variance, avoiding pessimistic worst-case bounds. Combining our optimization framework with sketching, we obtain the first algorithm with provable guarantees and nearly-linear running time for influence maximization on observed cascades, optimal up to logarithmic factors. For IC, we prove a novel tail bound replacing a factor $n$ with $\tau$ (the number of diffusion steps) in sample complexity, improving over prior work when $\tau$ is small, as is common due to small-world phenomena. Experiments confirm substantial speedups while maintaining solution quality.

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

0 major / 3 minor

Summary. The paper studies influence maximization under general stochastic models with black-box sample access, the observed-cascades model, and the independent-cascade (IC) model. It introduces a low-adaptivity optimization framework that improves sample complexity and running time over Sadeh et al. (2020) and serves as the basis for the remaining results; an adaptive algorithm that uses empirical variance to avoid worst-case bounds; a sketching-based algorithm that yields the first provably correct nearly-linear-time solution for observed cascades (optimal up to logarithmic factors); and a new tail bound for the IC model that replaces an n-factor with τ (the number of diffusion steps). Experiments report substantial speed-ups while preserving solution quality.

Significance. If the stated guarantees and running-time claims hold, the work supplies the first nearly-linear-time algorithm with provable approximation guarantees for influence maximization on observed cascades, a setting directly relevant to large-scale network data. The low-adaptivity framework, the variance-guided adaptation, and the τ-based tail bound (exploiting the small-world regime common in practice) constitute concrete, usable improvements over the cited prior work. The combination of sketching with the optimization framework is a technically interesting contribution that could influence subsequent algorithmic design in submodular optimization under limited adaptivity.

minor comments (3)
  1. [Abstract / §1] The abstract and introduction should explicitly state the precise sample-complexity improvement (in big-O notation) achieved by the low-adaptivity framework relative to Sadeh et al. (2020) so that the claimed advance is immediately quantifiable.
  2. [IC-model tail bound] In the IC-model section, the new tail bound should be stated as a numbered theorem or lemma with the exact dependence on τ made visible, and the proof should be cross-referenced to the location where the n-to-τ replacement occurs.
  3. [Experiments] The experimental section would benefit from reporting the number of independent runs, standard deviations, and the precise network sizes used when claiming “substantial speedups,” to allow readers to assess reproducibility and effect size.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary, significance assessment, and recommendation of minor revision. The report accurately captures the main contributions of the low-adaptivity framework, variance-guided adaptation, sketching for observed cascades, and the τ-based tail bound.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper introduces a low-adaptivity optimization framework and tail-bound improvements for influence maximization, explicitly building on and improving over the cited Sadeh et al. (2020) work without any equations or claims reducing by construction to fitted parameters, self-definitions, or unverified self-citation chains. Assumptions such as black-box sample access and small τ are stated as standard modeling choices rather than hidden premises that force the results. No load-bearing step in the described derivation equates a prediction or uniqueness claim to its own inputs, leaving the central nearly-linear runtime result with independent algorithmic content.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; any modeling assumptions (black-box access, small τ) are implicit in the problem statements rather than introduced by the paper.

pith-pipeline@v0.9.1-grok · 5690 in / 1155 out tokens · 36969 ms · 2026-06-26T10:37:54.224957+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

46 extracted references · 1 canonical work pages

  1. [1]

    The adaptive complexity of maximizing a submodular function

    Eric Balkanski and Yaron Singer. The adaptive complexity of maximizing a submodular function. In STOC , pages 1138--1151. ACM , 2018

  2. [2]

    The limitations of optimization from samples

    Eric Balkanski, Aviad Rubinstein, and Yaron Singer. The limitations of optimization from samples. J. ACM , 69 0 (3): 0 21:1--21:33, 2022

  3. [3]

    Chayes, and Brendan Lucier

    Christian Borgs, Michael Brautbar, Jennifer T. Chayes, and Brendan Lucier. Maximizing social influence in nearly optimal time. In SODA , pages 946--957. SIAM , 2014

  4. [4]

    Exact and efficient generation of geometric random variates and random graphs

    Karl Bringmann and Tobias Friedrich. Exact and efficient generation of geometric random variates and random graphs. In ICALP (1) , Lecture Notes in Computer Science, pages 267--278. Springer, 2013

  5. [5]

    Scalable influence maximization for prevalent viral marketing in large-scale social networks

    Wei Chen, Chi Wang, and Yajun Wang. Scalable influence maximization for prevalent viral marketing in large-scale social networks. In KDD , pages 1029--1038. ACM , 2010

  6. [6]

    Time-critical influence maximization in social networks with time-delayed diffusion process

    Wei Chen, Wei Lu, and Ning Zhang. Time-critical influence maximization in social networks with time-delayed diffusion process. In AAAI , pages 591--598. AAAI Press, 2012

  7. [7]

    Wei Chen, Laks V. S. Lakshmanan, and Carlos Castillo. Information and Influence Propagation in Social Networks. Synthesis Lectures on Data Management. Morgan & Claypool Publishers, 2013

  8. [8]

    Network inference and influence maximization from samples

    Wei Chen, Xiaoming Sun, Jialin Zhang, and Zhijie Zhang. Network inference and influence maximization from samples. In ICML , Proceedings of Machine Learning Research, pages 1707--1716. PMLR , 2021 a

  9. [9]

    Best of both worlds: Practical and theoretically optimal submodular maximization in parallel

    Yixin Chen, Tonmoy Dey, and Alan Kuhnle. Best of both worlds: Practical and theoretically optimal submodular maximization in parallel. In NeurIPS, pages 25528--25539, 2021 b

  10. [10]

    Size-estimation framework with applications to transitive closure and reachability

    Edith Cohen. Size-estimation framework with applications to transitive closure and reachability. J. Comput. Syst. Sci., 55 0 (3): 0 441--453, 1997

  11. [11]

    Summarizing data using bottom-k sketches

    Edith Cohen and Haim Kaplan. Summarizing data using bottom-k sketches. In PODC , pages 225--234. ACM , 2007

  12. [12]

    Tighter estimation using bottom k sketches

    Edith Cohen and Haim Kaplan. Tighter estimation using bottom k sketches. Proc. VLDB Endow. , 1 0 (1): 0 213--224, 2008

  13. [13]

    Edith Cohen, Daniel Delling, Thomas Pajor, and Renato F. Werneck. Sketch-based influence maximization and computation: Scaling up with guarantees. In CIKM , pages 629--638. ACM , 2014 a

  14. [14]

    Edith Cohen, Daniel Delling, Thomas Pajor, and Renato F. Werneck. Timed influence: Computation and maximization. CoRR, abs/1410.6976, 2014 b

  15. [15]

    Scalable influence estimation in continuous-time diffusion networks

    Nan Du, Le Song, Manuel Gomez - Rodriguez, and Hongyuan Zha. Scalable influence estimation in continuous-time diffusion networks. In NIPS , pages 3147--3155, 2013

  16. [16]

    Uncovering the temporal dynamics of diffusion networks

    Manuel Gomez - Rodriguez, David Balduzzi, and Bernhard Sch \" o lkopf. Uncovering the temporal dynamics of diffusion networks. In ICML , pages 561--568. Omnipress, 2011

  17. [17]

    Threshold models of collective behavior

    Mark Granovetter. Threshold models of collective behavior. American journal of sociology, 83 0 (6): 0 1420--1443, 1978

  18. [18]

    Influence maximization revisited: Efficient sampling with bound tightened

    Qintian Guo, Sibo Wang, Zhewei Wei, Wenqing Lin, and Jing Tang. Influence maximization revisited: Efficient sampling with bound tightened. ACM Trans. Database Syst. , 47 0 (3): 0 12:1--12:45, 2022

  19. [19]

    Bevilacqua, Xiaokui Xiao, and Laks V

    Keke Huang, Sibo Wang, Glenn S. Bevilacqua, Xiaokui Xiao, and Laks V. S. Lakshmanan. Revisiting the stop-and-stare algorithms for influence maximization. Proc. VLDB Endow. , 10 0 (9): 0 913--924, 2017

  20. [20]

    Kleinberg, and \' E va Tardos

    David Kempe, Jon M. Kleinberg, and \' E va Tardos. Maximizing the spread of influence through a social network. Theory Comput., 11: 0 105--147, 2015

  21. [21]

    Influence maximization from cascade information traces in complex networks in the absence of network structure

    Naimisha Kolli and Balakrishnan Narayanaswamy. Influence maximization from cascade information traces in complex networks in the absence of network structure. IEEE Trans. Comput. Soc. Syst. , 6 0 (6): 0 1147--1155, 2019

  22. [22]

    SNAP Datasets : Stanford large network dataset collection

    Jure Leskovec and Andrej Krevl. SNAP Datasets : Stanford large network dataset collection. http://snap.stanford.edu/data, June 2014

  23. [23]

    VanBriesen, and Natalie S

    Jure Leskovec, Andreas Krause, Carlos Guestrin, Christos Faloutsos, Jeanne M. VanBriesen, and Natalie S. Glance. Cost-effective outbreak detection in networks. In KDD , pages 420--429. ACM , 2007 a

  24. [24]

    Glance, and Matthew Hurst

    Jure Leskovec, Mary McGlohon, Christos Faloutsos, Natalie S. Glance, and Matthew Hurst. Patterns of cascading behavior in large blog graphs. In SDM , pages 551--556. SIAM , 2007 b

  25. [25]

    A survey on influence maximization: From an ml-based combinatorial optimization

    Yandi Li, Haobo Gao, Yunxuan Gao, Jianxiong Guo, and Weili Wu. A survey on influence maximization: From an ml-based combinatorial optimization. ACM Trans. Knowl. Discov. Data , 17 0 (9): 0 133:1--133:50, 2023

  26. [26]

    Thai, Renhao Xue, James Song, Meikang Qiu, and Liang Zhao

    Chen Ling, Junji Jiang, Junxiang Wang, My T. Thai, Renhao Xue, James Song, Meikang Qiu, and Liang Zhao. Deep graph representation learning and optimization for influence maximization. In ICML , Proceedings of Machine Learning Research, pages 21350--21361. PMLR , 2023

  27. [27]

    Time constrained influence maximization in social networks

    Bo Liu, Gao Cong, Dong Xu, and Yifeng Zeng. Time constrained influence maximization in social networks. In ICDM , pages 439--448. IEEE Computer Society, 2012

  28. [28]

    Empirical bernstein bounds and sample-variance penalization

    Andreas Maurer and Massimiliano Pontil. Empirical bernstein bounds and sample-variance penalization. In COLT , 2009

  29. [29]

    Lazier than lazy greedy

    Baharan Mirzasoleiman, Ashwinkumar Badanidiyuru, Amin Karbasi, Jan Vondr \' a k, and Andreas Krause. Lazier than lazy greedy. In AAAI , pages 1812--1818. AAAI Press, 2015

  30. [30]

    On the submodularity of influence in social networks

    Elchanan Mossel and S \' e bastien Roch. On the submodularity of influence in social networks. In STOC , pages 128--134. ACM , 2007

  31. [31]

    Parkes, and Yaron Singer

    Harikrishna Narasimhan, David C. Parkes, and Yaron Singer. Learnability of influence in networks. In NIPS , pages 3186--3194, 2015

  32. [32]

    Nemhauser, Laurence A

    George L. Nemhauser, Laurence A. Wolsey, and Marshall L. Fisher. An analysis of approximations for maximizing submodular set functions - I . Math. Program., 14 0 (1): 0 265--294, 1978

  33. [33]

    Learning the graph of epidemic cascades

    Praneeth Netrapalli and Sujay Sanghavi. Learning the graph of epidemic cascades. In SIGMETRICS , pages 211--222. ACM , 2012

  34. [34]

    Nguyen, My T

    Hung T. Nguyen, My T. Thai, and Thang N. Dinh. Stop-and-stare: Optimal sampling algorithms for viral marketing in billion-scale networks. In SIGMOD Conference , pages 695--710. ACM , 2016

  35. [35]

    Nguyen, Thang N

    Hung T. Nguyen, Thang N. Dinh, and My T. Thai. Revisiting of 'revisiting the stop-and-stare algorithms for influence maximization'. In CSoNet, Lecture Notes in Computer Science, pages 273--285. Springer, 2018

  36. [36]

    Inferring graphs from cascades: A sparse recovery framework

    Jean Pouget - Abadie and Thibaut Horel. Inferring graphs from cascades: A sparse recovery framework. In ICML , JMLR Workshop and Conference Proceedings, pages 977--986. JMLR.org, 2015

  37. [37]

    Domingos

    Matthew Richardson and Pedro M. Domingos. Mining knowledge-sharing sites for viral marketing. In KDD , pages 61--70. ACM , 2002

  38. [38]

    Sample complexity bounds for influence maximization

    Gal Sadeh, Edith Cohen, and Haim Kaplan. Sample complexity bounds for influence maximization. In ITCS , LIPIcs, pages 29:1--29:36. Schloss Dagstuhl - Leibniz-Zentrum f \" u r Informatik, 2020

  39. [39]

    Online processing algorithms for influence maximization

    Jing Tang, Xueyan Tang, Xiaokui Xiao, and Junsong Yuan. Online processing algorithms for influence maximization. In SIGMOD Conference , pages 991--1005. ACM , 2018

  40. [40]

    Influence maximization: near-optimal time complexity meets practical efficiency

    Youze Tang, Xiaokui Xiao, and Yanchen Shi. Influence maximization: near-optimal time complexity meets practical efficiency. In SIGMOD Conference , pages 75--86. ACM , 2014

  41. [41]

    Influence maximization in near-linear time: A martingale approach

    Youze Tang, Yanchen Shi, and Xiaokui Xiao. Influence maximization in near-linear time: A martingale approach. In SIGMOD Conference , pages 1539--1554. ACM , 2015

  42. [42]

    Depth-first search and linear graph algorithms

    Robert Endre Tarjan. Depth-first search and linear graph algorithms. SIAM J. Comput. , 1 0 (2): 0 146--160, 1972

  43. [43]

    An experimental study of the small world problem, pages 130--148

    Jeffrey Travers and Stanley Milgram. An experimental study of the small world problem, pages 130--148. Princeton University Press, Princeton, 2006. ISBN 9781400841356. doi:doi:10.1515/9781400841356.130

  44. [44]

    A note on concentration of submodular functions

    Jan Vondr \' a k. A note on concentration of submodular functions. CoRR, abs/1005.2791, 2010

  45. [45]

    Modeling the spread of infectious diseases through influence maximization

    Shunyu Yao, Neng Fan, and Jie Hu. Modeling the spread of infectious diseases through influence maximization. Optim. Lett., 16 0 (5): 0 1563--1586, 2022

  46. [46]

    Information diffusion meets invitation mechanism

    Shiqi Zhang, Jiachen Sun, Wenqing Lin, Xiaokui Xiao, Yiqian Huang, and Bo Tang. Information diffusion meets invitation mechanism. In WWW (Companion Volume) , pages 383--392. ACM , 2024