Pith. sign in

REVIEW 6 major objections 5 minor 51 references

Scalable Explanation of Inferences on Large Graphs

T0 review · 6 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper proposes GraphExp, a beam-search method that explains belief-propagation inferences by extracting small trees whose target-node marginal closely matches the full graph's marginal.

desk verdict A clean new problem formulation with a sensible beam search, but the empirical claims outrun the hand-set potentials and the missing user study. read the letter →

arxiv 1908.06482 v2 pith:B44APWVO submitted 2019-08-13 cs.SI cs.LG

Chao Chen , Yifei Liu , Xi Zhang , Sihong Xie This is my paper · ORCID
classification cs.SIcs.LG
keywords beliefpropagationMarkovrandomfieldsmodelexplanationgraphsparsificationbeamsearchsymmetricKLdivergencescalableinferencefrauddetection
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 tries to establish that belief-propagation inferences on large cyclic graphical models can be explained by small tree-shaped subgraphs: for any target node, one can extract a subgraph with at most a handful of nodes, run belief propagation on it, and get a marginal for that node close to the marginal computed on the full graph. The authors formulate this as minimizing the symmetric KL divergence between the full-graph marginal and the subgraph marginal under an acyclicity and size constraint. They prove the optimization is NP-hard and that its objective is neither monotonic nor submodular, ruling out standard greedy guarantees, and they propose a beam-search algorithm, GraphExp, with two variants and a pruning strategy for scalability. Empirically, on ten networks with up to millions of nodes, the method reports lower symmetric KL divergence than LIME, embedding-based, and random baselines. If correct, this gives end-users a small, inspectable tree on which the inference's marginal is reproduced, supporting verification and debugging of model outputs.

What carries the argument

The load-bearing object is the constrained cross-entropy objective of Eq. (4): symmetric KL divergence between the marginal of a target node on the full cyclic MRF and the same marginal on a candidate acyclic subgraph. The search starts from the target node alone and greedily adds one neighbor at a time under an acyclicity constraint, evaluating each extension with a fresh belief-propagation run. Two instantiations are given: GE-G, which directly minimizes the global objective at each step, and GE-L, which back-traces messages and beliefs locally and stops when a prior explains a message. The beam keeps multiple candidate trees to provide diverse explanations, and the safe pruning strategy drops low-ranking candidates without, according to the experiments, degrading faithfulness. The non-submodularity theorem explains why the algorithm cannot rely on a $(1-1/e)$ approximation guarantee and instead must be evaluated empirically.

What would settle it

Run GraphExp and the baselines on the same ten networks with MRF parameters estimated by maximum likelihood or cross-validation rather than hand-set, and compare the symmetric KL values; if the advantage of GraphExp over the baselines narrows or reverses, the claimed superiority depends on the synthetic parameterization.

Watch

Extended reading notes

Core claim

The central claim is that explaining a belief-propagation marginal is equivalent to finding a small acyclic subgraph $\tilde{G}$ of the original MRF $G$ that minimizes $d(b_X,\tilde{b}_X)=KL(b_X\|\tilde{b}_X)+KL(\tilde{b}_X\|b_X)$, where $b_X$ is the target marginal on $G$ and $\tilde{b}_X$ the marginal computed on $\tilde{G}$. For any target $X$, a solution is a tree grown from $X$ one node at a time, and the paper proves that every output of its beam search is a tree. The paper also proves that exhaustive search for the optimal subgraph is NP-hard and that the objective is neither monotonic nor submodular, so no greedy algorithm carries a constant-factor guarantee. GraphExp therefore uses beam search to keep several candidate trees, evaluates each extension by rerunning belief propagation on the enlarged tree, and prunes the search space when the graph is dense. The reported result is that on ten networks from four domains the combined explanation reliably produces smaller symmetric KL divergence than the compared alternatives.

Load-bearing premise

The experiments assume that the synthetic model parameters chosen for each network—homophily potentials with 0.9 on the diagonal and 0.9 priors for labeled nodes—reflect the true dependency structure; if they do not, the reported faithfulness rankings may not transfer to deployed systems.

Editorial extensions

If this is right

  • On any MRF where belief propagation converges, a user can inspect a tree of at most $C$ nodes and see the message-passing computations that produce an approximation of the target marginal, without tracing the entire call graph.
  • Explanations are acyclic by construction, so the recursive self-reference that arises from cycles in loopy belief propagation is avoided.
  • Multiple diverse explanations from the beam let users choose the tree that matches their mental model, and the combined explanation (Comb) is the most faithful in the reported experiments.
  • The pruning strategy gives a practical speedup on dense graphs while keeping faithfulness roughly unchanged, supporting interactive use on networks with millions of edges.
  • The extracted trees can be used to audit a deployed detector, as demonstrated by the finding that some false negatives in Yelp spam detection are driven by reviewer prolificacy.

Reading between the lines

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

  • A natural extension the authors do not pursue is to apply the same extraction objective to other inference algorithms, such as loopy belief propagation with different message schedules or sampling-based marginals, treating the extracted tree as a surrogate that must match the algorithm's output rather than the model's true posterior.
  • The symmetric KL objective could be used as a training signal: one could search over trees not only to explain a fixed model but to choose model parameters that make the faithful explanation smaller, connecting explanation length to model simplicity.
  • Because the faithfulness measure is only evaluated under hand-set potentials, a testable next step is to compare extracted trees when the MRF parameters are learned from data; the authors' claims would be strengthened if the same advantage persists under calibrated potentials.
  • The Yelp security finding suggests the explanations can double as a diagnostic for biased detectors; one could formalize this by measuring whether explanation trees for false negatives systematically overweight high-degree neighbors across multiple datasets.
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

6 major / 5 minor

Summary. This paper addresses interpretability of belief propagation (BP) inferences on Markov random fields (MRFs). The authors formulate the task as finding a small acyclic subgraph that, when used for BP, yields a marginal for a target node close to the full-graph marginal, measured by symmetric KL divergence. They claim that the optimization is NP-hard and that the objective is neither submodular nor monotone, and they propose GraphExp, a beam-search algorithm with two variants (GE-G and GE-L), parallelization, and a pruning strategy. Experiments on 10 networks compare the method against random, DeepWalk-based embedding, and LIME baselines, and a visualization case study is discussed.

Significance. If the claims are fully supported, the paper offers a principled and scalable way to generate diverse, simple explanations for a common inference algorithm, filling an important gap: prior work on graphical-model explanation did not handle cycles or target the inference process itself. The constrained cross-entropy formulation is clean, the local/global search trade-off is a reasonable design contribution, and the scalability analysis is concrete. However, the NP-hardness result is only cited rather than proved, the non-submodularity counterexample is not numerically verified, and the empirical evidence rests on hand-set potentials and an evaluation metric identical to the optimized objective. The significance of the work is therefore potentially high but not yet established.

major comments (6)
  1. [Section IV-B (Experimental Setting) and Table IV] The MRF parameters for all seven non-Yelp networks are hand-set: compatibility matrices have 0.9 on the diagonal and 0.1/(c-1) elsewhere, and labeled-node priors are 0.9 on the true class. These strongly homophilous potentials make it easy for a small tree containing the target and a few labeled neighbors to reproduce the marginal, so the reported low symmetric KL values and the ranking over baselines may be specific to this parameter choice. The paper's central claim of superior performance on real applications therefore needs support from calibrated models or real posterior distributions.
  2. [Section IV-D and Table IV] The evaluation uses Eq. (4), which is exactly the objective minimized by GE-G. While baselines are compared fairly on the same metric, this measures how well the method achieves its own training target, not an independent notion of explanation quality. More importantly, the comparison is confounded by subgraph size: Comb builds subgraphs of average size 6.1–7.9 while GE-G(k=3) has average size at most 5.0, and the authors note in Section IV-D that faithfulness is positively correlated with size. No error bars or test statistics are reported that would support the claim of statistical significance.
  3. [Section III and Abstract] The claim that the optimization in Eq. (4) is NP-hard is not proved; the text states it is 'similar to exhaustive feature selection' and cites [11] and [1]. No reduction from a known NP-hard problem (e.g., Steiner tree or k-median on graphs) to the constrained, acyclic subgraph problem is given. The abstract's 'We prove' is therefore misleading.
  4. [Theorem II.1] The counterexample is asserted but not verified. The proof says 'One can run BP on G_i ... to find \tilde{b}_X' and states an inequality, but it reports neither the beliefs on X nor the resulting symmetric KL distances. As written, the reader cannot check the claimed violation of submodularity and monotonicity; the proof needs explicit numbers or a small script.
  5. [Section IV and Section IV-G] The manuscript announces a user study ('we conduct user study to demonstrate the usability of GE-G') but no user-study results appear. The utility example on Yelp spam detection also lacks quantitative evidence: the logistic regression model is described but its coefficients, performance, and statistical significance are not provided. Claims about usability and utility are thus unsupported.
  6. [Section III-B and Fig. 4] The 'safe pruning' claim that pruning does not hurt faithfulness is supported only by three curves on Yelp networks, without error bars, and no formal guarantee is given. The abstract's unconditional phrasing 'without hurting faithfulness' exceeds what the experiments establish.
minor comments (5)
  1. [Throughout] There are several typos: Section IV has 'demostrate' for 'demonstrate', Section IV-G has 'reivewer' for 'reviewer', Fig. 3 caption has 'exlaining' for 'explaining', and Table II has 'Random varaibles' for 'Random variables'.
  2. [Section IV-B] 'Compatibility metrics' should be 'compatibility matrices'.
  3. [Algorithm 1] The pseudocode does not specify how the 'top k extensions' are ranked; this is only described in the text. Please add a pointer or make the pseudocode self-contained.
  4. [Fig. 7 caption] The caption appears garbled ('C e n t e r e d: t a r g e t n o d e'); please fix the formatting.
  5. [Theorem III.1] The proof of Theorem III.1 is trivial; consider moving it to a remark.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the faithfulness objective is both minimized and evaluated, but baselines are compared on the same metric and an external utility check is provided.

full rationale

The paper's derivation chain is self-contained. The central optimization in Eq. (4) defines faithfulness as symmetric KL divergence between the full-graph and subgraph marginals, and the GE-G algorithm explicitly minimizes that same objective during beam search. Table IV then reports that same objective as the evaluation metric. This is a consistent use of a single definition of faithfulness, not a derivation that assumes its conclusion: the random, embedding, LIME, and GE-L baselines are all scored on the same metric without being allowed to optimize it, so the relative comparison is meaningful. No parameters are fitted to data and then renamed as predictions; MRF potentials are either taken from an external source [34] or set by hand in Section IV-B. No load-bearing self-citation appears: the cited Yelp parameters [34], datasets [50], and DeepWalk [33] are external. Theorem III.1, stating that Algorithm 1 outputs a tree, follows immediately from the algorithm's explicit rule of adding one node through one edge without forming a loop; it is a trivial consequence of the construction rather than a circular proof of a substantive claim. The external utility analysis in Section IV-G, which uses the generated subgraphs to identify a security issue in Yelp spam detection, provides an independent check of explanation usefulness and further separates the method's value from its own objective. Hand-set homophily potentials and the absence of a calibrated-model reproduction are correctness and generalization concerns, not circularity.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central approach rests on a small number of modeling choices: the symmetric KL objective as faithfulness, the tree-and-size constraints as interpretability, and the BP equations as the inference to explain. The empirical section additionally relies on hand-assigned potentials and priors for most networks, which are the main threat to generalizing the reported comparisons.

free parameters (5)
  • max subgraph size C = 5
    Set to 5 based on Miller's 7 plus or minus 2 working memory cap; controls explanation complexity and faithfulness.
  • beam size k = 1 or 3
    Beam width in the search; larger beam improves faithfulness at higher compute cost.
  • compatibility matrix diagonal = 0.9
    Ad hoc homophily setting for non-Yelp networks: 0.9 on diagonal, 0.1/(c-1) off-diagonal; not derived from data.
  • prior for labeled nodes = 0.9 for true class
    Hand-set priors for labeled nodes on non-Yelp networks; no calibration.
  • pruning ratio p = up to 99%
    Heuristic pruning threshold in GE-G; claimed safe empirically on Yelp networks only.
assumptions (4)
  • standard math BP converges to a unique fixed point on acyclic graphs, and Eq. (2)-(3) define the inference to be explained.
    Standard result cited to Pearl [32]; the paper's explanation target is the BP marginal, not the true posterior.
  • domain assumption Symmetric KL divergence d(bX, tilde bX) is the correct faithfulness measure for explanations.
    Eq. (4) defines the objective; the paper justifies via information loss but does not validate against human-perceived explanation quality.
  • domain assumption A tree subgraph of at most C nodes is interpretable to humans.
    Cites Miller's 7 plus or minus 2 and working memory; not empirically validated in this paper.
  • ad hoc to paper Synthetic compatibility potentials and priors on non-Yelp networks represent realistic MRFs.
    Section IV-B: assigns 0.9 diagonal and uniform-ish off-diagonal potentials; these are not learned from data, so the faithfulness numbers are internal to the paper's setup.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scalable Explanation of Inferences on Large Graphs." pith.science (2026). https://pith.science/paper/B44APWVO

@misc{pith2026190806482,
  author       = {Pith},
  title        = {Pith review of: Scalable Explanation of Inferences on Large Graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B44APWVO}},
  note         = {Machine review of arXiv:1908.06482}
}
read the original abstract

Probabilistic inferences distill knowledge from graphs to aid human make important decisions. Due to the inherent uncertainty in the model and the complexity of the knowledge, it is desirable to help the end-users understand the inference outcomes. Different from deep or high-dimensional parametric models, the lack of interpretability in graphical models is due to the cyclic and long-range dependencies and the byzantine inference procedures. Prior works did not tackle cycles and make \textit{the} inferences interpretable. To close the gap, we formulate the problem of explaining probabilistic inferences as a constrained cross-entropy minimization problem to find simple subgraphs that faithfully approximate the inferences to be explained. We prove that the optimization is NP-hard, while the objective is not monotonic and submodular to guarantee efficient greedy approximation. We propose a general beam search algorithm to find simple trees to enhance the interpretability and diversity in the explanations, with parallelization and a pruning strategy to allow efficient search on large and dense graphs without hurting faithfulness. We demonstrate superior performance on 10 networks from 4 distinct applications, comparing favorably to other explanation methods. Regarding the usability of the explanation, we visualize the explanation in an interface that allows the end-users to explore the diverse search results and find more personalized and sensible explanations.

Figures

Figures reproduced from arXiv: 1908.06482 by the authors.

Figure 1
Figure 1. A cyclic graphical model G for the Zachary karate club network, with BP inference outcomes shown in two colors. We focus on explaining how BP calculates the belief on X, highlighted in the subgraph G˜(1). Due to the cycles and long-range dependencies on G, a complete explanation is recursive and long. With messages, beliefs, and priors, GraphExp extracts a limited-size tree G˜(3), on which X has a marginal similar t… view at source ↗
Figure 2
Figure 2. A graphical model on which submodularity and monotonicity of the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Explaining BP on a graphical model with four variables ( [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Left the computing time reduces superlinearly as the number of cores increases from 1 to 14. The parentheses enclose the number of hour(s) per unit. The square brackets enclose the running time of GE-L. For example, GE-G takes about 192 hours using one core on YelpZip,…
Figure 5
Figure 5. Figure 5: Explanation faithfulness (the smaller the better) on citation networks while the ratio of labeled nodes increases. Random (L) is the random baseline that imitate GE-L. Similar for Random (G). 2 (2) 3 (2) 4 (2) 5 (2) Subgraph size (YelpZip) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0…
Figure 6
Figure 6. Figure 6: Relationship between size of subgraph and faithfulness Centered:targetnode Red:constructedbefore Green:latestadded Number:probability [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Larger red nodes are the nodes explained. Other red nodes are the explaining nodes. The green nodes are the newly added ones. the above mistakes should not depend on the prolificacy of the connected reviewers and products. However, we found that a sizable subset of fal…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 42 canonical work pages

  1. [11]

    An introduction to variable and feature selection

    Isabelle Guyon and Andr ´e Elisseeff. An introduction to variable and feature selection. Journal of machine learning research , 3(Mar):1157–1182, 2003

    work page 2003

  2. [1]

    On the approximability of minimizing nonzero variables or unsatisfied relations in linear systems

    Edoardo Amaldi and Viggo Kann. On the approximability of minimizing nonzero variables or unsatisfied relations in linear systems. Theoretical Computer Science , 209(1-2):237–260, 1998

    work page 1998

  3. [2]

    How to explain individual classification decisions

    David Baehrens, Timon Schroeter, Stefan Harmeling, Motoaki Kawanabe, Katja Hansen, and Klaus-Robert M ˜Aˇzller. How to explain individual classification decisions. Journal of Machine Learning Research, 11(Jun):1803–1831, 2010

    work page 2010

  4. [3]

    Diverse m-best solutions in markov random fields

    Dhruv Batra, Payman Yadollahpour, Abner Guzman-Rivera, and Gregory Shakhnarovich. Diverse m-best solutions in markov random fields. In European Conference on Computer Vision , pages 1–16. Springer, 2012

    work page 2012

  5. [4]

    Case-based explanation of non-case-based learning methods

    Rich Caruana, Hooshang Kangarloo, JD Dionisio, Usha Sinha, and David Johnson. Case-based explanation of non-case-based learning methods. In Proceedings of the AMIA Symposium, page

    work page

  6. [5]

    Sensitivity analysis in markov networks

    Hei Chan and Adnan Darwiche. Sensitivity analysis in markov networks. In International Joint Conference on Artificial In- telligence, volume 19, page 1300. LAWRENCE ERLBAUM ASSOCIATES LTD, 2005

    work page 2005

  7. [6]

    Reading tea leaves: How humans interpret topic models

    Jonathan Chang, Sean Gerrish, Chong Wang, Jordan L Boyd- Graber, and David M Blei. Reading tea leaves: How humans interpret topic models. In Advances in neural information processing systems, pages 288–296, 2009

    work page 2009

  8. [7]

    A differential approach to inference in bayesian networks

    Adnan Darwiche. A differential approach to inference in bayesian networks. Journal of the ACM (JACM) , 50(3):280– 305, 2003

    work page 2003

Show all 51 references
  1. [8]

    Sparse inverse covariance estimation with the graphical lasso

    Jerome Friedman, Trevor Hastie, and Robert Tibshirani. Sparse inverse covariance estimation with the graphical lasso. Bio- statistics, 9(3):432–441, 2008

    2008

  2. [9]

    Residual splash for optimally parallelizing belief propagation

    Joseph Gonzalez, Yucheng Low, and Carlos Guestrin. Residual splash for optimally parallelizing belief propagation. In Artifi- cial Intelligence and Statistics , pages 177–184, 2009

    2009

  3. [10]

    European union regulations on algorithmic decision-making and a right to explanation

    Bryce Goodman and Seth Flaxman. European union regulations on algorithmic decision-making and a right to explanation. AI Magazine, 38(3):50–57, 2017

    2017

  4. [12]

    Norm-product belief propa- gation: Primal-dual message-passing for approximate inference

    Tamir Hazan and Amnon Shashua. Norm-product belief propa- gation: Primal-dual message-passing for approximate inference. IEEE Transactions on Information Theory , 56(12):6294–6316, 2010

    2010

  5. [13]

    Variable selection for gaussian graphical models

    Jean Honorio, Dimitris Samaras, Irina Rish, and Guillermo Cecchi. Variable selection for gaussian graphical models. In Artificial Intelligence and Statistics , pages 538–546, 2012

    2012

  6. [14]

    An efficient network querying method based on conditional random fields

    Qiang Huang, Ling-Yun Wu, and Xiang-Sun Zhang. An efficient network querying method based on conditional random fields. Bioinformatics, 27(22):3173–3178, 2011

    2011

  7. [15]

    Fast and scalable distributed loopy belief propagation on real-world graphs

    Saehan Jo, Jaemin Yoo, and U Kang. Fast and scalable distributed loopy belief propagation on real-world graphs. In Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining , pages 297–305. ACM, 2018

    2018

  8. [16]

    Composing graphical models with neural networks for structured representations and fast in- ference

    Matthew Johnson, David K Duvenaud, Alex Wiltschko, Ryan P Adams, and Sandeep R Datta. Composing graphical models with neural networks for structured representations and fast in- ference. In Advances in neural information processing systems , pages 2946–2954, 2016

    2016

  9. [17]

    Inference of beliefs on billion-scale graphs

    U Kang, Duen Horng, et al. Inference of beliefs on billion-scale graphs. 2010

    2010

  10. [18]

    Markov random fields and their applications

    Ross Kindermann. Markov random fields and their applications. American mathematical society , 1980

    1980

  11. [19]

    Understanding black-box predictions via influence functions

    Pang Wei Koh and Percy Liang. Understanding black-box predictions via influence functions. In Proceedings of the 34th International Conference on Machine Learning-Volume 70, pages 1885–1894. JMLR. org, 2017

    2017

  12. [20]

    Probabilistic graphical models: principles and techniques

    Daphne Koller and Nir Friedman. Probabilistic graphical models: principles and techniques . MIT press, 2009

    2009

  13. [21]

    Increasing the in- terpretability of recurrent neural networks using hidden markov models

    Viktoriya Krakovna and Finale Doshi-Velez. Increasing the in- terpretability of recurrent neural networks using hidden markov models. arXiv preprint arXiv:1606.05320 , 2016

    2016 arXiv

  14. [22]

    An evaluation of the human-interpretability of explanation

    Isaac Lage, Emily Chen, Jeffrey He, Menaka Narayanan, Been Kim, Sam Gershman, and Finale Doshi-Velez. An evaluation of the human-interpretability of explanation. arXiv preprint arXiv:1902.00006, 2019

    1902 arXiv

  15. [23]

    Rationalizing neural predictions

    Tao Lei, Regina Barzilay, and Tommi Jaakkola. Rationalizing neural predictions. arXiv preprint arXiv:1606.04155 , 2016

    2016 arXiv

  16. [24]

    Protein– protein interaction site prediction based on conditional random fields

    Ming-Hui Li, Lei Lin, Xiao-Long Wang, and Tao Liu. Protein– protein interaction site prediction based on conditional random fields. Bioinformatics, 23(5):597–604, 2007

    2007

  17. [25]

    The structure and function of explanations

    Tania Lombrozo. The structure and function of explanations. Trends in cognitive sciences , 10(10):464–470, 2006

    2006

  18. [26]

    Intelligible models for classification and regression

    Yin Lou, Rich Caruana, and Johannes Gehrke. Intelligible models for classification and regression. In Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining , pages 150–158. ACM, 2012

    2012

  19. [27]

    Reviews, reputation, and revenue: The case of yelp

    Michael Luca. Reviews, reputation, and revenue: The case of yelp. com. Com (March 15, 2016). Harvard Business School NOM Unit Working Paper , (12-016), 2016

    2016

  20. [28]

    The magical number seven, plus or minus two: Some limits on our capacity for processing information

    George A Miller. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological review, 63(2):81, 1956

    1956

  21. [29]

    Collective graph identification

    Galileo Mark Namata, Stanley Kok, and Lise Getoor. Collective graph identification. In Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining, pages 87–95. ACM, 2011

    2011

  22. [30]

    An analysis of approximations for maximizing submod- ular set functions

    George L Nemhauser, Laurence A Wolsey, and Marshall L Fisher. An analysis of approximations for maximizing submod- ular set functions. Mathematical programming, 14(1):265–294, 1978

    1978

  23. [31]

    An explanation mechanism for bayesian inferencing systems

    Steven W Norton. An explanation mechanism for bayesian inferencing systems. In Machine Intelligence and Pattern Recognition, volume 5, pages 165–173. Elsevier, 1988

    1988

  24. [32]

    Probabilistic reasoning in intelligent systems: networks of plausible inference

    Judea Pearl. Probabilistic reasoning in intelligent systems: networks of plausible inference . Elsevier, 2014

    2014

  25. [33]

    Deepwalk: Online learning of social representations

    Bryan Perozzi, Rami Al-Rfou, and Steven Skiena. Deepwalk: Online learning of social representations. In Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining , pages 701–710. ACM, 2014

  26. [34]

    Collective opinion spam detection: Bridging review networks and metadata

    Shebuti Rayana and Leman Akoglu. Collective opinion spam detection: Bridging review networks and metadata. In Pro- ceedings of the 21th acm sigkdd international conference on knowledge discovery and data mining , pages 985–994. ACM, 2015

    2015

  27. [35]

    Why should i trust you?: Explaining the predictions of any classifier

    Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. Why should i trust you?: Explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD international con- ference on knowledge discovery and data mining , pages 1135–

  28. [36]

    Right for the right reasons: Training differentiable models by constraining their explanations

    Andrew Slavin Ross, Michael C Hughes, and Finale Doshi- Velez. Right for the right reasons: Training differentiable models by constraining their explanations. arXiv preprint arXiv:1703.03717, 2017

    2017 arXiv

  29. [37]

    Efficient search for diverse coherent explanations

    Chris Russell. Efficient search for diverse coherent explanations. arXiv preprint arXiv:1901.04909 , 2019

    1901 arXiv

  30. [38]

    Constructing the call graph of a pro- gram

    Barbara G Ryder. Constructing the call graph of a pro- gram. IEEE Transactions on Software Engineering , (3):216– 226, 1979

    1979

  31. [39]

    Deep inside convolutional networks: Visualising image clas- sification models and saliency maps

    Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. Deep inside convolutional networks: Visualising image clas- sification models and saliency maps. arXiv preprint arXiv:1312.6034, 2013

    2013 arXiv

  32. [40]

    Integrating rich user feedback into intelligent user interfaces

    Simone Stumpf, Erin Sullivan, Erin Fitzhenry, Ian Oberst, Weng-Keen Wong, and Margaret Burnett. Integrating rich user feedback into intelligent user interfaces. In Proceedings of the 13th international conference on Intelligent user interfaces , pages 50–59. ACM, 2008

    2008

  33. [41]

    Explanation in bayesian belief networks

    Henri J Suermondt. Explanation in bayesian belief networks. 1993

    1993

  34. [42]

    An introduction to conditional random fields

    Charles Sutton, Andrew McCallum, et al. An introduction to conditional random fields. Foundations and Trends R© in Machine Learning, 4(4):267–373, 2012

    2012

  35. [43]

    Relational learning via latent social dimensions

    Lei Tang and Huan Liu. Relational learning via latent social dimensions. In Proceedings of the 15th ACM SIGKDD inter- national conference on Knowledge discovery and data mining , pages 817–826. ACM, 2009

    2009

  36. [44]

    Learning to infer social ties in large networks

    Wenbin Tang, Honglei Zhuang, and Jie Tang. Learning to infer social ties in large networks. In Joint european conference on machine learning and knowledge discovery in databases , pages 381–397. Springer, 2011

    2011

  37. [45]

    An analysis of physi- cian attitudes regarding computer-based clinical consultation systems

    Randy L Teach and Edward H Shortliffe. An analysis of physi- cian attitudes regarding computer-based clinical consultation systems. Computers and Biomedical Research , 14(6):542–558, 1981

    1981

  38. [46]

    Counter- factual explanations without opening the black box: Automated decisions and the gpdr

    Sandra Wachter, Brent Mittelstadt, and Chris Russell. Counter- factual explanations without opening the black box: Automated decisions and the gpdr. Harv. JL & Tech., 31:841, 2017

    2017

  39. [47]

    influence sketching: Finding influential samples in large-scale regressions

    Mike Wojnowicz, Ben Cruz, Xuan Zhao, Brian Wallace, Matt Wolff, Jay Luan, and Caleb Crable. influence sketching: Finding influential samples in large-scale regressions. In 2016 IEEE International Conference on Big Data (Big Data) , pages 3601–

    2016

  40. [48]

    Linear program- ming relaxations and belief propagation–an empirical study

    Chen Yanover, Talya Meltzer, and Yair Weiss. Linear program- ming relaxations and belief propagation–an empirical study. Journal of Machine Learning Research , 7(Sep):1887–1907, 2006

    1907

  41. [49]

    Smoke screener or straight shooter: Detecting elite sybil attacks in user-review social networks

    Haizhong Zheng, Minhui Xue, Hao Lu, Shuang Hao, Haojin Zhu, Xiaohui Liang, and Keith Ross. Smoke screener or straight shooter: Detecting elite sybil attacks in user-review social networks. arXiv preprint arXiv:1709.06916 , 2017

    2017 arXiv

  42. [50]

    Predicting multicellular function through multi-layer tissue networks

    Marinka Zitnik and Jure Leskovec. Predicting multicellular function through multi-layer tissue networks. Bioinformatics, 33(14):i190–i198, 2017

    2017

  43. [212]

    American Medical Informatics Association, 1999

    1999

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.