REVIEW 2 major objections 5 minor 93 references
A Compositional Theory of Curvature in Probabilistic Circuits
T0 review · 2 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash
Pith's one-line read Each sum node's curvature contribution in a probabilistic circuit splits exactly into squared flow times local sharpness, which explains why global sharpness regularization underfits and motivates a gated fix.
desk verdict Clean theory, useful decomposition, but the headline experiments are undercut by an implementation mismatch in Algorithm 2; still worth refereeing. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the exact global–local factorization of the Hessian trace, $T_n(x) = F_n(x)^2 t_n(x)$, obtained by substituting the edge-flow identity $F_{nc}(x)/\theta_{nc} = \rho_{nc}(x) F_n(x)$, with $\rho_{nc} = p_c/p_n$ the output ratio of child $c$ to its sum parent $n$, into the definition of the node's trace contribution. Two companion results carry the argument. Locally, the Hessian of a sum node's negative log-output with respect to its edge weights is the rank-one outer product $\rho_n \rho_n^\top$, whose unique nonzero eigenvalue equals $t_n = \|\rho_n\|_2^2$, so $t_n$ simultaneously is the local trace, the maximum curvature, and the total Hessian magnitude. Globally, the circuit-flow recursion is unrolled to show that in a tree, flow through a node is the product of routing responsibilities over upstream sum edges only, which yields the geometric attenuation bound $T_n(x) \le \rho^{2 d_\Sigma(n)} t_n(x)$ and the depth bias of the global trace. The algorithmic payoff is the gated EM update $\theta_{nc} = (N_{nc} + \sqrt{N_{nc}^2 + 4\lambda_n \mu \omega_n N_{nc}})/(2\lambda_n)$, identical in form to the global trace update with the node-specific strength $\mu\omega_n$.
What would settle it
Take any trained smooth, decomposable PC with strictly positive outgoing sum weights and any input $x$; compute every node's $T_n(x)$, $F_n(x)^2$, and $t_n(x)$ from a single forward–backward pass and check whether $T_n(x) = F_n(x)^2 t_n(x)$ holds to numerical precision, a single violation falsifies the identity. To test the gating claim, compare the Algorithm 2 gate $\sum_c \bigl(\frac{1}{N}\sum_i p_c(x_i)/p_n(x_i)\bigr)^2$ with the true empirical local trace $\frac{1}{N}\sum_i \sum_c (p_c(x_i)/p_n(x_i))^2$ on trained circuits: if the two rank nodes differently, the reported gains are not attributable to local-curvature gating.
Extended reading notes
Core claim
On the paper's own terms, the central discovery is an exact identity: in any smooth, decomposable probabilistic circuit with strictly positive sum weights, the contribution $T_n(x)$ of a sum node $n$ to the Hessian trace of the negative log-likelihood factorizes as $T_n(x) = F_n(x)^2 t_n(x)$, where $F_n$ is the circuit flow through $n$, the sensitivity of the root probability to the node's output, and $t_n(x) = \sum_c (p_c(x)/p_n(x))^2$ is a purely local sharpness term determined by the node's own mixture. Summing over sum nodes recovers the total trace, $\mathrm{Tr}(\nabla^2_\theta \ell) = \sum_n F_n(x)^2 t_n(x)$. Around this identity the paper builds a full characterization: the local Hessian of a sum node is rank one with unique nonzero eigenvalue exactly $t_n$; in a tree-structured circuit flow is the product of routing responsibilities along upstream sum edges, so the global trace is depth-biased and locally sharp deep nodes are discounted geometrically in the number of upstream sum edges; and two nodes' global rankings reverse precisely when $t_i/t_j > (F_j/F_i)^2$, so the most globally curved nodes need not be the locally sharpest ones. These results motivate the paper's proposal: keep the trace penalty but gate it per node by the empirical local curvature through $\omega_n = \hat{t}_n / \max_m \hat{t}_m$, which preserves the closed-form EM update with node-specific strength $\mu\omega_n$ and, empirically, recovers the fit that uniform global regularization sacrifices.
Load-bearing premise
The reported gains of the adaptive regularizer rest on the assumption that the gate it computes in Algorithm 2, the square of the dataset-averaged child-to-parent output ratio, faithfully tracks the true average local curvature, even though the two quantities are not mathematically the same.
Editorial extensions
If this is right
- Global sharpness regularization is depth-biased: because flow is attenuated along upstream sum edges, the trace penalty lands preferentially on shallow, heavily used nodes, and uniform application can lower both training and test likelihood even as the model becomes flatter.
- The global trace contribution $T_n$ and the local curvature $t_n$ induce different node rankings, with a reversal exactly when $t_i/t_j > (F_j/F_i)^2$; selecting nodes by their global contribution can therefore worsen performance, as the experiments observe.
- Local curvature $t_n$ alone determines a sum node's entire ambient second-order geometry, since its local Hessian is rank one with eigenvalue $t_n$, so it is a sound basis for allocating regularization.
- A locality-gated trace regularizer preserves the closed-form, linear-time EM update, recovering the unregularized model's fit on most benchmarks while keeping the low-data generalization gains of sharpness-aware training.
- In the high-data regime the gated method matches or beats the unregularized baseline on the majority of the 20 DEBD datasets, whereas global trace regularization improves only two and degrades the remaining eighteen.
Reading between the lines
- The same decomposition suggests curvature-aware compression: nodes low in both flow and local curvature contribute negligibly to the trace, so measuring $F_n^2$ and $t_n$ separately could reveal which mixture components can be pruned without touching the structurally load-bearing ones.
- Because the identity holds per input, the measurement pipeline could also support targeted robustness analysis, for example flagging locally sharp but low-flow nodes that have little influence on typical inputs yet might destabilize under distribution shift.
- The paper itself notes that its simple gate is a proof of concept and that any monotone gate preserves the closed-form update; a direct testable extension is a gate that jointly accounts for local geometry and contextual usage, such as $F_n^2 t_n$ or a rank-based variant.
- A natural check of the adaptive method's mechanism is whether gating on the true per-sample average of squared output ratios, rather than the square of the dataset-averaged ratio used in the paper's Algorithm 2, changes which nodes get regularized and by how much.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This paper develops a compositional theory of curvature in probabilistic circuits. It proves that, for smooth and decomposable PCs, the per-node contribution to the Hessian trace of the negative log-likelihood factorizes exactly as T_n(x) = F_n(x)^2 t_n(x), where F_n is the circuit flow through node n and t_n(x) = sum_c (p_c(x)/p_n(x))^2 is a local trace. It further shows that the local Hessian of a sum node is rank one with unique nonzero eigenvalue t_n, analyzes how flow attenuation through sum edges creates a depth bias in the global trace, and derives an exact condition for ranking reversals between global contribution and local curvature. Motivated by experiments showing that global trace regularization underfits at higher data fractions, the paper proposes a gated regularizer that scales the trace penalty by a monotone function of the empirical local trace, preserving closed-form EM updates. Experiments on 20 DEBD datasets compare unregularized, global-trace, and gated training, reporting that gated training recovers much of the fit lost by global regularization.
Significance. The theoretical core is sound and valuable: the factorization is exact, the rank-one local-Hessian characterization is clean, and the depth-bias analysis provides a concrete mechanism for the observed underfitting of global trace regularization. The empirical observations, including concentration of the global trace in few nodes and the failure of top-global-contribution selection, are interesting and largely consistent with the theory. However, the paper's algorithmic implementation and its derivation contain two load-bearing problems: the gate computed in Algorithm 2 does not match the local trace defined in Eq. (6), and the proof of the closed-form update contains a mathematical error. Because the experimental claims are presented as validating the local-curvature mechanism, these issues must be resolved before the paper can be accepted.
major comments (2)
- [Algorithm 2, lines 9-10; 'Local-Curvature Gating' and Eq. (6)] The gate implemented in Algorithm 2 computes \hat t_n = \sum_{c in ch(n)} ( (1/|D|) \sum_{x in D} p_c(x)/p_n(x) )^2, whereas the paper defines the empirical local trace as \hat b_{t_n} = (1/N) \sum_{i=1}^N \sum_{c in ch(n)} (p_c(x_i)/p_n(x_i))^2. These two quantities are not equal: by Jensen's inequality the latter is at least the former for each child term, and the induced node rankings can differ. Since the adaptive method is motivated as allocating regularization according to the local curvature of Eq. (6), the results in Table 2 and Figures 7-8 cannot be interpreted as validating that mechanism unless Algorithm 2 is corrected or the authors show that the two quantities produce the same node ranking on the evaluated circuits. No code is released, so the reader cannot determine which quantity was actually used. This is a load-bearing issue for the central empirical claim.
- [Gated EM Update and Proposition 3; Appendix 'Adaptive Gated Regularization', Proposition 6] The derivation of the closed-form update in Proposition 3 is not mathematically consistent. In the appendix, the Lagrangian for the regularized objective is differentiated to obtain \partial L/\partial \theta_{nc} = N_{nc}/\theta_{nc} - \lambda + \mu \omega_n N_{nc}/\theta_{nc}^2, but the correct derivative of -\mu \omega_n \sum_c (N_{nc}/\theta_{nc})^2 is N_{nc}/\theta_{nc} - \lambda + 2\mu \omega_n N_{nc}^2/\theta_{nc}^3. The stated quadratic equation \lambda \theta_{nc}^2 - N_{nc}\theta_{nc} - \mu \omega_n N_{nc} = 0 therefore does not follow from that Lagrangian. In addition, the main text's objective uses S_{nc} = \sum_i F_{nc}(x_i)^2, while the appendix uses N_{nc}, and the 'surrogate used by global trace-regularized EM' is never explicitly defined. Since Eq. (7) is the proposed update rule, the authors need to provide a correct derivation or explicitly state and prove the surrogate step.
minor comments (5)
- [Algorithm 2 and Eq. (6)] The symbol \hat t_n is used for two different quantities: the empirical local trace in the main text and the dataset-averaged marginal-ratio summary in Algorithm 2. Distinct notation should be used to prevent exactly the confusion described above.
- [Algorithm 2, line 2] The initialization \omega_n \leftarrow g(0) is undefined for the proposed gate g(\hat t) = \hat t / \max \hat t, since 0/0 is not defined. The authors should specify the value of g(0) or initialize gates after the first forward-backward pass.
- [Figure 6 caption] The figure reports 'Normalised contribution (%)' for both the global and local traces, but these quantities are normalized separately on different scales; the caption should say so explicitly to avoid implying a common scale.
- [Introduction and references] The paper relies heavily on the authors' prior work (Suresh et al. 2026) for the global trace regularizer and the EM surrogate; the introduction should state this dependency more explicitly when framing the contribution.
- [Q3, Table 2] The text claims global regularization 'degrades performance on the remaining 18' datasets at the full data fraction, which is consistent with the table, but on bbc and dna the global regularizer improves substantially; a one-sentence acknowledgement of these exceptions would make the summary more precise.
Circularity Check
No significant circularity: Theorem 1 is derived from the flow identity and the trace definition rather than assumed, and the self-citations to prior work are transparent and not load-bearing.
full rationale
The central decomposition Theorem 1 is not circular. T_n(x) is defined in Eq. (5) as the sum over children of (F_nc/theta_nc)^2, and the edge-flow identity F_nc/theta_nc = rho_nc F_n (Eq. (2), restated in the supplementary proof) gives T_n(x) = F_n(x)^2 * sum_c (p_c(x)/p_n(x))^2. The local trace t_n is defined as that residual sum of squared output ratios, so the factorization is proven by substitution rather than assumed. Proposition 1 independently computes the local Hessian as rho rho^T, which is rank one with trace t_n, so the geometric reading of t_n is derived. Lemma 1, Corollary 1, and Proposition 2 are algebraic consequences of the flow recursion and Theorem 1. The adaptive gate in Eq. (6) and the closed-form update in Proposition 3 are presented as design choices, with the update explicitly inherited from the authors' prior AAAI work through a named surrogate; this is a disclosed self-citation, not a smuggled premise, and the trace formula itself is re-justified in the text by multilinearity of p_r in each sum weight. No prediction in the paper is statistically forced by a fitted parameter. A non-circular concern is that Algorithm 2 computes the gate from dataset-averaged ratios squared, namely \hat t_n = sum_c ((1/|D|) sum_{x in D} p_c(x)/p_n(x))^2, whereas Eq. (6) defines b_t_n = (1/N) sum_i sum_c (p_c(x_i)/p_n(x_i))^2; these quantities differ, so the implemented gate may not exactly match the paper's proposed local-curvature mechanism. This is an implementation and validation issue, not a circularity of the derivation.
Assumptions & free parameters
free parameters (3)
- regularization strength mu =
selected per dataset/fraction via validation log-likelihood
- simplex Lagrange multiplier lambda =
not specified in the paper
- gate function g =
omega_n = b_t_n / max_m b_t_m (Eq 6)
assumptions (5)
- domain assumption The Hessian trace of the NLL over sum weights equals the sum of squared edge-flow ratios: Tr = sum_nc (F_nc/theta_nc)^2 (Eq 3)
- standard math Basic calculus and linear algebra: outer product rho*rho^T has rank one with eigenvalue ||rho||^2; derivative of -log sum theta_c p_c is -rho
- domain assumption The EM surrogate from global trace regularization remains valid when mu is replaced by node-specific mu*omega_n
- ad hoc to paper The gate computed as sum_c (dataset-averaged p_c/p_n)^2 ranks nodes similarly to the mean over samples of sum_c (p_c/p_n)^2
- domain assumption PCs are smooth and decomposable with alternating sum/product layers
Cite this review
Pith. "Pith review of A Compositional Theory of Curvature in Probabilistic Circuits." pith.science (2026). https://pith.science/paper/AL4P6EFK
@misc{pith2026260812869,
author = {Pith},
title = {Pith review of: A Compositional Theory of Curvature in Probabilistic Circuits},
year = {2026},
howpublished = {\url{https://pith.science/paper/AL4P6EFK}},
note = {Machine review of arXiv:2608.12869}
}
read the original abstract
Probabilistic Circuits (PCs) are generative models that support exact inference and, unlike deep neural networks, admit an exact and tractable measure of loss-surface curvature: the trace of the Hessian of the log-likelihood. Recent work regularizes this trace globally to bias learning toward flatter, better generalizing optima. We show that treating sharpness as a global regularizer can be misspecified for PCs, whose curvature is inherently compositional. We prove that each sum node's contribution to the Hessian trace factorizes exactly into its circuit flow, which measures how heavily the node is used, and a local sharpness term determined by its output distribution. This decomposition provides insights into why global sharpness regularization is depth biased and can lead to underfitting. Building on it, we introduce an adaptive sharpness aware regularizer that penalizes nodes based on intrinsic local curvature and preserves closed form EM updates. We also show that empirically, this targeted regularization recovers the generalization that global regularization sacrifices while retaining the robustness and benefits of sharpness aware learning.
Figures
Figures from the paper (11 more)
Reference graph
Works this paper leans on
-
[1]
Relative Flatness and Generalization , year =
Henning Petzka and Michael Kamp and Linara Adilova and Cristian Sminchisescu and Mario Boley , bibsource =. Relative Flatness and Generalization , year =. Advances in Neural Information Processing Systems 34: Annual Conference on Neural Information Processing Systems 2021 , pages =
2021
-
[2]
Forty-third International Conference on Machine Learning , year=
Geometry-Aware Probabilistic Circuits via Voronoi Tessellations , author=. Forty-third International Conference on Machine Learning , year=
-
[3]
Proceedings of the 40th International Conference on Machine Learning , pages =
A Modern Look at the Relationship between Sharpness and Generalization , author =. Proceedings of the 40th International Conference on Machine Learning , pages =. 2023 , volume =
2023
-
[4]
Advances in Neural Information Processing Systems , volume=
Adaptable logical control for large language models , author=. Advances in Neural Information Processing Systems , volume=
-
[5]
International Conference on Machine Learning , pages=
Sharp minima can generalize for deep nets , author=. International Conference on Machine Learning , pages=. 2017 , organization=
2017
-
[6]
International Conference on Learning Representations , year=
Lossless Compression with Probabilistic Circuits , author=. International Conference on Learning Representations , year=
-
[7]
5th International Conference on Learning Representations, 2017 , title =
Nitish Shirish Keskar and Dheevatsa Mudigere and Jorge Nocedal and Mikhail Smelyanskiy and Ping Tak Peter Tang , bibsource =. 5th International Conference on Learning Representations, 2017 , title =
2017
-
[8]
Tractable Control for Autoregressive Language Generation , year =
Honghua Zhang and Meihua Dang and Nanyun Peng and Guy Van den Broeck , bibsource =. Tractable Control for Autoregressive Language Generation , year =. International Conference on Machine Learning,2023 , pages =
2023
Show all 93 references
-
[9]
AAAI Conference on Artificial Intelligence, 2025 , volume=
A Unified Framework for Human-Allied Learning of Probabilistic Circuits , author=. AAAI Conference on Artificial Intelligence, 2025 , volume=
2025
-
[10]
How to Turn Your Knowledge Graph Embeddings into Generative Models via Probabilistic Circuits , year =
Lorenzo Loconte and Nicola Di Mauro and Robert Peharz and Antonio Vergari , booktitle =. How to Turn Your Knowledge Graph Embeddings into Generative Models via Probabilistic Circuits , year =
-
[11]
Advances in Neural Information Processing Systems 35: Annual Conference on Neural Information Processing Systems 2022 , title =
Kareem Ahmed and Stefano Teso and Kai. Advances in Neural Information Processing Systems 35: Annual Conference on Neural Information Processing Systems 2022 , title =
2022
-
[12]
Probabilistic circuits that know what they don't know , year =
Fabrizio Ventola and Steven Braun and Zhongjie Yu and Martin Mundt and Kristian Kersting , bibsource =. Probabilistic circuits that know what they don't know , year =. Uncertainty in Artificial Intelligence, 2023 , pages =
2023
-
[13]
The 28th International Conference on Artificial Intelligence and Statistics , year =
Sahil Sidheekh and Pranuthi Tenali and Saurabh Mathur and Erik Blasch and Kristian Kersting and Sriraam Natarajan , title =. The 28th International Conference on Artificial Intelligence and Statistics , year =
-
[14]
Image Inpainting via Tractable Steering of Diffusion Models , year =
Anji Liu and Mathias Niepert and Guy Van den Broeck , bibsource =. Image Inpainting via Tractable Steering of Diffusion Models , year =. The Twelfth International Conference on Learning Representations, 2024 , timestamp =
2024
-
[15]
Balasubramanian , bibsource =
Adepu Ravi Sankar and Yash Khasbage and Rahul Vigneswaran and Vineeth N. Balasubramanian , bibsource =. A Deeper Look at the Hessian Eigenspectrum of Deep Neural Networks and its Applications to Regularization , year =. Thirty-Fifth
-
[16]
Lipton , booktitle =
Simran Kaur and Jeremy Cohen and Zachary C. Lipton , booktitle =. On the Maximum Hessian Eigenvalue and Generalization , year =
-
[17]
Ugur Güney and Yann Dauphin and Léon Bottou , booktitle =
Levent Sagun and Utku Evci and V. Ugur Güney and Yann Dauphin and Léon Bottou , booktitle =. Empirical analysis of the hessian of over-parametrized neural networks , year =
-
[18]
Asymmetric Valleys: Beyond Sharp and Flat Local Minima , year =
Haowei He and Gao Huang and Yang Yuan , bibsource =. Asymmetric Valleys: Beyond Sharp and Flat Local Minima , year =. Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, December 8-14, 2019, Vanco...
2019
-
[19]
Improving Generalization in Federated Learning by Seeking Flat Minima , year =
Debora Caldarola and Barbara Caputo and Marco Ciccone , booktitle =. Improving Generalization in Federated Learning by Seeking Flat Minima , year =
-
[20]
Geometry of the Loss Landscape in Overparameterized Neural Networks: Symmetries and Invariances , year =
Berfin Simsek and Fran. Geometry of the Loss Landscape in Overparameterized Neural Networks: Symmetries and Invariances , year =. 38th International Conference on Machine Learning 2021 , pages =
2021
-
[21]
SIAM Journal on Mathematics of Data Science , title =
Yaim Cooper , issn =. SIAM Journal on Mathematics of Data Science , title =
-
[22]
Advances in Neural Information Processing Systems 31: Annual Conference on Neural Information Processing Systems 2018, , title =
Hao Li and Zheng Xu and Gavin Taylor and Christoph Studer and Tom Goldstein , bibsource =. Advances in Neural Information Processing Systems 31: Annual Conference on Neural Information Processing Systems 2018, , title =
2018
-
[23]
Variational Inference with Normalizing Flows , year =
Danilo Jimenez Rezende and Shakir Mohamed , bibsource =. Variational Inference with Normalizing Flows , year =. 32nd International Conference on Machine Learning, 2015 , pages =
2015
-
[24]
Gradient Norm Aware Minimization Seeks First-Order Flatness and Improves Generalization , year =
Xingxuan Zhang and Renzhe Xu and Han Yu and Hao Zou and Peng Cui , bibsource =. Gradient Norm Aware Minimization Seeks First-Order Flatness and Improves Generalization , year =
-
[25]
38th International Conference on Machine Learning, 2021 , pages =
Jungmin Kwon and Jeongseop Kim and Hyunseo Park and In Kwon Choi , bibsource =. 38th International Conference on Machine Learning, 2021 , pages =
2021
-
[26]
Building Expressive and Tractable Probabilistic Generative Models:
Sahil Sidheekh and Sriraam Natarajan , bibsource =. Building Expressive and Tractable Probabilistic Generative Models:. Thirty-Third International Joint Conference on Artificial Intelligence, 2024 , pages =
2024
-
[27]
Cutset networks: A simple, tractable, and scalable approach for improving the accuracy of chow-liu trees , year =
Rahman, Tahrima and Kothalkar, Prasanna and Gogate, Vibhav , booktitle =. Cutset networks: A simple, tractable, and scalable approach for improving the accuracy of chow-liu trees , year =
-
[28]
A Differential Approach to Inference in Bayesian Networks , year =
Darwiche, Adnan , issue_date =. A Differential Approach to Inference in Bayesian Networks , year =. Journal of the ACM , number =
-
[29]
Sharpness-aware Minimization for Efficiently Improving Generalization , year =
Pierre Foret and Ariel Kleiner and Hossein Mobahi and Behnam Neyshabur , bibsource =. Sharpness-aware Minimization for Efficiently Improving Generalization , year =. 9th International Conference on Learning Representations, 2021 , timestamp =
2021
-
[30]
Scaling Probabilistic Circuits via Monarch Matrices , year =
Zhang, Honghua and Dang, Meihua and Wang, Benjie and Ermon, Stefano and Peng, Nanyun and Van den Broeck, Guy , booktitle =. Scaling Probabilistic Circuits via Monarch Matrices , year =
-
[31]
Visualizing the Loss Landscape of Neural Nets , year =
Hao Li and Zheng Xu and Gavin Taylor and Christoph Studer and Tom Goldstein , bibsource =. Visualizing the Loss Landscape of Neural Nets , year =. Advances in Neural Information Processing Systems 31: Annual Conference on Neural Information Processing Systems 2018 , pages =
2018
-
[32]
Flat minima , year =
Hochreiter, Sepp and Schmidhuber, J. Flat minima , year =. Neural computation , number =
-
[33]
Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning , year =
Yarin Gal and Zoubin Ghahramani , bibsource =. Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning , year =. 33rd International Conference on Machine Learning, 2016 , pages =
2016
-
[34]
Einsum Networks: Fast and Scalable Learning of Tractable Probabilistic Circuits , year =
Robert Peharz and Steven Lang and Antonio Vergari and Karl Stelzner and Alejandro Molina and Martin Trapp and Guy Van den Broeck and Kristian Kersting and Zoubin Ghahramani , bibsource =. Einsum Networks: Fast and Scalable Learning of Tractable Probabilistic Circuits , year =....
2020
-
[35]
Strudel: Learning Structured-Decomposable Probabilistic Circuits , year =
Meihua Dang and Antonio Vergari and Guy Van den Broeck , booktitle =. Strudel: Learning Structured-Decomposable Probabilistic Circuits , year =
-
[36]
de Campos and Robert Peharz and Antonio Vergari and Erik Quaeghebeur , bibsource =
Gennaro Gala and Cassio P. de Campos and Robert Peharz and Antonio Vergari and Erik Quaeghebeur , bibsource =. Probabilistic Integral Circuits , year =. International Conference on Artificial Intelligence and Statistics, 2024 , pages =
2024
-
[37]
Advances in Neural Information Processing Systems 35: Annual Conference on Neural Information Processing Systems, 2022 , title =
Meihua Dang and Anji Liu and Guy Van den Broeck , bibsource =. Advances in Neural Information Processing Systems 35: Annual Conference on Neural Information Processing Systems, 2022 , title =
2022
-
[38]
The Eleventh International Conference on Learning Representations, 2023 , title =
Anji Liu and Honghua Zhang and Guy Van den Broeck , bibsource =. The Eleventh International Conference on Learning Representations, 2023 , title =
2023
-
[39]
Sum of squares circuits , year =
Loconte, Lorenzo and Mengel, Stefan and Vergari, Antonio , booktitle =. Sum of squares circuits , year =
-
[40]
Transactions on Machine Learning Research , issn=
Tractable Representation Learning with Probabilistic Circuits , author=. Transactions on Machine Learning Research , issn=
-
[41]
Sladek and Stefan Mengel and Martin Trapp and Arno Solin and Nicolas Gillis and Antonio Vergari , bibsource =
Lorenzo Loconte and Aleksanteri M. Sladek and Stefan Mengel and Martin Trapp and Arno Solin and Nicolas Gillis and Antonio Vergari , bibsource =. The Twelfth International Conference on Learning Representations, 2024 , title =
2024
-
[42]
Tractable Regularization of Probabilistic Circuits , year =
Anji Liu and Guy Van den Broeck , bibsource =. Tractable Regularization of Probabilistic Circuits , year =. Advances in Neural Information Processing Systems 34: Annual Conference on Neural Information Processing Systems, 2021 , pages =
2021
-
[43]
Simplifying, regularizing and strengthening sum-product network structure learning , year =
Antonio Vergari and Nicola Di Mauro and Floriana Esposito , booktitle =. Simplifying, regularizing and strengthening sum-product network structure learning , year =
-
[44]
On the Relationship between Sum-Product Networks and Bayesian Networks , year =
Han Zhao and Mazen Melibari and Pascal Poupart , bibsource =. On the Relationship between Sum-Product Networks and Bayesian Networks , year =. 32nd International Conference on Machine Learning, 2015 , pages =
2015
-
[45]
Random Sum-Product Networks:
Robert Peharz and Antonio Vergari and Karl Stelzner and Alejandro Molina and Martin Trapp and Xiaoting Shao and Kristian Kersting and Zoubin Ghahramani , bibsource =. Random Sum-Product Networks:. Thirty-Fifth Conference on Uncertainty in Artificial Intelligence, 2019 , pages =
2019
-
[46]
Group Fairness by Probabilistic Modeling with Latent Fair Decisions , year =
YooJung Choi and Meihua Dang and Guy Van den Broeck , bibsource =. Group Fairness by Probabilistic Modeling with Latent Fair Decisions , year =. Thirty-Fifth
-
[47]
Probabilistic sentential decision diagrams , year =
Doga Kisa and Guy Van Den Broeck and Arthur Choi and Adnan Darwiche , booktitle =. Probabilistic sentential decision diagrams , year =
-
[48]
HyperSPNs: Compact and Expressive Probabilistic Circuits , year =
Andy Shih and Dorsa Sadigh and Stefano Ermon , bibsource =. HyperSPNs: Compact and Expressive Probabilistic Circuits , year =. Advances in Neural Information Processing Systems 34: Annual Conference on Neural Information Processing Systems, 2021 , pages =
2021
-
[49]
On the relationship between monotone and squared probabilistic circuits , year =
Wang, Benjie and Van den Broeck, Guy , booktitle =. On the relationship between monotone and squared probabilistic circuits , year =
-
[50]
Domingos , bibsource =
Hoifung Poon and Pedro M. Domingos , bibsource =. Sum-Product Networks:. Twenty-Seventh Conference on Uncertainty in Artificial Intelligence, 2011 , pages =
2011
-
[51]
IEEE Transactions on Pattern Analysis and Machine Intelligence , title =
Robert Peharz and Robert Gens and Franz Pernkopf and Pedro Domingos , issn =. IEEE Transactions on Pattern Analysis and Machine Intelligence , title =
-
[52]
Machine Learning , title =
Antonio Vergari and Nicola Di Mauro and Floriana Esposito , issn =. Machine Learning , title =
-
[53]
IEEE Transactions on Pattern Analysis and Machine Intelligence , title =
Raquel Sanchez-Cauce and Iago Paris and Francisco Javier Diez , issn =. IEEE Transactions on Pattern Analysis and Machine Intelligence , title =
-
[54]
Advances in Neural Information Processing Systems , volume=
Tractable learning for complex probability queries , author=. Advances in Neural Information Processing Systems , volume=
-
[55]
YooJung Choi and Antonio Vergari and Guy Van den Broeck , title =
-
[56]
AAAI Conference on Artificial Intelligence , volume=
Markov network structure learning: A randomized feature generation approach , author=. AAAI Conference on Artificial Intelligence , volume=
-
[57]
Uncertainty in Artificial Intelligence , pages=
VQ-Flows: Vector quantized local normalizing flows , author=. Uncertainty in Artificial Intelligence , pages=. 2022 , organization=
2022
-
[58]
Advances in Neural Information Processing Systems , volume=
A compositional atlas of tractable circuit operations for probabilistic inference , author=. Advances in Neural Information Processing Systems , volume=
-
[59]
Nalisnick and Danilo Jimenez Rezende and Shakir Mohamed and Balaji Lakshminarayanan , journal=
George Papamakarios and Eric T. Nalisnick and Danilo Jimenez Rezende and Shakir Mohamed and Balaji Lakshminarayanan , journal=. Normalizing Flows for Probabilistic Modeling and Inference , year =
-
[60]
Stochastic Backpropagation and Approximate Inference in Deep Generative Models , year =
Danilo Jimenez Rezende and Shakir Mohamed and Daan Wierstra , bibsource =. Stochastic Backpropagation and Approximate Inference in Deep Generative Models , year =. 31st International Conference on Machine Learning, 2014 , pages =
2014
-
[61]
Kingma and Max Welling , bibsource =
Diederik P. Kingma and Max Welling , bibsource =. Auto-Encoding Variational Bayes , year =. 2nd International Conference on Learning Representations, 2014 , timestamp =
2014
-
[62]
Goodfellow and Jean Pouget
Ian J. Goodfellow and Jean Pouget. Advances in Neural Information Processing Systems 27: Annual Conference , title =
-
[63]
Scaling Tractable Probabilistic Circuits:
Anji Liu and Kareem Ahmed and Guy Van den Broeck , booktitle =. Scaling Tractable Probabilistic Circuits:
-
[64]
Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems, 2023 , title =
Jihun Yun and Eunho Yang , bibsource =. Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems, 2023 , title =
2023
-
[65]
SPFlow: An Easy and Extensible Library for Deep Probabilistic Learning Using Sum-Product Networks , year =
Robert Peharz and Alejandro Molina and Krzysztof Trapp and Zoubin Ghahramani and Kristian Kersting , journal =. SPFlow: An Easy and Extensible Library for Deep Probabilistic Learning Using Sum-Product Networks , year =
-
[66]
arXiv , author =:1604.07243 , primaryclass =
Learning Arbitrary Sum-Product Network Leaves with Expectation-Maximization , year =. arXiv , author =:1604.07243 , primaryclass =
-
[67]
Gordon , bibsource =
Han Zhao and Pascal Poupart and Geoffrey J. Gordon , bibsource =. A Unified Approach for Learning the Parameters of Sum-Product Networks , year =. Advances in Neural Information Processing Systems 29: Annual Conference on Neural Information Processing Systems, 2016 , pages =
2016
-
[68]
Domingos , bibsource =
Robert Gens and Pedro M. Domingos , bibsource =. Discriminative Learning of Sum-Product Networks , year =. Advances in Neural Information Processing Systems 25: 26th Annual Conference on Neural Information Processing Systems, 2012 , pages =
2012
-
[69]
On the Sample Complexity of Learning Sum-Product Networks , year =
Ishaq Aden. On the Sample Complexity of Learning Sum-Product Networks , year =. The 23rd International Conference on Artificial Intelligence and Statistics, 2020 , pages =
2020
-
[70]
Domingos , bibsource =
Robert Peharz and Sebastian Tschiatschek and Franz Pernkopf and Pedro M. Domingos , bibsource =. On Theoretical Properties of Sum-Product Networks , year =. Eighteenth International Conference on Artificial Intelligence and Statistics, 2015 , series =
2015
-
[71]
Machine Learning , title =
Mattia Desana and Christoph Schnörr , issn =. Machine Learning , title =
-
[72]
Journal of Statistical Mechanics: Theory and Experiment , title =
Lucas Böttcher and Gregory Wheeler , issn =. Journal of Statistical Mechanics: Theory and Experiment , title =
-
[73]
Probabilistic Flow Circuits: Towards Unified Deep Models for Tractable Probabilistic Inference , year =
Sahil Sidheekh and Kristian Kersting and Sriraam Natarajan , bibsource =. Probabilistic Flow Circuits: Towards Unified Deep Models for Tractable Probabilistic Inference , year =. Uncertainty in Artificial Intelligence, 2023 , pages =
2023
-
[74]
Deep Tractable Probabilistic Models , year =
Sahil Sidheekh and Saurabh Mathur and Athresh Karanam and Sriraam Natarajan , booktitle =. Deep Tractable Probabilistic Models , year =
-
[75]
Interventional Sum-Product Networks: Causal Inference with Tractable Probabilistic Models , year =
Matej Zecevic and Devendra Singh Dhami and Athresh Karanam and Sriraam Natarajan and Kristian Kersting , bibsource =. Interventional Sum-Product Networks: Causal Inference with Tractable Probabilistic Models , year =. Advances in Neural Information Processing Systems 34: Annua...
2021
-
[76]
Haas and Kristian Kersting and Sriraam Natarajan , booktitle =
Athresh Karanam and Saurabh Mathur and Predrag Radivojac and David M. Haas and Kristian Kersting and Sriraam Natarajan , booktitle =. Explaining Deep Tractable Probabilistic Models: The sum-product network case , year =
-
[77]
M. F. Hutchinson , issn =. Communications in Statistics - Simulation and Computation , title =
-
[78]
AAAI Conference on Artificial Intelligence , volume=
Continuous mixtures of tractable probabilistic models , author=. AAAI Conference on Artificial Intelligence , volume=
-
[79]
Transactions on Machine Learning Research , issn=
What is the Relationship between Tensor Factorizations and Circuits (and How Can We Exploit it)? , author=. Transactions on Machine Learning Research , issn=
-
[80]
Chayes and Levent Sagun and Riccardo Zecchina , bibsource =
Pratik Chaudhari and Anna Choromanska and Stefano Soatto and Yann LeCun and Carlo Baldassi and Christian Borgs and Jennifer T. Chayes and Levent Sagun and Riccardo Zecchina , bibsource =. Entropy-SGD: Biasing Gradient Descent Into Wide Valleys , year =. 5th International Confe...
2017
-
[81]
29th International Conference on Machine Learning, 2012 , title =
James Martens and Ilya Sutskever and Kevin Swersky , bibsource =. 29th International Conference on Machine Learning, 2012 , title =
2012
-
[82]
AAAI Conference on Artificial Intelligence, 2012 , title =
Jan Van Haaren and Jesse Davis , bibsource =. AAAI Conference on Artificial Intelligence, 2012 , title =
2012
-
[83]
Journal of Machine Learning Research , title =
Daniel Lowd and Jesse Davis , issn =. Journal of Machine Learning Research , title =
-
[84]
AAAI Conference on Artificial Intelligence , volume=
The Gradient of Algebraic Model Counting , author=. AAAI Conference on Artificial Intelligence , volume=
-
[85]
AAAI Conference on Artificial Intelligence , volume=
Probabilistic neural circuits , author=. AAAI Conference on Artificial Intelligence , volume=
-
[86]
International Conference on Artificial Intelligence and Statistics , pages=
Sum-product-quotient networks , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2018 , organization=
2018
-
[87]
International Journal of Approximate Reasoning , volume=
Conditional sum-product networks: Modular probabilistic circuits via gate functions , author=. International Journal of Approximate Reasoning , volume=. 2022 , publisher=
2022
-
[88]
arXiv preprint arXiv:2508.05537 , year=
Tractable Sharpness-Aware Learning of Probabilistic Circuits , author=. arXiv preprint arXiv:2508.05537 , year=
-
[89]
International Conference on Artificial Intelligence and Statistics , pages=
Bayesian structure scores for probabilistic circuits , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2023 , organization=
2023
-
[90]
2026 , eprint=
PAR: Plausibility-aware Amortized Recourse Generation , author=. 2026 , eprint=
2026
-
[91]
International Conference on Learning Representations , volume=
Generating likely counterfactuals using sum-product networks , author=. International Conference on Learning Representations , volume=
-
[92]
Proceedings of the AAAI Conference on Artificial Intelligence , author=
Tractable Sharpness-Aware Learning of Probabilistic Circuits , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2026 , month=. doi:10.1609/aaai.v40i30.39771 , number=
2026 doi
-
[93]
ArXiv , year=
Fantastic Generalization Measures and Where to Find Them , author=. ArXiv , year=
Reviewed August 15, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.