Scholars@Duke

On Neural Networks as Infinite Tree-Structured Probabilistic Graphical Models.

Publication ,  Conference
Li, B; Thomson, AJ; Nassif, H; Engelhard, MM; Page, D
Published in: Advances in neural information processing systems
January 2024
Published version (DOI)

Deep neural networks (DNNs) lack the precise semantics and definitive probabilistic interpretation of probabilistic graphical models (PGMs). In this paper, we propose an innovative solution by constructing infinite tree-structured PGMs that correspond exactly to neural networks. Our research reveals that DNNs, during forward propagation, indeed perform approximations of PGM inference that are precise in this alternative PGM structure. Not only does our research complement existing studies that describe neural networks as kernel machines or infinite-sized Gaussian processes, it also elucidates a more direct approximation that DNNs make to exact inference in PGMs. Potential benefits include improved pedagogy and interpretation of DNNs, and algorithms that can merge the strengths of PGMs and DNNs.

Duke Scholars

Matthew M. Engelhard
Author Matthew M. Engelhard Biostatistics & Bioinformatics, Division of Translational Bi ...
David Page
Author David Page Biostatistics & Bioinformatics, Division of Biostatistics

Published In

Advances in neural information processing systems

DOI

10.48550/arxiv.2305.17583

ISSN

1049-5258

Publication Date

January 2024

Volume

37

Start / End Page

4598 / 4628

Related Subject Headings

  • 4611 Machine learning
  • 1702 Cognitive Sciences
  • 1701 Psychology
 

Citation

APA
Chicago
ICMJE
MLA
NLM
Li, B., Thomson, A. J., Nassif, H., Engelhard, M. M., & Page, D. (2024). On Neural Networks as Infinite Tree-Structured Probabilistic Graphical Models. In Advances in neural information processing systems (Vol. 37, pp. 4598–4628). https://doi.org/10.48550/arxiv.2305.17583
Li, Boyao, Alexander J. Thomson, Houssam Nassif, Matthew M. Engelhard, and David Page. “On Neural Networks as Infinite Tree-Structured Probabilistic Graphical Models.” In Advances in Neural Information Processing Systems, 37:4598–4628, 2024. https://doi.org/10.48550/arxiv.2305.17583.
Li B, Thomson AJ, Nassif H, Engelhard MM, Page D. On Neural Networks as Infinite Tree-Structured Probabilistic Graphical Models. In: Advances in neural information processing systems. 2024. p. 4598–628.
Li, Boyao, et al. “On Neural Networks as Infinite Tree-Structured Probabilistic Graphical Models.” Advances in Neural Information Processing Systems, vol. 37, 2024, pp. 4598–628. Epmc, doi:10.48550/arxiv.2305.17583.
Li B, Thomson AJ, Nassif H, Engelhard MM, Page D. On Neural Networks as Infinite Tree-Structured Probabilistic Graphical Models. Advances in neural information processing systems. 2024. p. 4598–4628.

Published In

Advances in neural information processing systems

DOI

10.48550/arxiv.2305.17583

ISSN

1049-5258

Publication Date

January 2024

Volume

37

Start / End Page

4598 / 4628

Related Subject Headings

  • 4611 Machine learning
  • 1702 Cognitive Sciences
  • 1701 Psychology