PILOGIC
Research

Exact reasoning,
on the record.

Exact reasoning in a Bayesian network hit a wall decades ago: past a certain complexity the standard algorithms ran out of time and memory, and most of the field settled for approximations. The 2008 paper behind PiLogic's engine took another route, turning the question into a counting problem and compiling the model once, so every later answer is fast and exact.

NASA ran the method on its power-system testbed the same year. An independent study put 18 methods through 672 benchmark problems in 2021, and this approach solved the most. PiLogic's engine, now in its fifth generation, is built on it and runs today under a two-year Cooperative Research and Development Agreement with the Air Force Research Laboratory.

16 entries · 2005 — 2026 · newest on topClick a spine to open it
  1. 01 / 16Test · 2026 · Resolve

    Resolve is tested head to head against the Kalman filter, the standard for sixty years, in post-mission analysis on the same data. It produces the better track. Post-mission today; real time is on the roadmap.

    AuthorsPiLogic CitationPiLogic (2026). Resolve versus the Kalman filter in post-mission tracking analysis. Internal evaluation; figures not yet published.
  2. 02 / 16White paper · 2025 · PiLogic

    The technical paper on Resolve, PiLogic's exact probabilistic tracking model for radar. Post-mission analysis today, real time on the roadmap. The full paper is available on request.

    Sent on request · the full paper is not hosted online
    AuthorsPiLogic CitationPiLogic (2025). PiLogic Resolve: A Radar Tracking Engine. Technical white paper, November 2025. Available on request.
  3. 03 / 16White paper · 2025 · PiLogic

    The technical paper on Manifest, PiLogic's diagnostics model for satellite electrical power systems. Two paragraphs of it are here. The full paper is available on request.

    Sent on request · the full paper is not hosted online
    AuthorsPiLogic CitationPiLogic (2025). PiLogic Manifest: A Satellite EPS Diagnosis Engine. Technical white paper, October 2025. Available on request.
  4. 04 / 16Conference paper · 2021 · IJCAI-21

    An independent study by researchers with no connection to PiLogic or UCLA ran 18 methods across 672 benchmark problems. Ace, the academic engine built on the 2008 method, solved the most of any of them, and returned exact answers rather than estimates.

    Opens at IJCAI proceedings · not hosted by PiLogic
    AuthorsDurgesh Agrawal · Yash Pote · Kuldeep S. Meel CitationAgrawal, D., Pote, Y. & Meel, K. S. (2021). Partition Function Estimation: A Quantitative Study. In Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence (IJCAI-21), 4276–4285.
  5. 05 / 16Journal paper · 2010 · IEEE Trans. SMC-A

    The journal version of the NASA Ames work. A 503-node model of the ADAPT testbed, compiled into an arithmetic circuit, returned its most probable explanation in about a quarter of a millisecond on average, an order of magnitude faster than the alternatives tested.

    Opens at IEEE Xplore · not hosted by PiLogic
    AuthorsOle J. Mengshoel · Mark Chavira · Keith Cascio · Scott Poll · Adnan Darwiche · Serdar Uckun CitationMengshoel, O. J., Chavira, M., Cascio, K., Poll, S., Darwiche, A. & Uckun, S. (2010). Probabilistic Model-Based Diagnosis: An Electrical Power System Case Study. IEEE Transactions on Systems, Man, and Cybernetics, Part A: Systems and Humans, 40(5), 874–885.
  6. 06 / 16Conference paper · 2008 · IAAI-08

    NASA-funded work that ran the method on ADAPT, the electrical power system testbed at NASA Ames. A model of more than 400 nodes, compiled once, returned diagnoses in under a millisecond on average.

    Opens at AAAI (proceedings PDF) · not hosted by PiLogic
    AuthorsOle J. Mengshoel · Adnan Darwiche · Keith Cascio · Mark Chavira · Scott Poll · Serdar Uckun CitationMengshoel, O. J., Darwiche, A., Cascio, K., Chavira, M., Poll, S. & Uckun, S. (2008). Diagnosing Faults in Electrical Power Systems of Spacecraft and Aircraft. In Proceedings of the Twentieth Innovative Applications of Artificial Intelligence Conference (IAAI-08), 1699–1705. AAAI Press.
  7. 07 / 16Journal paper · 2008 · JSAT

    An analysis of Clone, a weighted Max-SAT solver from the same UCLA group, whose bounds come from compiling the formula so that the search runs in a reduced space.

    Opens at IOS Press · not hosted by PiLogic
    AuthorsKnot Pipatsrisawat · Akop Palyan · Mark Chavira · Arthur Choi · Adnan Darwiche CitationPipatsrisawat, K., Palyan, A., Chavira, M., Choi, A. & Darwiche, A. (2008). Solving Weighted Max-SAT Problems in a Reduced Search Space: A Performance Analysis. Journal on Satisfiability, Boolean Modeling and Computation, 4, 191–217.
  8. 08 / 16Journal paper · 2008 · Artificial Intelligence

    The standard reference for the technique the engine is built on. It turns a probabilistic question into a counting problem and compiles the model once, so exact answers come back where textbook methods run out of time and memory.

    Opens at eScholarship (UC open access) · not hosted by PiLogic
    AuthorsMark Chavira · Adnan Darwiche CitationChavira, M. & Darwiche, A. (2008). On probabilistic inference by weighted model counting. Artificial Intelligence, 172(6–7), 772–799.
  9. 09 / 16Conference paper · 2007 · UAI-07

    Recasts the mini-bucket approximation as exact inference on a simplified model made by splitting nodes, so that approximations can borrow every advance in exact inference.

    Opens at arXiv · not hosted by PiLogic
    AuthorsArthur Choi · Mark Chavira · Adnan Darwiche CitationChoi, A., Chavira, M. & Darwiche, A. (2007). Node Splitting: A Scheme for Generating Upper Bounds in Bayesian Networks. In Proceedings of the Twenty-Third Conference on Uncertainty in Artificial Intelligence (UAI 2007).
  10. 10 / 16PhD dissertation · 2007 · UCLA

    The doctoral thesis behind the 2008 method. Its title names the problem the rest of this record addresses: exact inference that is not bounded by treewidth, the structural limit that stalled the field.

    Citation only · a public copy has not been located
    AuthorsMark Chavira CitationChavira, M. (2007). Beyond Treewidth in Probabilistic Inference. Ph.D. dissertation, University of California, Los Angeles.
  11. 11 / 16Conference paper · 2007 · IJCAI-07

    A second way to compile a Bayesian network, built on variable elimination and algebraic decision diagrams, that exploits local structure far better than earlier elimination methods and can answer many queries at once.

    Opens at IJCAI proceedings · not hosted by PiLogic
    AuthorsMark Chavira · Adnan Darwiche CitationChavira, M. & Darwiche, A. (2007). Compiling Bayesian Networks Using Variable Elimination. In Proceedings of the Twentieth International Joint Conference on Artificial Intelligence (IJCAI-07), 2443–2449.
  12. 12 / 16Conference paper · 2006 · SAT-06

    One of the three papers the 2008 reference builds on. How a network is written into logic decides how well a model counter can break the problem into independent parts.

    Opens at Springer · not hosted by PiLogic
    AuthorsMark Chavira · Adnan Darwiche CitationChavira, M. & Darwiche, A. (2006). Encoding CNFs to Empower Component Analysis. In Theory and Applications of Satisfiability Testing, SAT 2006, Lecture Notes in Computer Science 4121, 61–74. Springer.
  13. 13 / 16Conference paper · 2006 · AAAI-06

    An exact algorithm for MAP, the hardest standard query in a Bayesian network, that is not limited by treewidth because its bounds are computed in linear time on a compiled arithmetic circuit.

    Opens at AAAI (proceedings PDF) · not hosted by PiLogic
    AuthorsJinbo Huang · Mark Chavira · Adnan Darwiche CitationHuang, J., Chavira, M. & Darwiche, A. (2006). Solving MAP Exactly by Searching on Compiled Arithmetic Circuits. In Proceedings of the Twenty-First National Conference on Artificial Intelligence (AAAI-06). AAAI Press.
  14. 14 / 16Journal paper · 2006 · Int. J. Approximate Reasoning

    A system for exact inference on relational Bayesian networks, compiled into arithmetic circuits and answered in time linear in the circuit's size, on instances with thousands of variables.

    Opens at ScienceDirect · not hosted by PiLogic
    AuthorsMark Chavira · Adnan Darwiche · Manfred Jaeger CitationChavira, M., Darwiche, A. & Jaeger, M. (2006). Compiling relational Bayesian networks for exact inference. International Journal of Approximate Reasoning, 42(1–2), 4–20.
  15. 15 / 16Conference paper · 2005 · IJCAI-05

    The paper that scaled compilation. Better logical encodings of a network's local structure cut compile times by an order of magnitude, compiled networks that had never been compiled before, and sped up online inference by orders of magnitude against jointree.

    Opens at IJCAI proceedings · not hosted by PiLogic
    AuthorsMark Chavira · Adnan Darwiche CitationChavira, M. & Darwiche, A. (2005). Compiling Bayesian Networks with Local Structure. In Proceedings of the Nineteenth International Joint Conference on Artificial Intelligence (IJCAI-05), 1306–1312.
  16. 16 / 16Conference paper · 2005 · UAI-05

    Defines what it means to compile a Bayesian network together with its evidence, and shows the payoff in maximum-likelihood estimation, sensitivity analysis and MAP, with results from genetic linkage analysis.

    Opens at arXiv · not hosted by PiLogic
    AuthorsMark Chavira · David Allen · Adnan Darwiche CitationChavira, M., Allen, D. & Darwiche, A. (2005). Exploiting Evidence in Probabilistic Inference. In Proceedings of the Twenty-First Conference on Uncertainty in Artificial Intelligence (UAI 2005), 112–119.

Don’t guess.
Compute.

Models Manifest Resolve
Company Company Careers
Resources Research News Contact
Compliance Privacy Policy Terms
PILOGIC Exact AI for aerospace
and defense.
© 2026 PiLogic · pilogic.ai