Model-Based Diagnosis with Uncertain Observations

Authors: Dean Cazes, Meir Kalech2766-2773

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental evaluation shows that this third algorithm can be very effective in cases where the number of faults is small and the uncertainty over the observations is not large.
Researcher Affiliation Academia Dean Cazes, Meir Kalech Ben-Gurion University of the Negev, Israel deanc@post.bgu.ac.il, kalech@bgu.ac.il
Pseudocode Yes Algorithm 1 lists a pseudo code for O2D. Algorithm 2 lists a pseudo code for D2O. Algorithm 3 is an extension of O2D, called most likely diagnosis (Mst Like Diag), that returns the most probable diagnoses without finding all diagnoses.
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets Yes To evaluate our algorithms for the abductive diagnosis form, we used a system from the 74XXX benchmark Boolean circuit an ALU called 74182, having |O| = R = 5 outputs, SYSIN = 9 inputs and COMPS = 18 components. All observations were selected from Feldman et al. s (2010) known benchmark, a total of 250 observations, equally divided to five minimal cardinality values of 1-5. ...we used Barinel (Abreu, Zoeteweij, and van Gemund 2009) algorithm... Next, we evaluated our algorithms using real bugs from Apache Commons-Lang, an open-source software project.
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, or detailed splitting methodology) needed to reproduce the data partitioning.
Hardware Specification No The paper does not provide specific hardware details (GPU/CPU models, processor types, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper mentions using 'Barinel' algorithm but does not specify its version number or any other software dependencies with versions.
Experiment Setup Yes For each algorithm, O2D (Alg. 1), D2O (Alg. 2) and Mst Like Diag (Alg. 3), observation and value of P, we collected the average execution runtime. ... Then, we set a probability P for an observed output to be wrong. ... we created a combination of n components and m tests, where 7 n, m 13. For each combination we created 20 random matrices and error vectors having the combination properties. For each matrix file with m tests, we executed all three algorithms with a fixed number of u tests with uncertain observation over the result of the test, where 7 u m. In addition, the third algorithm was executed with different faulty output probabilities ([0.1, 0.5]).