GaSPing for Utility

Authors: Mengyang Gu, Debarun Bhattacharjya, Dharmashankar Subramanian2637-2644

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

Reproducibility Variable Result LLM Response
Research Type Experimental Through extensive simulation experiments as well as two real datasets from management science, we demonstrate that the proposed approach results in better function fitting. We conduct experiments with synthetic data, and perhaps more importantly, with real-world data from the management science literature.
Researcher Affiliation Collaboration Department of Statistics and Applied Probability, University of California, Santa Barbara, CA, USA Research AI, IBM T. J. Watson Research Center, Yorktown Heights, NY, USA
Pseudocode No The paper describes mathematical models and derivations but does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes using the default setting in the Robust Ga SP R Package (Gu, Palomo, and Berger 2019).
Open Datasets Yes a real dataset from Abdellaoui, Bleichrodt, and Paraschiv (2007), collected from a prospect theory based scheme. ... To demonstrate the performance of the Ga SP model for utility functions with multiple attributes, we study a real dataset with three attributes (Fischer, Jia, and Luce 2000).
Dataset Splits No The paper describes training and test sets (e.g., 'remaining assessed tuples are used as the training set' and 'saving them as the test set') but does not explicitly mention or detail a separate validation set for model tuning or selection.
Hardware Specification No The paper does not provide any specific details regarding the hardware (e.g., CPU, GPU models, memory, or cloud computing resources) used to conduct the experiments.
Software Dependencies No The paper refers to the 'Robust Ga SP R Package (Gu, Palomo, and Berger 2019)' but does not explicitly state its version number or any other software dependencies with their specific versions.
Experiment Setup Yes Out of sample mean squared error MSE = n i=1 {ˆu(x i ) u(x i )}2/n is utilized for comparison, where x i X is the ith equally spaced held-out point and n = 1, 001 is used for testing throughout this section. We assume U(xmin) = 0 and U(xmax) = 1, where xmin = 0 and xmax = 105 are lower and upper bounds of X in simulated studies. For the QPD, we choose the basis function to be g1(u) = 1, g2(u) = Φ 1 (u), g3(u) = uΦ 1 (u), g4(u) = u. ... we randomly sample n loss = 4 and n gain = 3 assessed tuples in the loss and gain domains for each person respectively, saving them as the test set, while the remaining assessed tuples are used as the training set.