Stochastic Network Design in Bidirected Trees

Authors: xiaojian wu, Daniel R. Sheldon, Shlomo Zilberstein

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We compared RDP with a greedy algorithm and a state-of-the-art algorithm for conservation planning, which uses sample average approximation and mixed integer programming (SAA+MIP) [4, 6, 11]. We compared SAA+MIP, RDP and Greedy on small river networks. On large networks, we only compare RDP with the greedy algorithm because SAA+MIP fails to solve problems of that size. Figure 3: RDP vs Greedy on symmetric passabilities. Table 1: Comparison of SAA, RDP and Greedy. Time is in seconds. Each unit of expected reward is 107 (square meters). ER increase means the increase in expected reward after taking the computed policy.
Researcher Affiliation Academia 1 School of Computer Science, University of Massachusetts Amherst 2 Department of Computer Science, Mount Holyoke College
Pseudocode No The paper describes the dynamic programming algorithm and its rounded variant using equations and textual explanations, but it does not include a clearly labeled pseudocode block or algorithm steps formatted like code.
Open Source Code No No statement explicitly providing open-source code for the methodology described in this paper was found. There are no links to repositories or mentions of code being released.
Open Datasets Yes Our experiments use data from the CAPS project [20] for river networks in Massachusetts (Fig. 2). [20] Kevin Mc Garigal, Bradley W. Compton, Scott D. Jackson, Ethan Plunkett, Kasey Rolih, Theresa Portante, and Eduard Ene. Conservation assessment and prioritization system (CAPS). Technical report, Department of Environmental Conservation, Univ. of Massachusetts Amherst, 2011.
Dataset Splits No The paper does not explicitly provide details about training, validation, or test dataset splits. It mentions using data from the CAPS project and the Connecticut River watershed but does not specify how these datasets were partitioned for the experiments.
Hardware Specification No SAA+MIP used 20 samples for the sample average approximation and IBM CPLEX on 12 CPU cores to solve the integer program. This mentions the number of CPU cores but does not specify the CPU model, manufacturer, or other detailed hardware specifications.
Software Dependencies No SAA+MIP used 20 samples for the sample average approximation and IBM CPLEX on 12 CPU cores to solve the integer program. While CPLEX is mentioned, no version number is provided.
Experiment Setup Yes In our experiments, setting these numbers to be 50, 50 and 150 for ˆνu, ˆµu and ˆzu, the algorithm became very fast and we were able to get very good solution quality. For road-crossings, most passabilities start close to 1 and are cheap to repair relative to dams. To model this, we set Au,v ={a1}, puv|a1 =pvu|a1 =1.0 and cuv|a1 = 5. In contrast, it is difficult and expensive to remove dams, so multiple strategies must be considered to improve their passability. We created actions Au ={a1, a2, a3} with action a1 having puv|a1 =pvu|a1 =0.2 and cuv|a1 =20; action a2 having puv|a2 =pvu|a2 =0.5 and cuv|a2 =40; and action a3 having puv|a3 =pvu|a3 =1.0 and cuv|a3 =100.