Dynamic Optimization of Landscape Connectivity Embedding Spatial-Capture-Recapture Information
Authors: Yexiang Xue, Xiaojian Wu, Dana Morin, Bistra Dilkina, Angela Fuller, J. Royle, Carla Gomes
AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We show that our method scales to realworld size problems and dramatically outperforms the solution quality of an expectation maximization approach and a sample average approximation approach. We compare our algorithm against various algorithms, including the Expectation Maximization (EM) approach with 4 different initialization strategies, the Sample Average Approximation (SAA) and the greedy algorithm. Our approach and SAA are encoded and run as MIP without warm starting from any initial solutions. We first run our experiments on a set of small synthetic instances (25 parcels), where our all-pair DWC encoding (discussed in Section 3) scales, providing a nice lower bound on the objective function. We run experiments with different budgets. |
| Researcher Affiliation | Collaboration | Yexiang Xue Dept. of Computer Science Cornell University yexiang@cs.cornell.edu Xiaojian Wu Dept. of Computer Science Cornell University xw458@cornell.edu Dana Morin NY Coop. Fish & Wildlife Res. Unit Dept. of Natural Resources Cornell University djm466@cornell.edu Bistra Dilkina College of Computing Georgia Institute of Technology bdilkina@cc.gatech.edu Angela Fuller U.S. Geological Survey NY Coop. Fish & Wildlife Res. Unit Dept. of Natural Resources Cornell University akf34@cornell.edu J. Andrew Royle U.S. Geological Survey Patuxent Wildlife Research Center aroyle@usgs.gov Carla P. Gomes Dept. of Computer Science Cornell University gomes@cs.cornell.edu |
| Pseudocode | Yes | Algorithm 1: XOR K(w : X × Y {0, 1}, k, T) Sample T pair-wise independent hash functions h(1) k , h(2) k , . . . , h(T ) k : Y {0, 1}k; Solve the following problem max x∈X,y(i)∈Y i=1 w(x, y(i)) s.t. h(i) k (y(i)) = 0, i = 1, . . . , T. Return true if the max value is larger than T/2 , otherwise return false. Algorithm 2: XOR MAX COUNT(w : X × Y {0, 1}, T) k = log2 |Y|; while k > 0 do if XOR K(w, k, T) then Return 2k; end k ← k − 1; end Return 1; |
| Open Source Code | No | The paper does not provide a direct link to open-source code or explicitly state that the code is publicly available. |
| Open Datasets | No | The paper mentions using "small synthetic instances (25 parcels)" and "larger instances (225 parcels)" and "large scale (1024 parcel) realistic instances". However, it does not provide any concrete access information (link, citation, repository) for these datasets. |
| Dataset Splits | No | The paper mentions running experiments and comparing against other algorithms, but it does not specify any dataset splits (e.g., training, validation, test percentages or counts) or cross-validation setup. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the experiments (e.g., GPU models, CPU types, memory). |
| Software Dependencies | No | The paper mentions running algorithms and using solvers (MIP encoding), but it does not specify any software dependencies with version numbers (e.g., Python 3.x, CPLEX 12.x, Gurobi 9.x). |
| Experiment Setup | Yes | The number of segments g is set to 3 for all-pair DWC algorithm (og = 1/3), mainly due to scaling limitations. The Sample Average Approximation algorithm randomly samples 500 (s, t) pairs in DWC objective as its objective function, which is the largest number without exceeding the memory limit. The number of replications T in the XOR encoding is set to 11, based on empirical performance. We give all algorithms 4-hour time limit and 4GB of memory. |