Delivering Guaranteed Display Ads under Reach and Frequency Requirements
Authors: Ali Hojjat, John Turner, Suleyman Cetintas, Jian Yang
AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our numerical tests suggest that with parallelization of the pattern generation process, the algorithm has a promising run time and memory usage. ... We tested the algorithm on randomly-generated graphs that we constructed in such a fashion to resemble appropriately-scaled versions of real-world instances. For example, Figure 3 demonstrates the progress of the algorithm on a small graph with 40 demand nodes and 300 supply nodes. |
| Researcher Affiliation | Collaboration | Ali Hojjat and John Turner Paul Merage School of Business University of California Irvine hojjats@uci.edu, john.turner@uci.edu Suleyman Cetintas and Jian Yang Advertising Sciences Group Yahoo Labs, Sunnyvale, CA cetintas@yahoo-inc.com, jianyang@yahoo-inc.com |
| Pseudocode | No | The paper describes algorithms and formulations mathematically and in text, but does not provide pseudocode or a clearly labeled algorithm block. |
| Open Source Code | No | The paper does not provide any explicit statements about open-sourcing the code or links to a code repository for the described methodology. |
| Open Datasets | No | We tested the algorithm on randomly-generated graphs that we constructed in such a fashion to resemble appropriately-scaled versions of real-world instances. |
| Dataset Splits | No | The paper describes using 'randomly-generated graphs' but does not specify any train/validation/test dataset splits. It describes the characteristics of these generated graphs (e.g., '40 demand nodes and 300 supply nodes', '3 subgroups with guaranteed visit lengths of {10, 20, 30} impressions'). |
| Hardware Specification | Yes | We used the AMPL modeling language with CPLEX solver on a dual core i5 2.5GHz CPU with 8GB of RAM to carry out the experiments. |
| Software Dependencies | No | We used the AMPL modeling language with CPLEX solver. The paper does not provide specific version numbers for AMPL or CPLEX. |
| Experiment Setup | Yes | We tested the algorithm on randomly-generated graphs that we constructed in such a fashion to resemble appropriately-scaled versions of real-world instances. For example, Figure 3 demonstrates the progress of the algorithm on a small graph with 40 demand nodes and 300 supply nodes. Each supply node was further partitioned into 3 subgroups with guaranteed visit lengths of {10, 20, 30} impressions. There are approximately 4600 arcs in the graph. ... In between the epochs, we solve the subproblems until (at most) 20 improving patterns were found. ... We solved the subproblems in an ad-hoc (essentially random) order. We used a diversity-seeking metric of the form (7), and did not use a pacing metric (i.e. πp( ) = 0). |