Robust Submodular Maximization: A Non-Uniform Partitioning Approach
Authors: Ilija Bogunovic, Slobodan Mitrović, Jonathan Scarlett, Volkan Cevher
ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We numerically demonstrate the performance of PRO in data summarization and influence maximization, demonstrating gains over both the greedy algorithm and the algorithm of (Orlin et al., 2016).In addition to the above contributions, we provide the first empirical study beyond what is demonstrated for = 1 in (Krause et al., 2008). We demonstrate several scenarios in which our algorithm outperforms both the greedy algorithm and the algorithm of (Orlin et al., 2016). |
| Researcher Affiliation | Academia | 1LIONS, EPFL, Switzerland 2LTHC, EPFL, Switzerland. |
| Pseudocode | Yes | Algorithm 1 Partitioned Robust Submodular optimization algorithm (PRO) |
| Open Source Code | No | The paper does not contain any statement or link regarding the public availability of source code for the described methodology. |
| Open Datasets | Yes | EGO-FACEBOOK and EGO-TWITTER were used previously in (Mcauley & Leskovec, 2014).TINY10K and TINY50K: We used two Tiny Images data sets of size 10k and 50k consisting of images each represented as a 3072-dimensional vector (Torralba et al., 2008).CM-MOLECULES: This dataset consists of 7211 small organic molecules, each represented as a 276 dimensional vector. Each vector is obtained by processing the molecule s Coulomb matrix representation (Rupp, 2015). |
| Dataset Splits | No | The paper discusses evaluation against worst-case removals but does not provide specific details on how the datasets were split into training, validation, and test sets for model development or evaluation. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used to conduct the experiments. |
| Software Dependencies | No | The paper mentions using a 'fast C++ implementation of branch-and-bound' and 'LAZYGREEDY implementation given in (Minoux, 1978)' but does not specify version numbers for these or other software dependencies. |
| Experiment Setup | Yes | In our experiments, the size of the robust part of the solution set (i.e., |S0|) is set to 2 and log for OSU and PRO, respectively. That is, we set = 1 in PRO, and similarly ignore constant and logarithmic factors in OSU, since both appear to be unnecessary in practice. |