Robust Influence Maximization for Hyperparametric Models
Authors: Dimitris Kalimeris, Gal Kaplun, Yaron Singer
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Additionally we validate our method empirically and prove that it outperforms the state-of-the-art robust influence maximization techniques. |
| Researcher Affiliation | Academia | 1Department of Computer Science, Harvard University, Cambridge, MA, USA. Correspondence to: Dimitris Kalimeris <kalimeris@g.harvard.edu>, Gal Kaplun <galkaplun@g.harvard.edu>. |
| Pseudocode | Yes | Algorithm 1 HIRO: Hyperparam Inf Robust Optimizer |
| Open Source Code | No | The paper does not provide any link or explicit statement about open-sourcing the code. |
| Open Datasets | No | The paper uses synthetically generated networks ('We generated four different synthetic networks...'), but does not provide any access information (link, DOI, citation) for a publicly available dataset. |
| Dataset Splits | No | The paper does not explicitly provide training/validation/test dataset splits (e.g., percentages, sample counts, or references to predefined splits) needed for reproduction. |
| Hardware Specification | No | The paper does not mention any specific hardware (e.g., GPU/CPU models, memory, or cloud instance types) used for running its experiments. |
| Software Dependencies | No | The paper does not list specific software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x, or specific solvers). |
| Experiment Setup | Yes | We used the sigmoid function as the hyperparameteric model to determine the diffusion probabilities, i.e. h(θ xe) = 1 1+exp( θ xe) as in (Kalimeris et al., 2018). We generated d random features in [ 1, 1] for every edge. We used d = 5, however our results are consistent across a large range of dimensions d and featuregenerating techniques, such as normal or uniform distributions over the unit hyper-cube [ 1, 1]d and it s discrete analog { 1, 1}d. We sampled Θϵ = {θ1, . . . , θl} from Θ = [ B, B]d and generated the family of influence functions Fϵ = {fi | θi Θϵ} for l = 20. In addition we set T = 10 HIRO iterations with the exception of Experiments 1 and 2, where l and T are the free variable, respectively. |