Variational Policy for Guiding Point Processes

Authors: Yichen Wang, Grady Williams, Evangelos Theodorou, Le Song

ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on synthetic and real-world data show that our algorithm can steer the user activities much more accurately and efficiently than other stochastic control methods.
Researcher Affiliation Academia 1College of Computing, Georgia Tech, Atlanta, GA, USA 2School of Aerospace Engineering, Georgia Tech. Correspondence to: Yichen Wang <yichen.wang@gatech.edu>.
Pseudocode Yes Algorithm 1 KL Model Predictive Control
Open Source Code No The paper does not provide concrete access to source code, nor does it explicitly state that the code will be made open source.
Open Datasets Yes We evaluate on a real-world Twitter dataset (Farajtabar et al., 2015)
Dataset Splits No We partition the data into ten intervals and use one interval for training and others for testing.
Hardware Specification No The paper does not provide specific hardware details such as CPU or GPU models used for running experiments.
Software Dependencies No The paper mentions methods like 'Euler forward method' and 'thinning algorithm' but does not specify any software dependencies with version numbers (e.g., programming languages, libraries, or frameworks).
Experiment Setup Yes For MPC, we set the optimization window T = T/10 and sample size I = 10, 000. ... The time window is divided into 500 timestamps. We set the initial opinion xi(0) = 10 and the target opinion ai = 1 for each user. For model parameters, we set β = 0.2, and adjacency matrix A generated uniformly on [0, 0.01] with sparsity of 0.001. We set the base intensity in (1) to be µ = 0.01; the influence matrix is the same as the adjacency matrix A. We set the cost tradeoff parameter to be γ = 10.