Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Cox-Hawkes: doubly stochastic spatiotemporal Poisson processes
Authors: Xenia Miscouridou, Samir Bhatt, George Mohler, Seth Flaxman, Swapnil Mishra
TMLR 2023 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We show the efficacy and flexibility of our approach in experiments on simulated data and use our methods to uncover the trends in a dataset of reported crimes in the US. 5 Experiments We demonstrate the applicability of our methods on both simulated and real data. |
| Researcher Affiliation | Academia | Xenia Miscouridou EMAIL I-X and Department of Mathematics, Imperial College London, Samir Bhatt EMAIL Department of Public Health, University of Copenhagen, School of Public Health, Imperial College London, George Mohler EMAIL Department of Computer Science, Boston College, Seth Flaxman EMAIL Department of Computer Science, University of Oxford, Swapnil Mishra EMAIL Saw Swee Hock School of Public Health and Institute of Data Science, National University of Singapore and National University Health System. All listed affiliations are academic institutions. |
| Pseudocode | Yes | Algorithm 1 Cluster based generative algorithm for Hawkes process simulation. Algorithm 2 Simulation of the LGCP events from the background. |
| Open Source Code | Yes | The code for simulation and inference for this class of models of Cox-Hawkes processes implemented in python and numpyro (Phan et al., 2019) can be found at https://github.com/misxenia/Spatiotemporal_Cox_Hawkes. |
| Open Datasets | Yes | On real settings we apply our methods to gunfire data used in Loeffler & Flaxman (2018) detected by an acoustic gunshot locator system to uncover the underlying patterns of crime contagion in space and time. We use gunshot data in 2013 recorded by an acoustic gunshot locator system (AGLS) in Washington DC and follow Loeffler & Flaxman (2018) for data preprocessing. |
| Dataset Splits | Yes | The experimental setup is as follow. We simulate 100 datasets (each of which give on average 300 events) over a fixed time window and a fixed spatial domain and then do a train-test split. |
| Hardware Specification | No | No specific hardware details (e.g., exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running the experiments are provided. |
| Software Dependencies | No | The code for simulation and inference for this class of models of Cox-Hawkes processes implemented in python and numpyro (Phan et al., 2019) can be found at https://github.com/misxenia/Spatiotemporal_Cox_Hawkes. However, specific version numbers for python or numpyro are not provided. |
| Experiment Setup | Yes | For inference we run 3 chains with 1, 500 samples each of which 500 were discarded as burn in, using a thinning size of 1. In Figure 3 we report the trace plots for the parameters ฮฑ, ฮฒ, ฯ which define the triggering kernel that governs excitation. We also report a0 which we used as the total mean of the latent Gaussian process ยต(t, x, y) = exp (ft(t) + a0 + fs(x, y)). For Experiment 3: We used 2 chains each with 4, 000 samples from which 2000 are discarded as warmup. |