Causal Discovery in Hawkes Processes by Minimum Description Length
Authors: Amirkasra Jalaldoust, Kateřina Hlaváčková-Schindler, Claudia Plant6978-6987
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We compare our algorithm with the state-of-the-art baseline methods on synthetic and real-world financial data. The synthetic experiments demonstrate superiority of our method in causal graph discovery compared to the baseline methods with respect to the size of the data. |
| Researcher Affiliation | Academia | 1 Department of Computer Science, Columbia University, New York, USA 2 Department of Mathematical Science, Sharif University of Technology, Tehran, Iran 3 Faculty of Computer Science, University of Vienna, Vienna, Austria 4 Institute of Computer Science, Czech Academy of Sciences, Prague, Czech Republic 5 ds:Uni Vie, University of Vienna, Vienna, Austria |
| Pseudocode | Yes | Algorithm 1: Estimate COMP(Mγ; v), Algorithm 2: Estimate Li v(γi; x), Algorithm 3: Causal Discovery by MDL |
| Open Source Code | Yes | Our implementation and all experimental data are available at https://dm.cs.univie.ac. at/research/downloads/ and https://github.com/Amirkasraj/ Hawkes MDL |
| Open Datasets | Yes | Our implementation and all experimental data are available at https://dm.cs.univie.ac. at/research/downloads/ and https://github.com/Amirkasraj/ Hawkes MDL. We use daily return volatility of sovereign bonds of 7 large and developed economies called G-7 including USA, Germany, France, Japan, UK, Canada, Italy from 2003-2014 as in (Demirer et al. 2018). |
| Dataset Splits | No | The paper does not provide specific dataset split information such as exact percentages for training, validation, and test sets, nor does it reference predefined splits with citations that include such details. |
| Hardware Specification | Yes | MDLH is implemented in python, and the experiments are performed on a one Core Intel Xeon machine with 16 GB RAM. |
| Software Dependencies | No | The paper mentions 'python' and 'Tick package (Bacry et al. 2017)' but does not provide specific version numbers for either of them, which is required for reproducibility. |
| Experiment Setup | Yes | Due to the limitation of our computation resources, the number of MC simulations N was 1000 in our algorithm in all cases; We only consider luckiness function v 1 and uniform distribution π. For ML, LS, and NPHC we have penalties: L1 (lasso), L2, elastic net, and none. For ADM4, we also used the nuclear lasso ratio; For each of the above settings we also considered the lasso-nuclear-ratio taking value in {0, 0.1, 05, 0.9, 1}. |