Learning Interpretable Models for Coupled Networks Under Domain Constraints
Authors: Hongyuan You, Sikun Lin, Ambuj Singh10727-10736
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We validate our method on multishell diffusion and task-evoked f MRI datasets from the Human Connectome Project, leading to both important insights on structural backbones that support various types of task activities as well as general solutions to the study of coupled networks.We present two sets of experiments. First, to understand our method s strengths and limitations, we utilize simulated datasets that allow us to inspect both reconstruction performance and model selection performance. ... Second, we conduct experiments on the Human Connectome Project (HCP) data (Van Essen et al. 2013). |
| Researcher Affiliation | Academia | Hongyuan You, Sikun Lin, Ambuj K. Singh Department of Computer Science University of California, Santa Barbara 2120A Harold Frank Hall Santa Barbara, California 93106 {hyou, sikun, ambuj}@ucsb.edu |
| Pseudocode | Yes | Algorithm 1: CONCORD-MRCE; Algorithm 2: CC-MRCE; Algorithm 3: Constrained-CONCORD |
| Open Source Code | No | The paper mentions implementations of CGGM and MRCE provided by other researchers but does not state that its own source code is openly available. |
| Open Datasets | Yes | We apply the proposed framework on the Human Connectome Project (HCP) dataset (Van Essen et al. 2013) |
| Dataset Splits | Yes | We ran a 10-fold cross-validation of our model for each task, splitting data in a 9-1 train-validation ratio (46-5 split for the 51 subjects). We selected the optimal hyperparameters λ1 and λ2 by a 5 5 grid search in the log-scale between 10 1.6 to 10 0.4, keeping the models with smallest Mean Squared Error (MSE) percentage on the validation sets, averaged across 10 folds. |
| Hardware Specification | No | No specific hardware details (e.g., GPU/CPU models, memory, or cloud instance types) are provided for running the experiments. |
| Software Dependencies | No | The paper mentions methods like CONCORD, FISTA, MRCE, CGGM, and VAE, but does not provide specific version numbers for any software components or libraries used in their own implementation. |
| Experiment Setup | Yes | We selected the optimal hyperparameters λ1 and λ2 by a 5 5 grid search in the log-scale between 10 1.6 to 10 0.4, keeping the models with smallest Mean Squared Error (MSE) percentage on the validation sets, averaged across 10 folds. In our case, both λ1 and λ2 have optimal values around 0.1 across tasks. |