Discovering Features with Synergistic Interactions in Multiple Views
Authors: Chohee Kim, Mihaela Van Der Schaar, Changhee Lee
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on synthetic, semi-synthetic, and real-world multiview datasets demonstrate that our model discovers relevant feature subsets with synergistic and non-synergistic interactions, achieving remarkable similarity to the ground truth. In this section, we evaluate the performance of Syn FS and multiple feature selection methods using synthetic, semi-synthetic, and real-world experiments. |
| Researcher Affiliation | Academia | 1Chung-Ang University, South Korea 2University of Cambridge, UK 3The Alan Turing Institute, UK. |
| Pseudocode | Yes | A.3. Pseudo-code of Syn FS Algorithm 1 Pseudo-code of Syn FS |
| Open Source Code | Yes | 2https://github.com/chohee K/Syn FS. |
| Open Datasets | Yes | We start by evaluating the synergistic and non-synergistic feature selection performance by utilizing a set of synthetic datasets employed in (Chen et al., 2018) and (Imrie et al., 2022) with known ground truth. We illustrate the interaction discovery performance of Syn FS on the MNIST (Le Cun et al., 1998) METABRIC (Pereira et al., 2016) is a multi-view dataset. 7,295 cancer cell lines collected by the Cancer Genome Atlas (TCGA)4. 4https://www.cancer.gov/tcga. purified populations of peripheral blood monocytes (PBMCs) using single-cell RNA (Zheng et al., 2017). |
| Dataset Splits | Yes | All the results are averaged over 10 random iterations of random 64/14/20 training/validation/testing splits. |
| Hardware Specification | No | The paper does not provide specific details on the hardware used, such as CPU or GPU models, memory, or specific cloud instance types. It only mentions general computing environments if at all (e.g. 'deep learning based methods'). |
| Software Dependencies | No | The paper mentions software like 'scikit-learn', 'ADAM optimizer', 'Re Lu activation', and provides GitHub links for 'STG' and 'Comp FS' but does not specify version numbers for these software dependencies (e.g., 'PyTorch 1.9', 'Python 3.8'). |
| Experiment Setup | Yes | Table 13: Hyperparameters for synthetic experiments. and Table 14: Hyperparameters for real-world datasets experiments. contain specific hyperparameter values for learning rates, hidden dimensions, epochs, and batch sizes. |