First-Order Manifold Data Augmentation for Regression Learning

Authors: Ilya Kaufman, Omri Azencot

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate FOMA on in-distribution generalization and out-of-distribution robustness benchmarks, and we show that it improves the generalization of several neural architectures.
Researcher Affiliation Academia 1Department of Computer Science, Ben-Gurion University of the Negev, Beer-Sheva, Israel.
Pseudocode Yes We provide an example Py Torch pseudocode in Fig. 1 (left).
Open Source Code Yes Our code is publicly available at https://github.com/ azencot-group/FOMA
Open Datasets Yes We use the following five datasets to evaluate the performance of in-distribution generalization. Two tabular datasets: Airfoil Self-Noise (Airfoil) (Brooks et al., 2014) and NO2 (Aldrin, 2004). Two time series datasets: Exchange-Rate and Electricity (Lai et al., 2018)...
Dataset Splits Yes As per reference (Hwang & Whang, 2021), the training, validation, and test sets consist of 1003, 300, and 200 examples, respectively. (Airfoil dataset)
Hardware Specification Yes The results are obtained with a single RTX3090 GPU.
Software Dependencies No The paper uses PyTorch as implied by the pseudocode (torch.linalg.svd), but no specific version numbers for PyTorch or other software dependencies are provided in the text.
Experiment Setup Yes Detailed experimental settings and hyperparameters are provided in App. E. We list the hyperparameters for every dataset in Table 8 and Table 9 for the methods FOMA and FOMAρ, respectively.