Curvature Enhanced Data Augmentation for Regression

Authors: Ilya Kaufman, Omri Azencot

ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive evaluations across multiple datasets and comparisons with state-of-the-art methods demonstrate that CEMS delivers superior performance in both in-distribution and out-of-distribution scenarios, while introducing only minimal computational overhead.
Researcher Affiliation Academia 1Department of Computer Science, Ben-Gurion University of the Negev, Beer-Sheva, Israel. Correspondence to: Ilya Kaufman <EMAIL>.
Pseudocode Yes Pseudo-code and illustration of CEMS are given in Alg. 1 and Fig. 2.
Open Source Code Yes Code is available at https: //github.com/azencot-group/CEMS.
Open Datasets Yes Airfoil Self-Noise (Airfoil) (Brooks et al., 2014), containing aerodynamic and acoustic measurements of airfoil blade sections, and NO2 (Aldrin, 2004), which predicts air pollution levels at specific locations. We also use two time series datasets: Exchange-Rate and Electricity (Lai et al., 2018)...
Dataset Splits Yes The dataset is divided into 1003 training examples, 300 validation examples, and 200 test examples as noted in (Hwang & Whang, 2021).
Hardware Specification Yes The results are obtained with a single RTX3090 GPU.
Software Dependencies No The paper mentions "Our PyTorch implementation" but does not specify any version numbers for PyTorch or other software dependencies.
Experiment Setup Yes We present the hyperparameters for each dataset in Table 10. In our main results, we apply our method to the input space or the latent space, and we report the configuration with the best performance. All hyperparameters were selected through cross-validation and evaluated on the validation set.