RCAP: New Algorithm for Efficient Model Training Through Intelligent Data Selection
Researchers have developed RCAP, a new dynamic dataset pruning algorithm that reduces computational costs during machine learning model training while maintaining accuracy, particularly for imbalanced datasets. The method works by adaptively selecting representative data samples for each class based on their training loss, adjusting selections each epoch. This matters because it enables faster model training without sacrificing performance—achieving 8.69x speedup while using only 10% of data on imbalanced datasets.
RCAP (Robust, Class-Aware, Probabilistic dynamic dataset pruning) addresses a key challenge in machine learning: reducing training time without losing model accuracy. The algorithm uses a closed-form mathematical solution to determine what fraction of samples from each class should be included in training, then adaptively adjusts these fractions each epoch based on class-wise aggregated loss. It prioritizes samples with high loss values for inclusion in training subsets. The researchers tested RCAP across six diverse datasets ranging from balanced to highly imbalanced, using five different models and three training paradigms (training from scratch, transfer learning, and fine-tuning). Results show RCAP consistently outperformed existing dataset pruning methods, particularly excelling at maintaining worst-group accuracy at high pruning rates—a critical metric for ensuring fairness across underrepresented classes.
What different sources said
- arXiv cs.LGCenter
RCAP: Robust, Class-Aware, Probabilistic Dynamic Dataset Pruning
Related
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.
New Semi-Supervised Method Improves Single-Cell RNA Sequencing Integration Using Virtual Adversarial Training
Researchers introduced scCRAFT+, a semi-supervised integration method for single-cell RNA sequencing that uses Virtual Adversarial Training to incorporate marker gene information. The method addresses a key limitation of existing approaches: over-mixing of closely related cell subtypes during data integration. The advancement could improve cell type identification and biological interpretation in genomics research.