arXiv:2405.03386v2 Announce Type: replace Abstract: Training with noisy class labels impairs neural networks' generalization performance. In this context, mixup is a popular regularization technique to improve training robustness by making memorizing false class labels more difficult. However, mixup neglects that multiple annotators, e.g., crowdworkers, typically provide class labels. Therefore, we propose an extension of mixup, which handles multiple class labels per instance while considering which class label originates from which annotator. Integrated into our multi-annotator classificatio
The increasing reliance on large datasets and crowd-sourced annotations for training complex AI models makes robust learning from noisy labels a critical, immediate challenge.
Improving the robustness of AI models against noisy data directly impacts the reliability and performance of AI systems in various applications, enhancing their trustworthiness and efficacy.
This research provides a refined method for training neural networks with noisy class labels, particularly from multiple annotators, leading to more resilient and accurate AI models.
- · AI model developers
- · Companies using crowd-sourcing for data annotation
- · Industries relying on AI-powered classification
- · Organizations using less robust data annotation methods
AI models will become more resilient to imperfect training data.
The cost-effectiveness and scalability of crowd-sourced data annotation will improve as model robustness increases.
More reliable AI systems could lead to wider adoption in sensitive applications where data quality is a major concern.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG