https://nalar.dev/tags/training-data/ Recent content in Training Data on Nalar Hugo en-us Sat, 05 Sep 2026 00:00:00 +0700 https://nalar.dev/handle-label-noise-in-supervised-learning/ Sat, 05 Sep 2026 00:00:00 +0700 https://nalar.dev/handle-label-noise-in-supervised-learning/ <p>Supervised learning assumes that training examples come with useful target labels. Real datasets rarely satisfy that assumption perfectly. A support ticket may be assigned to the wrong queue, an image may receive the wrong class, or two annotators may interpret an ambiguous policy differently.</p> <p>These errors create <strong>label noise</strong>: the recorded target does not reliably represent the target the model is supposed to learn. Enough noise can teach a model contradictory patterns, distort evaluation, and make apparently difficult modeling problems into data-quality problems.</p>