Valeriy M., PhD, MBA, CQF<p>🔍 The paper prove that even when data isn't exchangeable (like in time series), coverage loss can be bounded—so in many real-world cases, it doesn’t matter much! </p><p>These methods power tools like NeuralProphet and Nixtla, and now they’ve got solid math backing them up.</p><p><a href="https://sigmoid.social/tags/timeseries" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>timeseries</span></a> <a href="https://sigmoid.social/tags/forecasting" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>forecasting</span></a></p>