Vladyslav Moroshan

Where to Begin: Efficient Pretraining via Sub-network Selection and Distillation

TempoPFN: Synthetic Pre-training of Linear RNNs for Zero-shot Time Series Forecasting