1 paper
Pradeep Singh, Mehak Sharma, Anupriya Dey +1
Transformers are the de-facto choice for sequence modelling, yet their quadratic self-attention and weak temporal bias can make long-range forecasting both expensive and brittle. W…