“…The burgeoning field of machine learning led to the rapid availability of new surrogate modeling paradigms in the integrated circuit design and signal integrity analysis [3], [4], [5], including for example support-vector machines [6], [7], [8], [9], Gaussian processes [10], [11], [12], neural networks [13], [14], and reinforcement learning [15], [16]. Alternatively, the method of polynomial chaos expansion also became widely popular for UQ [17], [18], [19], [20], [21].…”