Search papers, labs, and topics across Lattice.
Ben-Gurion University of the Negev
2
0
4
A novel I/O-aware reformulation of wavelet convolution slashes memory usage and accelerates training speed, making it a game-changer for deep learning efficiency.
SoftDTW on GPUs just got a whole lot faster and memory-efficient, unlocking applications to longer sequence analysis thanks to a 98% memory reduction.
