Re:Show HN: Matrix Multiplication with Half the Multiplications
by waynecochran on Friday March 15, 2024 @11:49AM
(#39717074)
The document said it outputs the exact same values as the conventional method. There is no accuracy trade off here.
[ Parent
| on HN ]
Re:Show HN: Matrix Multiplication with Half the Multiplications
by pclmulqdq on Friday March 15, 2024 @12:45PM
(#39717735)
The paper cited is about hardware, where there is no accuracy tradeoff because you control the numerical precision completely and use fixed point. In a software implementation, neither is true. There is no chance that you will get the exact same values out of this method that you do out of other FP matmuls.
Re:Show HN: Matrix Multiplication with Half the Multiplications
by p1esk on Friday March 15, 2024 @12:02PM
(#39717261)
For floating point? Are you sure?
Re:Show HN: Matrix Multiplication with Half the Multiplications
by waynecochran on Friday March 15, 2024 @01:49PM
(#39718591)
Opening statement of README
This repository contains the source code for ML hardware architectures that
require nearly half the number of multiplier units to achieve the same
performance, by executing alternative inner-product algorithms that trade
nearly half the multiplications for cheap low-bitwidth additions, while still
producing identical output as the conventional inner product.