Efficient Symmetric Band Matrix-Matrix Multiplication on GPUs

Author(s):  
Ernesto Dufrechou ◽  
Pablo Ezzatti ◽  
Enrique S. Quintana-Ortí ◽  
Alfredo Remón
2018 ◽  
Vol 07 (02) ◽  
pp. 1850002
Author(s):  
Sheehan Olver ◽  
Andrew Swan

We prove that the Poisson/Gaudin–Mehta phase transition conjectured to occur when the bandwidth of an [Formula: see text] symmetric band matrix grows like [Formula: see text] is naturally observable in the rate of convergence of the level density to the Wigner semi-circle law. Specifically, we show for periodic and non-periodic band matrices the rate of convergence of the fourth moment of the level density is independent of the boundary conditions in the localized regime [Formula: see text] with a rate of [Formula: see text] for both cases, whereas in the delocalized regime [Formula: see text] where boundary effects become important, the rate of convergence for the two ensembles differs significantly, slowing to [Formula: see text] for non-periodic band matrices. Additionally, we examine the case of thick non-periodic band matrices [Formula: see text], showing that the fourth moment is maximally deviated from the Wigner semi-circle law when [Formula: see text], and provide numerical evidence that the eigenvector statistics also exhibit critical behavior at this point.


Author(s):  
Peter Benner ◽  
Alfredo Remon ◽  
Ernesto Dufrechou ◽  
Pablo Ezzatti ◽  
Enrique S. Quintana-Orti

1987 ◽  
Vol 4 (3) ◽  
pp. 239-258 ◽  
Author(s):  
Kam Hoi Cheng ◽  
Sartaj Sahni

Author(s):  
Yaniv Aspis ◽  
Krysia Broda ◽  
Alessandra Russo ◽  
Jorge Lobo

We introduce a novel approach for the computation of stable and supported models of normal logic programs in continuous vector spaces by a gradient-based search method. Specifically, the application of the immediate consequence operator of a program reduct can be computed in a vector space. To do this, Herbrand interpretations of a propositional program are embedded as 0-1 vectors in $\mathbb{R}^N$ and program reducts are represented as matrices in $\mathbb{R}^{N \times N}$. Using these representations we prove that the underlying semantics of a normal logic program is captured through matrix multiplication and a differentiable operation. As supported and stable models of a normal logic program can now be seen as fixed points in a continuous space, non-monotonic deduction can be performed using an optimisation process such as Newton's method. We report the results of several experiments using synthetically generated programs that demonstrate the feasibility of the approach and highlight how different parameter values can affect the behaviour of the system.


Sign in / Sign up

Export Citation Format

Share Document