Fast Linear Interpolation for Piecewise-Linear Functions, GAMs, and Deep Lattice Networks
We present fast implementations of linear interpolation operators for both piecewise linear functions and multi-dimensional look-up tables. We use a compiler-based solution (using MLIR) for accelerating this family of workloads. On real-world multi-layer lattice models and a standard CPU, we show these strategies deliver $5-10\times$ faster runtimes compared to a C++ interpreter implementation that uses prior techniques, producing runtimes that are 1000s of times faster than TensorFlow 2.0 for single evaluations.
Code (0)
등록된 구현이 없습니다.
Tasks
CPUSimilar Papers 제목 키워드 기반
Gray level image enhancement using the Bernstein polynomials
This paper presents a method for enhancing the gray level images. This presented method takes part from the category of point operations and it is based on piecewise linear functions. The interpolation nodes of these fun…
Image EnhancementTightening the mixed integer linear formulation for the piecewise linear approximation in general dimensions
This paper addresses the problem of tightening the mixed-integer linear programming (MILP) formulation for continuous piecewise linear (CPWL) approximations of data sets in arbitrary dimensions. The MILP formulation leve…
The Quadratic Local Variance Gamma Model: an arbitrage-free interpolation of class C3 for option prices
This paper generalizes the local variance gamma model of Carr and Nadtochiy, to a piecewise quadratic local variance function. The formulation encompasses the piecewise linear Bachelier and piecewise linear Black local v…
Representing Piecewise Linear Functions by Functions with Small Arity
A piecewise linear function can be described in different forms: as an arbitrarily nested expression of $\min$- and $\max$-functions, as a difference of two convex piecewise linear functions, or as a linear combination o…
Representing Piecewise-Linear Functions by Functions with Minimal Arity
Any continuous piecewise-linear function $F\colon \mathbb{R}^{n}\to \mathbb{R}$ can be represented as a linear combination of $\max$ functions of at most $n+1$ affine-linear functions. In our previous paper [``Representi…