Accurately computing the log-sum-exp and softmax functions

نویسندگان
چکیده

برای دانلود باید عضویت طلایی داشته باشید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Accurately Computing log(1− exp(− |a|)) Assessed by the Rmpfr package

In this note, we explain how f(a) = log(1− e−a) = log(1− exp(−a)) can be computed accurately, in a simple and optimal manner, building on the two related auxiliary functions log1p(x) (= log(1 + x)) and expm1(x) (= exp(x)− 1 = ex − 1). The cutoff, a0, in use in R since 2004, is shown to be optimal both theoretically and empirically, using Rmpfr high precision arithmetic. As an aside, we also sho...

متن کامل

Growth Estimates for Exp-Log Functions

Exp-log functions are those obtained from the constant 1 and the variable X by means of arithmetic operations and the function symbols exp() and logll. This paper gives an explicit algorithm for determining eventual dominance of these functions modulo an oracle for deciding zero equivalence of constant terms. This also provides another proof that the dominance problem for exp-log functions is T...

متن کامل

Global optimal solutions to nonconvex optimisation problems with a sum of double-well and log-sum-exp functions

This paper presents a canonical dual approach for solving a nonconvex global optimisation problem with a sum of double-well and log-sum-exp functions. Such a problem arises extensively in mechanics, robot designing, information theory and network communication systems. It includes fourth-order polynomial minimisation problems and minimax problems. Based on the canonical duality theory, this non...

متن کامل

Twofold exp and log

This article is about twofold arithmetic [1, 2]. Here I introduce algorithms and experimental code for twofold variant of C/C++ standard functions exp() and log(), and expm1() and log1p(). Twofold function y0 + y1 ≈ f(x0 + x1) is nearly 2x-precise so can assess accuracy of standard one. Performance allows assessing on-fly: twofold texp() over double is ~10x times faster than expq() by GNU quadm...

متن کامل

Softmax-Margin CRFs: Training Log-Linear Models with Cost Functions

We describe a method of incorporating taskspecific cost functions into standard conditional log-likelihood (CLL) training of linear structured prediction models. Recently introduced in the speech recognition community, we describe the method generally for structured models, highlight connections to CLL and max-margin learning for structured prediction (Taskar et al., 2003), and show that the me...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: IMA Journal of Numerical Analysis

سال: 2020

ISSN: 0272-4979,1464-3642

DOI: 10.1093/imanum/draa038