A guide to experimental algorithmics by Catherine C. McGeoch

By Catherine C. McGeoch

"Computational experiments on algorithms can complement theoretical research by means of displaying what algorithms, implementations, and speed-up equipment paintings most sensible for particular machines or difficulties. This booklet publications the reader during the nuts and bolts of the most important experimental questions: What should still I degree? What inputs should still I try out? How do I research the knowledge? Answering those questions wishes principles from set of rules design Read more...

Show description

Read or Download A guide to experimental algorithmics PDF

Similar programming languages books

TCP/IP Tutorial and Technical Overview

The TCP/IP protocol suite has turn into the de facto common for desktop communications in ultra-modern networked global. the ever present implementation of a selected networking common has resulted in a massive dependence at the functions enabled through it. at the present time, we use the TCP/IP protocols and the web not just for leisure and knowledge, yet to behavior our enterprise by means of appearing transactions, trading items, and supplying prone to shoppers.

Sams teach yourself Cobol in 24 hours

Sams train your self COBOL in 24 Hours teaches the fundamentals of COBOL programming in 24 step by step classes. every one lesson builds at the past one offering a high-quality beginning in COBOL programming strategies and methods. Coupled with the resource code and the compiler to be had from Fujitsu, this hands-on advisor is the best, quickest strategy to commence growing ordinary COBOL compliant code.

CMMI for Development®: Guidelines for Process Integration and Product Improvement (3rd Edition) (SEI Series in Software Engineering)

CMMI® for improvement (CMMI-DEV) describes top practices for the advance and upkeep of goods and providers throughout their lifecycle. by means of integrating crucial our bodies of information, CMMI-DEV presents a unmarried, complete framework for enterprises to evaluate their improvement and upkeep approaches and increase functionality.

Additional resources for A guide to experimental algorithmics

Example text

This design would require 49, 600 = 24 × 10 × 10 × 31 design points, totaling 496,000 random trials. Assuming average runtimes around 100 seconds per trial (as reported by Culberson and Luo [12]) the experiment should finish in 574 days, or just more than 18 months. This design is still too big. Additional factor-reduction strategies are needed to get this experiment down to reasonable size. Here are some general tactics for shrinking experimental designs, illustrated with SIG. As always, pilot experiments can provide the information needed to exploit these ideas.

2 Choosing Factors and Design Points The motivating question in an algorithmic experiment typically falls into one of these four broad categories. 1. Assessment. These experiments look at general properties, relationships, and ranges of outcomes. Is there a performance bottleneck in Greedy? What are the range and distribution of color counts for a given input class? What input properties affect performance the most? 2. The horse race. This type of experiment looks for winners and losers in the space of implementation ideas.

The workhorse study comprises experiments built upon precisely stated problems: Estimate, to within 10 percent, the mean comparison costs for data structures A and B, on instances drawn randomly from input class C; bound the leading term of the (unknown) cost function F (n). Designs for workhorse experiments require some prior understanding of algorithm mechanisms and of the test environment. This understanding may be gleaned from pilot experiments; furthermore, a great deal of useful intelligence – which ideas work and do not work, which input classes are hard and easy, and what to expect from certain algorithms – may be found by consulting the experimental literature.

Download PDF sample

Rated 4.71 of 5 – based on 25 votes