Obviously an algorithm that may get better results than existing algorithm is probably a good one. But how to evaluate an algorithm that solve a new practical problem (or subproblem)? There are no existing algorithms that can be compared. How to make an algorithm that solves a new practical problem to be accepted by a popular academic journal?
In the industry (utilization practices in health care) I work in, the evolution of legacy operational algorithms is a constant process. The environmental factors, in my instance, influencing the need to evolve an algorithm, is the triangulation between current best practices (evidence based medicine), demographic shifts, and epidemicologica priorities. Additionally, need to have clear, precise, research questions - what are the emerging questions one is trying to answer. The comparison between the legacy algorithm and the prototype being developed is weighed by the outcomes to these questions. What is the value/relevancy gap of the outcomes between the legacy and the prototype given the new enironmental factors.
I concur with Kirk's response. When I worked in commercial industry, I experienced similar concerns. However, some of our algorithms were proprietary for the purposes of logistics and routing of vehicles and parcels. In each case, there was always a specific purpose for algorithm development or research problem to address. As such, our algorithms emerged according to business needs and market demands. The down side -- since most all of our developments were proprietary, we were unable to have them published in journals.
- Badges
- Science topic
- Similar topics
- Academic Journals