2

Many research journals and Ph.D. thesis has the requirement of originality for the content.

According to the dictionary, the meaning for the word originality is as follows

The ability to think independently and creatively.

The quality of being novel or unusual.

Suppose there exist two algorithms a and b. And if I get an idea of mixing both of them and obtaining a new algorithm named c that is useful.

I can, then, say that algorithm c has the property of usefulness. But I have doubts whether algorithm c has the property of originality since it is a mixture of two non-original algorithms. The originality maybe just be in the idea of mixing the both which many people might already think.

Under which category does the mixing of multiple algorithms for obtaining a new algorithm belong? Is it independent, creative, novel, or unusual?

hanugm
  • 7,779
  • 9
  • 43
  • 79
  • 1
    It could be novel or not. It is impossible to say in general. Einstein's theory of special relativity is novel, but based on earlier work as far back as Galileo. But mixing algorithms could also be trivial and obvious. – Buffy Oct 21 '21 at 14:37
  • Questions about CS as a research field aren't on topic. – Azor Ahai -him- Oct 21 '21 at 17:55

2 Answers2

2

It certainly can be.

A random forest algorithm is an ensemble model which is an extension and mixing of decision tree algorithms along with a statistical bagging algorithm. When first developed, it was an original idea by any sensible definition of the word, and most certainly an idea worthy of publication.

Many methods in machine learning are designed around things that are greater than the sum of their parts -- that combining a bunch of weak learners (algos that are only slightly better than random chance) can make something that is a strong learner (an algo that has an arbitrarily high level of accuracy). These new models are original, novel, useful, etc.

https://en.wikipedia.org/wiki/Bootstrap_aggregating https://en.wikipedia.org/wiki/Decision_tree_learning
https://en.wikipedia.org/wiki/Random_forest

eps
  • 1,093
  • 7
  • 11
1

There is no telling whether it is novel enough based solely on the info you have provided.

I would describe the required quality like so: the result, as a whole, must solve some identifiable problem better than it was achieved previously. To provide a couple examples:

  • In number theory, if you use an algorithm to produce a sequence of numbers and call it a day, this is probably not very useful. If this sequence of numbers has some properties allowing to attack some well-known conjecture, this could be absolutely groundbreaking.
  • In materials science, if you find a novel way to paint rubber balls some indistinct olive, that would probably also not be useful, unless you could specifically say why are indistinct olive rubber balls great. Whoever came up with the Tyrian purple was likely hailed as one of the greatest researchers/engineers of his age.
Lodinn
  • 8,379
  • 11
  • 45