Is it a good idea to write a paper on new algorithms based on an existing algorithm?

Question

I am an undergraduate student interested in doing research.

There is a paper I have recently read. In the paper, a new (general) algorithm is proposed (in a bottom-Q1 journal). It is not a groundbreaking discovery. However, because the algorithm is very general, there are many possibilities to develop new algorithms based on that algorithm and to improve the efficiency of the algorithm.

Is it a good idea to write a paper on new algorithms based on that algorithm? It is because I have some ideas in my mind now. Can I submit the paper to a journal? (probably with low enough IF)

Answer 1

For your own personal development, yes, it is worth writing the paper. Ask a local professor to evaluate it and give you feedback on it. It might take a few rewrites before if feels "satisfactory". This is typical for academic writing.

It can be submitted to a journal, but would only be accepted if it is sufficiently "novel" to appeal to other mathematicians or mathematically inclined computer scientists. It is a fairly steep hurdle, but an experienced local professor can advise you there.

Note, however, that there are some journals dedicated to publishing student work. The need for a "wow" factor is a bit less there so it has a better chance of publication.

But writing the paper will help you firm and clarify the ideas. Learning, through practice, to be a clear writer in your field is a necessary step and the experience gives you some skill that will be important later. Have fun. Good luck.

---

Note also that almost all work in mathematics and CS is based on earlier work. The exceptions are rare.

Answer 2

Yes, that is a good idea

Extending existing research in a new direction, or in a way that adds some novel improvement, is a large part of what academic research does. What you are proposing is consistent with the types of research projects academics undertake that lead to publications. If these algorithm extensions are something you are interested in then it sounds like an enjoyable project that could potentially lead to publication.

If you pursue this project, make sure your paper gives the appropriate background to your problem and cites this other paper. You could consider targeting the same journal where this other paper was published in the first instance. Since you are only an undergraduate student, you might have some difficulty writing to the standard of work expected in academic journals, so you might also consider getting some preliminary feedback from an experienced academic in your field prior to submission.

Answer 3

Of course! Building on prior work is the way people do research.

If you look at the literature on (say) matrix or graph operations, you'll find plenty of work where people refine a general-purpose algorithm to exploit some special feature of the problem, like sparsity or symmetry. There's also a great deal of interest in adapting algorithms to fit different types of computation: parallel/distributed processing, resource-constrained environments, or branchless/cache-oblivious versions of an algorithm.

Without knowing the specific details of your problem and proposed solution, and field, it is hard to say what would be expected for a publication. This will also vary a great deal from venue to venue. In general, I think it is often easier to publish algorithmic advances that improve the asymptotic runtime (e.g., O(N²) to O(N log N)) or performance on a benchmark. Optimizations of a specific implementation via clever coding may be harder to publish as a CS paper, but might find a home in the "applied" literature if you really polish the implementation and build it into a library/toolkit.

There are plenty of exceptions—and pitfalls—so a conversation with an experienced researcher in your field would be a great next step. You may also want to consider whether this could be the basis for an Honors Thesis or Independent Study course, which would provide a more formal structure for advising.