Every Researcher once started his PhD program seeks picking the right research road. Googling for "How to find a good topic for a PhD research proposal?" should result in some important notes. I would like to add some notes & advises:
- Your supervisor is a critical concern her. Work with him where he has interest and updated with the field.
Friends Advises:
- Avoid problems with very top results (e.g. accuracy 90%). Pushing the performance is very challenging.
- Avoid very old unsolvable problems. Too many researchers already give a trial and performance is weak.
- Identify new hot problems ... potentially large unexplored space. To find them:
-- Attending a single conference can do the job
-- Talking with "different" researchers
-- Attending talks
-- Doing internships
For more success:
- Identify which scientific tools/ways recently push performance highly. In other words, the trend of current years. For example, right now many researchers use Deep Learning in Machine Learning Problems. 3 years ago researches heavily used SVM. Before that probabilistic approaches were the key.
Last update: 13 January, 2015.