Deep Learning/Etc.

๋ฆฌ์„œ์น˜ ์•„์ด๋””์–ด๋ฅผ ์–ป๋Š” ๋ฒ•

๋ฐ˜์‘ํ˜•

์ถœ์ฒ˜ : https://twitter.com/jbhuang0604/status/1423499757591400448

 

Jia-Bin Huang on Twitter

“How to come up with research ideas? Excited about starting doing research but have no clue?๐Ÿคท‍โ™‚๏ธ๐Ÿคท๐Ÿป‍โ™€๏ธ Here are some simple methods that I found useful in identifying initial directions. Check out the thread below ๐Ÿ‘‡”

twitter.com

 

Find a different dimension

Just learn a cool idea from others?
Think about how you could extend it to another dimension.
Ex: Text / audio / image / video / graph

 

Relax assumptions

Identify the underlying assumptions of existing work and try relaxing them to make it work in more unconstrained settings.

 

Make more assumptions

Take a general approach and tailor it to your SPECIFIC problem.
You can then leverage all the domain knowledge (i.e., make more assumptions) to improve the method.

 

Combine two ideas/problems

"To steal ideas from one person is plagiarism.
To steal from many is research." - Wilson Mizner

 

Grab a powerful hammer and find all the nails

Pay attention to new emerging tools in the community. 
Apply and adapt them on your problem.

 

Add an adjective

Given an existing idea X, add an adjective to make it 
- slow → fast
- batch → online
- sensitive → robust
- centralized → distributed
- single-step → progressive
- single-level → hierarchical
- fixed → adaptive, sth-aware
- data-hungry → data-efficient

 

Stress test the state-of-the-art

Don't simply run on the fixed, boring benchmark datasets.
Try it out on diverse, unconstrained examples and see how it fails.
It's a great way to identify limitations of existing work.
This is where your work can fill the gap.

๋ฐ˜์‘ํ˜•