Research question

Iterate through the literature survey process by exploring the research gaps. You will find your research question when you find a gap that does not have a satisfactory answer, or by finding a solution/idea to answer the research gap in a limited context.

Tips

Can the question be answered:

  • using an appropriate table or graph?
    • type a: models as rows, datasets+metrics as columns
    • type b: graph of metrics with evident trends
  • without conducting any further experiments?
    • type a: 'is it possible', 'can we estimate'
    • type b: 'optimal model', 'most efficient method'
  • in a broad context?
    • type a: 'tested on 3 standard benchmarks'
    • type b: 'significantly outperformed state of the art on our completely unbiased small dataset'
    • type c: 'significantly outperformed by 0.01%'
    • type d: 'tested on publicly released dataset specifically altered by ...'

Examples

Bonus: Creating a graph of research questions

graph of research questions

Can we apply dropout to linear regression?


Define topics and express connections (using d2):

# Topics

est: Estimation of mechanical properties
shear: shear coefficients
thermal: thermal behaviour

regression: linear regression
networks: neural networks
transformers: transformers
dropout: dropout

# Connections

est -> shear
est -> thermal

shear -> regression
shear -> networks
shear -> transformers

thermal -> regression
thermal -> networks
thermal -> transformers

networks -> dropout
transformers -> dropout