Main discussion thread

[gsaurabhr]

Starting off on this project

• [x] Read some literature
  • [x] #2
  • [x] #3
• [x] #4
• [x] #5
• [x] #7
• [ ] #10

[Tanvig]

Key points from the meeting on 30th May:

• Complete the visualization part.

• Try to find the code implemented in the paper.

• Find out using domain knowledge what frequency range we should use for our analysis (original sampling frequency is 1000 Hz). For the time being, we have decided on 100 Hz.

• Remove the line noise (it was around 60 Hz).

• Keep the epoch size/window to at least 20 sec (since the transitions are spread out and a longer temporal context is required). In such case, our temporal dimension would be 20 sec * 100 Hz = 2000 frames

• Find the appropriate dimension size for preserving the spatial distribution of the electrodes.

• Start with the R(2+1)D model for the implementation part. For the time being, we can ignore the EOG channels.

[Ayush-Tibrewal]

https://github.com/akaraspt/deepsleepnet - 2nd paper code , we will start implementing it.