Wall of Ideas

Preamble

The wall ideas is a digital version of the list on the whiteboard. It's a list of stray ideas that come up randomly when vacuuming or loading the dishwasher or during lab meeting. There's more ideas here than we can possibly explore, but it's a handy tool for those times when you think, "well, now what?"

The List

• Sensibility project (Jen, Erica)
  • What sensorimotor modalities are salient (i.e., are "sensible") when you think about different categories?
  • What makes them sensible? When answering this question, make sure your argument isn't circular!
• Classification Constrains Connectivity
  • We already showed the reverse.
  • The ABCD project data is going to sort of touch on this, but I wasn't planning on comparing a MCN vs. conventional connectivity
  • Would probably benefit from some synthetic or benchmark dataset where we have a reference outcome
• In what domains does connectivity help classification?
  • We showed the classifier was better when the autoencoder was present. Does this hold in other domains?
  • Thinking genetics. The animal folks have some data.
• Graph Neural Networks (Chandola)
  • Convolutional Neural Networks in the graph domain
• Randazzo Data
  • Classifiers applied to EEG data to diagnose HoH from controls
• Feature Production Stream of Consciousness
  • We have feature production data and imagery and fMRI data for some number of participants.
  • Is there a relationship between the probability of listing features of different types and their imagery?
• Bayes Category Sensitivity (Avery)
  • Using category affiliation scores for a series of classifier networks in a Monte Carlo design
  • Uses Horikawa data
• Feature Production to MRI Babelfish (Josh)
  • We have fMRI patterns and feature listings for the same concepts
  • Can we find a relationship between, e.g., listing many auditory features and regional fMRI activity?
  • Similar to the sensibility project, but on an individual differences level