A project for NMA_CN_2022 by Chaosfan group (Farnaz, Jing, Peng, Ula, Yidan, Yuyang)
Question:
- What's the difference of state transitions in different brain areas? (motor_ctx or vis_ctx)
- How are state transitions correlated across brain areas? Which areas lead/follow?
- How are state transitions changed across different trial types(correct/miss/passive)?
- Can we predict trial_types from hidden state sequences?
Scientific question
- What actually happened during information transfer from vis_ctx to motor_ctx?
- behaviour level:
- vis_stimuli(L/equal/R) ---- move_direction(L/No/R) ---- result/reward(correct/incorrect)
- observed level/firing rate level/decoder level:
- decoding accuracy (using firing rate in vis_ctx) better in vis_stimuli than move_direction
- decoding accuracy (using firing rate in motor_ctx) better in move_direction than vis_stimuli
- unobserved level/hidden states level:
- hidden states found in vis_ctx meaning vis_stimuli(L/equal/R)
- states across trials
- best num of hidden state
- statistical features of hidden states sequences
- hidden states found in motor_ctx meaning move_direction(L/No/R)
- states across time/trials
- best num of hidden state
- statistical features of hidden states sequences
- hidden states found in where meaning result/reward(correct/incorrect)
- hidden states found in vis_ctx meaning vis_stimuli(L/equal/R)
- decoder using hidden states in vis_ctx/motor_ctx
- ... (unfinished)
- behaviour level:
Dataset:
- Steinmetz dataset
Work plan:
7/26
- decoder using firing rate/hidden states in vis_ctx/motor_ctx Jing, Yidan
- hidden states across trials in vis_stimuli Farnaz, Peng
- statistical features of hidden states sequences in vis_ctx/motor_ctx Yuyang
- hidden states found in where meaning result/reward(correct/incorrect) Ula
7/25
- decoder using hidden state sequence to predict behaviour/trial_types Jing
- statistal features in different trial_types/different event_periods Yuyang
- !!! what the meaning of the hidden states
- wheel speed focus on motor Ula
- pupil area focus on vis Farnaz
- reward Yidan
- !!! organize our data/results/code together peng
- !!! logic of explaining our results
7/22
- hidden states across areas (vis, motor, thalamus) Ula
- hidden states across trial types Peng
- best num of hidden states Jing, Yidan
- cross-validation
- BIC
- analysis statistical features of hidden states sequences farnaz, Yuyang
7/15:
- HMM intro: Yuyang, Jing, Yidan
- Accessing data: Ula, Peng, Farnaz
To do list:
- explore how to use Steinmetz dataset, refer to dataset notebook
- organize data refer to this repo
- explore how to use PossionHMM to find hidden states, refer to ssm notebook; and how to analysis hidden states, refer to this one
- ssm packages
- find the hidden states across different areas
- focus on certain areas, eg, motor areas, visual cortex
- quantify the difference of state sequence across different areas
- transition time
- interval distribution
- quantify the time delay (lead/follow relation) across different areas
- find the fidden states across different trial_types
- quantify the difference of state sequence across trial_types
- transition time
- interval distribution
- quantify the difference of state sequence across trial_types
- construct a decoder for predicting trial_types from hidden state sequences
- compare the state sequence difference in different trial types across different brain areas
- to find which areas play a important role in defferent trials