Implementation of ML Models to Predict Movement Direction of a Rat Based on Neural Signals

Movement Direction:

• Left: -1
• Right: 1
• No Go: 0

1. Logistic Regression Model

• Accuracy: Achieves 93% for predicting left/right movement
• Limitation: Unable to predict "no go" due to the constraint of binary classification

2. Bayesian Convolutional Network

• Implementation:
  • Simplified the module wrapper in BBBConv2d from the BayesianCNN package
• Performance:
  • Achieved a true positive rate of 84%
  • Frequently predicts "no go" instead of left/right, leading to a conservative decision-making process
  • Results in a false negative rate of 38%