Please enable JavaScript.

Coggle requires JavaScript to display documents.

Complex Spike Model - Coggle Diagram

- - - - Now, to present the information well, I created a plot of the signal using the matlab plot library then used a for loop to create an overlay of red highlighted windows that indicated if said windows contains complex spike behavior.
- - - - Load in label text file and LFP/SPK matlab files then create numpy array of the values for better processing within Python.
        
        In order to better process the large amount of data in accordance with the nature of CNN models, the data is split into windows.
        
        Using a For Loop, we can then break the signals up into windows approximately 62.5ms width with 31.25ms steps in between then reshape those windows in 1D columns that are then appended into an array.
        
        The above Data Prep is then repeated for the comparison dataset, which uses the next set of data as to allow the model to validate the model will work correctly outside of training.
        
        CNN Model Architecture
        
        The previous student Ethan used a sequential CNN model, which is more simple as it runs each step and layer one by one. Due to the nature of our data being separate, I opted to use a multi-input CNN model using the functional API in Keras
        
        2 more items...
        
        To start, we create the two input layers from the previously created LFP and SPK window arrays and ensure that they are in the proper 1D shape.
        
        In the very same for loop, we are also creating the labelled array "window_labels" that will later be used as the ground truth for the model.
        
        Knowing the samples are 60 sec long, the signal can be broken up into a designated amount of windows that will later determine our batch size per epoch.