CS6375

Project 2 for CS 6375 - Machine Learning

Files:

• gh2.py (Provided by Professor) - Original file provided by the professor.
• our_mnist.py (Provided by Professor) - Modified MNIST data to only use 1500 pieces of data.
• script1.py - Modified gh2.py to use functions to dynamically create layers. Contains final parameters submitted for project.
• script1_checkLayerShapes.py - Given fixed parameters, iterate over different number of layers and number of nodes in each layer to determine optimal sizes of each for this specific problem.
• script1_automated.py - Iterate over different ranges of values for all hyperparameters and identify the optimal values for each hyperparameter.
• script2.py - Modified gh2.py to use functions to dynamically create layers. Contains final parameters submitted for project.
• script2_AdamcheckLayerShapes.py - Given fixed parameters, iterate over different number of layers and number of nodes in each layer to determine optimal sizes of each for this specific problem.
• script2_Adamautomated.py - Iterate over different ranges of values for all hyperparameters and identify the optimal values for each hyperparameter.

Sample Output:

• OutputScript1AverageParametersTop5.xls - The top 5 average accuracies with corresponding parameters I achieved for Script 1 (Gradient Descent Optimizer). Average accuracies measured by running each configuration of hyperparameters 3 times and taking the average test accuracy.
• OutputScript1BestParametersTop5.xls - The top 5 best accuracies with corresponding parameters I achieved for Script 1 (Gradient Descent Optimizer).
• OutputScript2AverageParametersTop5.xls - The top 5 average accuracies with corresponding parameters I achieved for Script 2 (Adam Optimizer). Average accuracies measured by running each configuration of hyperparameters 3 times and taking the average test accuracy.
• OutputScript2BestParametersTop5.xls - The top 5 best accuracies with corresponding parameters I achieved for Script 2 (Adam Optimizer).