ExperimentsMachineLearning

Subject: COMP472 Artificial Intelligence

Teacher: Leila Kosseim

Team Members

• Alejandro Leonardo García Navarro - alexgaarciia
• Lucía Cordero Sánchez - lucia-corsan
• Simon Dunand - SquintyG33Rs

URL to the repository (private)

• https://github.com/alexgaarciia/ExperimentsMachineLearning

Languages and software

• Language used; Python
• Done in Pycharm, but works for any other Python IDE

Project A1 - Penguins & Abalone

• Datasets: Penguins and Abalone datasets.

• Main tasks:

1. Load datasets and divide into train and test, see if balanced.
2. Train and test classifiers:
   • Decision Tree with default parameters.
   • Decision Tree with different choices of criterion, maximum depth and minimum samples split.
   • Multilayered Perceptron with 2 hidden layers, SGD (Stochastic Gradient Descent) and sigmoid activation function.
   • Previous MLP but experimenting with the activation functions, architecture and solver.
3. Statistics of the previous classifiers: confusion matrix, recall, F1 (these three for the classes) and accuracy, macro-average/weighted-average/variation F1 (for the model).
4. Analysis of performance.

• Files:

1. main: Contains the solution to all the exercises proposed in the instructions.
2. functions: Contains all the classifiers and functions to (a)compute some metrics, (b)print information, and (c)evaluate all the models.

• Instructions: run the "main" file to obtain all the plots and training/testing the classifiers.