/cse555-final-project

This is the final project for cse555 _ pattern recognition _ course at UB in fall 2021

Primary LanguageJupyter Notebook

cse555-final-project

dataset generation

we generate 6 sets of datasets from MNIST that consists of balanced dataset, imbalanced dataset, balanced dataset with symmetric noise, balanced dataset with asymmetric noise, imbalanced dataset with symmetric noise and imbalanced dataset with asymmetric noise.

baseline ML models

implementation of random forest and logistic regression models with evaluation metrics.

baseline DL models

implementation of LDAM-CRW model and Symmetric Cross Entropy loss

proposed models

Combination of UMAP with Random Forest as proposed ML model that improves RF baseline and implementation of focal loss within Symmetric Cross Entropy loss to improve DL baseline.