CSE-6363-001-MACHINE LEARNING
Academic Projects @ UTA
Prof. ALEX JON DILLHOFF
Assmt -1 "Utilizing the Iris Dataset for Regression and Classification Analysis: Exploring Linear Regression Models and Discriminative Methods"
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/assignment1.pdf The assignment involves utilizing the Iris flower dataset for both regression and classification tasks. For regression, 12 linear regression models are created and trained using batch gradient descent with mean squared error as the loss function. Regularization effects are explored by comparing regularized and non-regularized models. For classification, Linear Discriminant Analysis, Logistic Regression, and Naive Bayes models are implemented and tested. Results are summarized in a report along with tables and plots illustrating model performance. The submission includes code, documentation, and instructions for reproducibility.
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/LDA.ipynb https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/Linear_Regression.ipynb https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/Logistic_regression.ipynb https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/Navie%20Bayes.ipynb https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/Report%20File_ML_Assmt1.pdf
Assmt -2 "Neural Network Library Implementation for XOR Problem and Handwritten Digit Recognition: Exploring Hyperparameters and Integrating Support Vector Machines"
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/assignment2.pdf This assignment involves building a neural network library and applying it to solve the XOR problem and recognize handwritten digits. You'll construct various neural network architectures, experiment with different hyperparameters, and evaluate model performance. Additionally, you'll explore Support Vector Machines (SVMs) in a forthcoming section. The submission includes code, a report detailing experiments and findings, and instructions for reproducibility.
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/ML_Assmt2.ipynb https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/ML_Assmt_2.pdf https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/x_test.npy https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/xor_network.pkl https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/y_test.npy https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/y_train.npy https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/commit/036fb4420faa0f155579ff2aca10a8125b63b76b
Assmt -3 "Enhancing Support Vector Machines: Negative Learning Rate, Non-linear Classification, and Multi-class Handling"
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/assignment3.pdf This assignment focuses on implementing Support Vector Machines (SVMs) with specific tasks:
Implementing SVMs when the learning rate (η) is negative. Adapting SVMs to support non-linear classification using polynomial kernels. Extending SVMs to handle multi-class classification using the One-versus-All approach. The implementation will be compared with sklearn.svm.SVC for accuracy and performance using various datasets. Submissions include code, a report summarizing findings, and instructions for reproducibility.
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/multiclass_SVM.ipynb https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/non_linear_SVM.ipynb https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/ML_Assignment%203.pdf
Assmt -4 "Gesture Recognition with Hidden Markov Models: Dataset Preparation, Model Building, and Evaluation"
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/assignment4.pdf
This assignment focuses on implementing Gesture Recognition models using Hidden Markov Models (HMMs). It involves preparing and normalizing the dataset, building HMM models for each gesture class, and evaluating them using test data. Additionally, a hyperparameter search is conducted, and the Forwards algorithm is implemented to compute the likelihood of input sequences. Visualizations of generated samples and discussion on their quality compared to real samples are also required. Submissions include code, a report detailing experiments and results, and instructions for reproducibility, packaged in a zip file for easy access and evaluation.
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/ML_Assmt_4.ipynb https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/Original_Dataset.csv https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/ML_Assmt_4.pdf
Assmt -5 "Deep Learning for Image Classification: Training and Evaluating Three Network Architectures on Food101 Dataset"
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/assignment5.pdf This assignment focuses on image classification with deep learning using the Food101 dataset. Three categories of networks are trained and evaluated:
Basic CNN: Implementing a basic convolutional neural network (CNN) with convolutional and fully connected layers. Train for up to 50 epochs, with optional early stopping to prevent overfitting.
All Convolutional Net: Creating an all convolutional model and comparing the total number of parameters with the basic CNN. Train for up to 50 epochs, with optional early stopping.
Regularization: Adding regularization in the form of data augmentation or dropout to one of the previous models. Train for up to 50 epochs and include final training and accuracy plots from TensorBoard in the report.
Transfer Learning: Fine-tuning a pre-trained model on the Food101 dataset. Describe the chosen pre-trained model, training plots, and final model accuracy in the report.
Bonus Round: Participants can compete in three categories for bonus points: training using a single epoch, building the smallest network, and creating the best all convolutional network.
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/ML_Assmt5.ipynb https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/ML_Assignment%205.pdf
ML Proj - https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/ML_Proj.ipynb
https://github.com/panchiwalashivani/CSE-6363-001-MACHINE-LEARNING/blob/main/ML_Proj.ipynb