Machine-Learning-Projects

Here are many machine learning projects ranging from classical models to Neural Networks

1. Judging a Book by its Cover

• Build Neural Network Architectures ,CNNs and RNNs on image(Book cover image) and text(Book title) dataset of Books, to predict the book genre amongst 30 possible categories
• Implemented own CNNs and RNNs models using PyTorch ,for this classification task.
• Used BERT (for text dataset) and EfficientNetB4 (for image dataset).For this Multimodal,combined (by Stacking) probabilities obtained from these models and then used SVM Classifier to get accuracy of 63.4%

Note: This project was created collaboratively by Lalit Meena and Prakul Virdi during COL774, Machine Learning course(Fall 2022,Prof. Parag Singla).

2. Image Classification

• Used Support Vector Machines (SVMs) to build models for binary classification and Multi-Class Classification (one-vs-one classifier setting) for CIFAR-10 image dataset
• Solved this SVM optimization problem using a general purpose convex optimization package(CVXOPT) as well as using a scikit-learn library function (based on LIBSVM)
• Compared their accuracy and training time for solving this SVM dual problem using Linear & Gaussian Kernel.Accuracy obtained- 85.6% (Binary ) & 62.44% (Multi-Class Classification)