This repository contains the complete set of assignments submitted for the Statistical Methods in AI (SMAI) course during Spring 2025 at IIIT Hyderabad. The assignments demonstrate implementation from scratch, use of deep learning frameworks, and in-depth analysis of models.
- Topics Covered:
- Linear Regression (Batch, Mini-Batch, Stochastic Gradient Descent)
- Regularization (Ridge, Lasso)
- KNN and ANN for classification and retrieval
- Decision Tree (custom and sklearn-based)
- K-Means and SLIC-based image segmentation
- Topics Covered:
- Multi-Layer Perceptron (MLP): Classification, Regression, and Multi-label tasks
- Gaussian Mixture Models (GMM) for medical image segmentation
- Principal Component Analysis (PCA) for dimensionality reduction
- Autoencoder and Variational Autoencoder (VAE) using PyTorch
- Topics Covered:
- CNN from scratch for facial image-based age regression
- Fine-tuned ResNet-18 for performance comparison
- Dataset preprocessing (UTKFace), training, and evaluation
- Python (Jupyter Notebooks)
- NumPy, Pandas, Matplotlib, Seaborn
- scikit-learn, OpenCV, scikit-image
- PyTorch, Torchvision
- ITK-SNAP (for medical image analysis)
- Regression: MSE, RMSE, R²
- Classification: Accuracy, Precision, Recall, F1-score, ROC-AUC
- Retrieval: MRR, Precision@100, Hit Rate
- Segmentation: Visual inspection and clustering performance
- All implementations are structured, documented, and annotated within respective notebooks.
- PyTorch was used where allowed (Autoencoders, CNNs, ResNet).
- Visualizations and observations are embedded for clarity and reproducibility.
Implemented by: P. Jakeer Hussain
Institute: IIIT Hyderabad
Course: Statistical Methods in AI (Spring 2025)