Lab Projects on Neural Networks and Machine Learning

This repository contains four distinct lab projects, each focusing on different aspects of neural networks and machine learning techniques.

Lab 1: Perceptron Algorithm and Experimentation

• Objective: Introduction to the Perceptron algorithm and its application on synthetic data.
• Components:
  • Implementation of the Perceptron algorithm.
  • Time and space complexity analysis.
  • Synthetic data generation and visualization.
  • Performance evaluation through multiple experiments under varying conditions.

Lab 2: Binary and Multi-Class Classification with Neural Networks

• Objective: Solving classification problems using Fully Connected Neural Networks (FCNN).
• Components:
  • Data generation for binary and multi-class classification tasks.
  • Model creation with Keras and Scikit-learn.
  • Training and evaluation of models.
  • Visualization of decision boundaries for classification tasks.

Lab 3: MNIST Digit Classification using Convolutional Neural Networks

• Objective: Construction and training of Convolutional Neural Networks (CNNs) for digit classification.
• Components:
  • Loading and preprocessing of the MNIST dataset.
  • Implementation of three CNN architectures of varying complexity.
  • Training, evaluation, and visualization of model performance.
  • Saving and loading trained models for future use.

Lab 4: Application of Different Deep Neural Networks on Real-World Data

• Objective: Applying various types of deep neural networks on real-world data.
• Components:
  • Use of Dense Neural Network, Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM).
  • Utilization of Keras Embedding Layer and GloVe for word embedding.
  • Evaluation of the performance of various models (LSTM, GRU, RNN + CNN) using different word embedding techniques (GloVe, Word2Vec, FastText, TF-IDF).