JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD B.Tech IV NEURAL NETWORKS AND DEEP LEARNING Previous Question Paper (June - 2024 - R16)

Artificial Neural Networks (ANN) are classified based on their architecture and data flow into several primary models:

• Single-Layer Perceptron: The simplest feedforward network consisting of a single layer of input nodes connected directly to an output node.
• Multi-Layer Perceptron (MLP): A feedforward network comprising an input layer, one or more hidden layers, and an output layer, capable of solving non-linearly separable problems.
• Recurrent Neural Networks (RNN): Networks that contain feedback loops, allowing information to persist and making them suitable for sequential data like time series or text.
• Convolutional Neural Networks (CNN): Specialized architectures designed primarily for processing grid-like data such as images, utilizing convolutional layers to automatically learn spatial hierarchies of features.

Supervised learning is a machine learning paradigm where a model learns a mapping from input data to output labels based on a provided training dataset. The process involves the following steps:

1. Data Preparation: The algorithm is provided with a labeled dataset consisting of input features and their corresponding correct target outputs.
2. Training Phase: The model processes the input data and makes predictions. It compares these predictions against the actual ground-truth labels using a loss function to quantify the error.
3. Optimization: An optimization algorithm, such as gradient descent, adjusts the model's internal parameters (weights and biases) to minimize the loss function, thereby improving the accuracy of future predictions.
4. Generalization: Once trained, the model is evaluated on unseen data to ensure it can accurately predict outputs for new, similar inputs rather than simply memorizing the training set.