JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD B.Tech IV Year ARTIFICIAL NEURAL NETWORKS Previous Question Paper (July/August - 2022 - R18)

A neural network is a computational model inspired by the structure and functional aspects of biological neural networks. It consists of an interconnected group of artificial neurons (nodes) organized in layers—typically an input layer, one or more hidden layers, and an output layer. Each connection has an associated weight that adjusts during the learning process to map input data to desired outputs.

The benefits of neural networks include:

• Non-linear Mapping: They can model complex, non-linear relationships between inputs and outputs that traditional statistical methods often fail to capture.
• Learning from Data: They possess the ability to learn from experience, automatically extracting features from raw data without requiring explicit programming for every rule.
• Fault Tolerance: Due to their distributed architecture, the network remains functional even if some neurons or connections are damaged or missing.
• Generalization: Once trained, they can provide accurate predictions or classifications for unseen data, demonstrating high adaptability.
• Parallel Processing: The inherent structure allows for simultaneous computation across multiple nodes, significantly increasing processing speed for large-scale tasks.

Neural network architectures are primarily classified based on the pattern of connections between neurons and the flow of information:

• Single-Layer Feedforward Network: The simplest form where input units are connected directly to output units. There are no hidden layers, and information flows in only one direction.
• Multi-Layer Feedforward Network: These contain one or more hidden layers between the input and output layers. The hidden layers enable the network to learn more complex patterns by extracting higher-level features.
• Recurrent Networks: Unlike feedforward networks, these include feedback loops where the output of a neuron is fed back as an input to the same or previous layers. This creates a form of internal memory, making them ideal for time-series data or sequential processing.
• Competitive Networks: In these architectures, neurons compete for the right to respond to a given input pattern. Only one output neuron or a small group of neurons is active at a time, which is commonly used for clustering and feature mapping (e.g., Self-Organizing Maps).