JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD B.Tech IV AI TECHNIQUES IN ELECTRICAL ENGINEERING Previous Question Paper (September - 2022 - R18)

Reinforcement Learning (RL) is a machine learning paradigm where an agent learns to make decisions by performing actions in an environment to maximize a cumulative reward signal. Unlike supervised learning, which relies on a labeled dataset of input-output pairs, or unsupervised learning, which seeks to identify hidden patterns in unlabeled data, RL is based on trial-and-error interaction.

Key differences include:

• Supervised Learning: The model is provided with the 'correct' answer for every input. It learns a mapping function from a static dataset.
• Unsupervised Learning: The model processes data without any labels or feedback, focusing on clustering or dimensionality reduction to find inherent structures.
• Reinforcement Learning: The agent receives delayed feedback in the form of rewards or penalties. It must balance exploration (trying new actions) and exploitation (choosing known high-reward actions) to optimize its long-term policy.

Competitive learning is a form of unsupervised neural network training where output units compete among themselves to be the one that responds to a given input pattern. Only the winning unit, often called the 'winner-take-all' neuron, is updated to become more similar to the input.

The paradigm typically involves the following steps:

1. Initialization: Weights of the output neurons are initialized with small random values.
2. Competition: For a given input vector, each output neuron calculates a similarity measure (e.g., Euclidean distance or dot product) with the input.
3. Selection: The neuron with the highest activation or smallest distance is declared the winner.
4. Adaptation: The weights of the winning neuron are adjusted to move closer to the input vector, effectively 'learning' that specific feature or cluster.

This process forces different neurons to specialize in different regions of the input space, making it highly effective for clustering and feature extraction tasks.