JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD B.Tech III ARTIFICIAL NEURAL NETWORKS Previous Question Paper (July - 2023 - R16)

Artificial Intelligence (AI) is a broad field of computer science focused on creating systems capable of performing tasks that typically require human intelligence, such as reasoning, problem-solving, and decision-making. In contrast, Neural Networks are a specific subset of machine learning inspired by the biological structure of the human brain, designed to recognize patterns and learn from data through interconnected nodes or neurons.

While AI encompasses a wide range of techniques including symbolic logic, expert systems, and search algorithms, neural networks represent a connectionist approach within AI that excels at handling complex, non-linear data like images, audio, and natural language.

Boltzmann learning is a stochastic learning rule used to train the weights of a Boltzmann machine, a type of recurrent neural network with symmetric connections. The primary goal is to adjust weights so that the probability distribution of the network's states matches the distribution of the training data.

The learning process consists of two phases:

1. Positive Phase: The visible units are clamped to the training data, and the network is allowed to reach thermal equilibrium. The correlation between the states of connected units is measured.
2. Negative Phase: The network runs freely without any input clamping. The correlation between the states of connected units is measured again.

The weight update rule is defined as:

Where η is the learning rate, p_ij⁺ is the probability that units i and j are both active during the positive phase, and p_ij^- is the probability that they are both active during the negative phase. This gradient-based approach effectively minimizes the Kullback-Leibler divergence between the model's distribution and the data distribution.