JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD B.Tech IV Year NEURAL NETWORKS AND DEEP LEARNING Previous Question Paper (September - 2022 - R16)

A perceptron is the simplest form of an artificial neural network, functioning as a linear binary classifier. It mimics the biological neuron by receiving multiple input signals, processing them through weighted connections, and producing a single binary output.

The working of a perceptron follows these steps:

1. Inputs x1, x2, ..., xn are multiplied by their respective weights w1, w2, ..., wn.
2. These weighted inputs are summed together with a bias term (b) to calculate the net input:
   z = Σ (xi * wi) + b
3. The net input is passed through an activation function, typically a step function, which determines the output:
   f(z) = 1 if z ≥ 0, else 0
4. If the predicted output does not match the target, the weights and bias are updated using the perceptron learning rule to minimize error.

A Hopfield network is a form of recurrent artificial neural network that serves as a content-addressable memory system. It is characterized by symmetric connections between neurons and a binary activation state (typically -1 or 1).

Key characteristics include:

• Architecture: All neurons are connected to every other neuron, except to themselves. The weight matrix is symmetric, meaning the connection from neuron i to j is equal to the connection from j to i.
• Energy Function: The network is defined by an energy function that always decreases or remains constant during state updates, ensuring the network converges to a stable state (a local minimum of the energy surface).
• Memory Storage: Patterns are stored as stable equilibrium points in the network. When a noisy or partial version of a stored pattern is provided as input, the network iteratively updates its states until it converges to the original, clean pattern.
• Limitation: The capacity of a Hopfield network to store patterns reliably is limited, typically defined as approximately 0.14 times the number of neurons in the network.