You are on page 1of 8

NEURAL

NETWORKS
PART-2

Shuraim. 🌟 Swipe
Terminologies in Neural Networks

Neuron/Node :- A The basic computational unit of


an ANN which processes input data and gives
output.
Input layer :- The initial layer of neurons that
recieves input data also called as features
Hidden layers :- Intermediate layers of neurons of
an ANN that come between input and output
layers
Output layer:- The final layer of neurons that
produces the network's output
Swipe
Backpropagation in neural network

Weights :- The parameters associated with


connections between neurons, determining the
strength of their influence

Bias :- An additional parameter associated with


each neuron, providing an offset to the weighted
sum of inputs

Backpropagation :- A training algorithm that


minimises the error during learning by adjusting
weights and biases

Swipe
Learning rate :- Determines the step size for weight
and bias updation during training

Gradient descent :- An optimization algorithm used


to update weights and minimize loss function

Loss function :- Measures the difference between


predicted and actual outputs and is used to guide
training process

Swipe
Activation function :- Is a function which introduces
non-linearity to neuron 's output enabling network
to model complex relationships in data.

Common activation functions are sigmoid, ReLU,


Linear and tanh and also softmax.

Swipe
Epoch :- One complete pass through entire training
dataset during training

Batch :- A subset of training data used for each forward


and backward pass during training

Swipe
Overfitting :- When an ANN performs well on training
dataset but does not perform well on unseen dataset due
to excessive complexity

Underfitting :- When an ANN is too simple to capture


underlying patterns in data

Swipe

You might also like