Gen-ai Fundamentals: Neural Networks

Hello friends,

I am starting a series on Gen-AI fundamentals. The goal is to present complex concepts into simple and easy to understand way. Today we shall cover basics of neural networks along with their key types such as Feedforward, CNN and RNN.

Before we plunge into details, let’s first understand what is the neural network. It’s also called as Artificial Neural Network (ANN)

Neural network representation (image source: wikipedia)

Artificial Neural Network (ANN) :

• In simple terms, neural network is a machine-learning model, inspired by the human brain, that processes information through interconnected artificial neurons layers
• By learning from data and adjusting the connection strengths (weights) between neurons layers, these networks excel at complex tasks like image recognition, natural language processing, and making predictions
• It’s used in wide range of applications such that Image and Speech Recognition (Identifying objects in images or understanding spoken language), Natural Language Processing (Powering chatbots, translation services, and summarization tools.), Medical Diagnosis (Assisting in the detection of anomalies in medical scans.) and Financial Market Analysis (Identifying patterns to predict market trends.) etc.

Now let’s go over their types. Although there are more categories, these neural networks types are basis for others.

Feedforward Neural Network (FNN):

• Feedforward network is a basic neural network where information flows only in one direction, from input to output, with no cycles or loops.
• It has Input layer, one or more hidden layers, and an output layer and processes data in a single, unidirectional pass.
• FNN lacks internal memory and hence it cannot retain information about previous inputs.
• Example includes- basic classification and regression tasks

Convolutional Neural Networks (CNN):

• CNNs are a type of feedforward neural network (FNN) designed for grid-like data.
• Filters (aka kernels) are used along with convolutional layers to detect spatial patterns.
• Like other feedforward networks, CNNs primarily consider the current input and have fixed input/output sizes.
• They are primarily used in analyzing visual and spatial data, such as images.

Recurrent Neural Networks (RNN):

• RNNs are designed for sequential data where order matters such as time series or text.
• They incorporates feedback loops that allow them to process sequential data and “remember” previous inputs.
• RNNs are used in natural language processing (NLP) for understanding context from past data and predicting the next word in a sequence.

Hope you liked the article. Please share your comment/ suggestion.