Forward Propagation (முன்னோக்கி செல்லுதல் / Passing the Message)

Let's order a Dosa in Saravana Bhavan! (சரவண பவன்ல தோசை ஆர்டர் பண்ணுவோமா!)

Technical Meaning: முன்னோக்கு பரவல் (Munnokku Paraval) - Pushing input data forward through the network to get a prediction.

The Core Idea

Forward Propagation is the process of pushing data forward through a neural network to get a prediction. The input data enters the first layer, gets multiplied by weights, passes through activation functions, and moves to the next layer until it reaches the final output layer, which spits out the model's guess.

The Origin Story

The term "propagation" simply means "to spread or transmit." Just like a ripple spreading across a pond when you throw a stone, the input data transmits its numerical values across every single connection in the network. During this forward journey, the network doesn't learn anything new—it is simply using what it currently knows to make a prediction.

The Tamil Analogy

Think of Forward Propagation like the workflow inside a busy Saravana Bhavan (சரவண பவன்) restaurant.

1. Input Layer: You place your order for a Masala Dosa with the waiter.
2. Hidden Layer 1: The waiter writes the token and shouts the order to the kitchen supervisor.
3. Hidden Layer 2: The supervisor assigns the exact ingredients to the Dosa Master.
4. Output Layer: The Dosa Master makes the dosa, and the delivery boy places the final dish on your table.

The information flows in only one direction: strictly forward from the customer to the final product on the table. Every person in the chain modifies the request slightly until the final output is generated!

Try It Yourself