2. Neural Networks
Neural Weights (முக்கியத்துவ அளவீடு / Importance Factors)
Let's make Kalyana Sambar! (கல்யாண சாம்பார் வைப்போமா!)
Technical Meaning: நரம்பியல் எடைகள் (Narambiyal Edaigal) - Values that determine the strength of the connection between neurons.
The Core Idea
In a neural network, a weight is a number that represents the strength of the connection between two neurons. When the network learns, it is simply adjusting these weights up or down until it consistently makes the right prediction.
The Origin Story
The human brain consists of billions of neurons connected by synapses. When we learn something new, the synaptic connections between specific neurons get stronger. Computer scientists modeled artificial neural networks after this exact biological process. The "weight" is the digital equivalent of a synapse's strength.
The Tamil Analogy
Think of weights like mastering the perfect Kalyana Veetu Sambar (கல்யாண வீட்டு சாம்பார்).
Your grandmother knows that a good sambar needs tamarind (புளி), sambar powder (சாம்பார் பொடி), and dal (பருப்பு). But what makes it perfect is the exact proportion (the weight) of each ingredient.
- If the tamarind weight is too high, it becomes too sour (புளிப்பு).
- If the sambar powder weight is too low, it lacks flavor.
Every time she cooks, she adjusts these "weights" slightly based on how it tastes (the feedback). Over years of experience, she finds the perfect, exact weights to consistently make the best sambar. A neural network does exactly this—adjusting its mathematical ingredients millions of times until the output is perfect.
Try It Yourself
Play with the connection weight below to see how the network's confidence in its output changes!
Concept 1 · Neural Networks
Learning is just adjusting weights
A neural network 'learns' by strengthening or weakening the connections (weights) between its nodes until it gets the right answer. Try dragging the slider to strengthen the connection.
Network Output
The network is learning the pattern, but isn't confident yet.