Hebb Network
Hebbian learning ek classic concept hai neural networks mein. Iska naam uske developer Donald Hebb ke naam par rakha gaya hai.
Hebbian learning ka basic idea yeh hai ki “jo saath saath activate hota hai, woh ek doosre se connect hota hai”. Matlab, agar ek neuron doosre neuron ko activate karta hai, to us connection ko strengthen kiya jata hai.
Iska simple example yeh hai: agar aap baar-baar kisi particular insaan ka chehra dekhte hain, to aapke brain mein woh specific neurons jo us insaan ke chehre ko recognize karte hain, woh activate hote rahenge. Is tarah, un neurons ke beech ki connection strong hoti jaati hai.
Agar aap is concept ko ek network mein apply karte hain, to jab do neurons repeatedly ek saath activate hote hain, unke beech ka connection strong hota hai. Is tarah, Hebbian learning ka use karte hue, neural networks khud se patterns seekh sakte hain.
Iska ek limitation yeh hai ki agar kuch neurons ek doosre ko bahut zyada activate karte hain, to woh overfitting ka issue create kar sakte hain, jisme network sirf specific examples ko yaad karta hai aur generalize nahi kar pata.