After experimenting with teaching my computer to draw black‑and‑white line art, I wondered: can it also learn words?
So I grabbed a list of 500 Indian names, built a 5‑gram vector, and fed it into a Vanilla Neural Network (written in Rust). Following are the network configuration:
║ Name: five-gram
║ Parameter Count: 758056
║ Hidden Layers: 4
║ Vocabulary Size: 25
║ Loss Function: Categorical Cross Entropy
║ Optimizer: Gradient Descent
Fifteen minutes later, it started generating brand‑new names that felt surprisingly authentic.
Some of my favorites: Yaman, Samanya, Samika, Praman, Sakhi, Debika, Mazhar, Maera, Narayani, Manyashree, Adhya, Manpreet, Jameera, Kash, Kaya, Nidhi…
It’s fascinating to watch creativity emerge from code.
Top comments (0)