Generative models woh AI models hote hain jo naya data create karte hain (text, image, audio, video, etc.). Neeche main point-to-point unke types bata raha hoon:
1) Generative Adversarial Networks (GANs)
Kaam: Real jaisa naya data banana (zyada tar images).
Structure: 2 networks – Generator + Discriminator (ek banata hai, doosra check karta hai).
Example Use: Face generation, deepfake, image enhancement.
2) Variational Autoencoders (VAEs)
Kaam: Data ko compress karke phir se naya similar data generate karna.
Structure: Encoder + Decoder.
Example Use: Image generation, anomaly detection.
3) Autoregressive Models
Kaam: Ek-ek karke next value predict karte hue data generate karna.
Example Use: Text generation (jaise GPT), music generation.
4) Transformer-based Models
Kaam: Attention mechanism se context samajhkar generation karna.
Example Use: Chatbots, translation, summarization.
Famous Example: OpenAI ka GPT model.
5) Diffusion Models
Kaam: Noise se dheere-dheere clear image banana.
Example Use: AI image generation (jaise Stable Diffusion).
6) Flow-based Models
Kaam: Exact probability calculation ke saath data generate karna.
Example Use: Image modeling, density estimation.
📌 Short Summary (Simple Language)
- GAN → Real jaisi images banane me expert
- VAE → Data ko compress karke naya data banana
- Autoregressive → Next word/step predict karke generation
- Transformer → Attention se smart text generation
- Diffusion → Noise se image banana
- Flow-based → Mathematical probability ke saath generation
Top comments (0)