Vector Quantization
التكميم المتجهي
تقنية تُجبر التمثيلات الكامنة على أن تكون متجهات منفصلة من قاموس مُتعلَّم بدلاً من قيم مستمرة، مما يُنشئ فضاءً كامناً مُنظَّماً.
A technique that forces latent representations to be discrete vectors from a learned codebook instead of continuous values, creating a structured latent space.
Also translated asالتكميم الشُعاعي، تكميم المتجهات، كمّنة المتجهات
First appears in this corpus in: Least Squares Quantization in PCM (1982)
Appears in these papers
- BEiT: BERT Pre-Training of Image Transformers2021in the sky ✦
- BEiT: BERT Pre-Training of Image Transformers2021in the sky ✦
- Billion-Scale Similarity Search with GPUs2017in the sky ✦
- Jukebox: A Generative Model for Music2020in the sky ✦
- Least Squares Quantization in PCM1982in the sky ✦
- High-Resolution Image Synthesis with Latent Diffusion Models2022in the sky ✦
- LLM.int8(): 8-Bit Matrix Multiplication for Transformers at Scale2022in the sky ✦
- MusicLM: Generating Music From Text2023in the sky ✦
- Neural Discrete Representation Learning2017in the sky ✦
- Neural Discrete Representation Learning2017in the sky ✦