Information Theory
Core Concepts
- Information content
- Entropy
- Joint entropy
- Conditional entropy
- Mutual information
- Cross-entropy
- KL divergence (Kullback-Leibler Divergence)
Applications in Large Models
Loss Function
- Cross-entropy loss is a measure of the difference between the predicted distribution and the true distribution.
Attention Mechanism
- When computing attention weights, the softmax operation relates to probability distributions and entropy.
Reinforcement Learning
- The optimization objective in policy gradient may include an entropy regularization term to encourage exploration.
- The core of TRPO / PPO algorithms is a KL divergence constraint.
Model Compression and Quantization
- Evaluating quantization information loss.
贡献者
这篇文章有帮助吗?
最近更新
Involution Hell© 2026 byCommunityunderCC BY-NC-SA 4.0