Logit

Logits (aka “Raw Scores”) are the direct result of the linear operations (weighted sums of inputs plus bias) in the final dense layer of a neural network.

Resources

• Understanding logits in AI and neural networks
• Logits in machine learning