Neural Networks
Software
Problems
Problem | Description | Output layer | Loss |
---|---|---|---|
Regression | Real number (-∞, ∞) | 1 neuron without act | Mean Squared Error (MSE) |
Binary Classification |
2 exclusive classes | 1 neuron + Sigmoid | Binary Cross Entropy (BCE) |
Multi-Class Classification |
N exclusive classes | N neurons + Softmax | Cross-Entropy (CE) |
Multi-Label Classification |
N non-exclusive classes | N neurons + Sigmoid | Binary Cross Entropy (BCE) |
- You can just consider the multi-label classifier as a combination of multiple independent binary classifiers.
- Cross-Entropy is called Categorical CrossEntropy in Keras.
Backpropagation