A logit is an unnormalized output score produced by a model before conversion into class probabilities.
A logit is an unnormalized output score from a neural network’s final layer. Larger logits indicate stronger evidence that an input belongs to a class; softmax converts them into probabilities. section 9.7
Computing class logits
For the penguins classifier, forward propagation computes hidden activations with tanh, then multiplies them by the output weights and adds output biases: logits = a1.dot(W2) + b2. Softmax and cross-entropy loss use these scores afterward, while prediction can select the class with the largest logit.
How Optimization Algorithms: AI techniques for design, planning, and control problems covers it
How deep it goes
Defines logits and computes them through hidden-layer activation, output weights, and biases before applying softmax.
Examples
Uses a penguins classifier with four inputs, ten hidden units, three output classes, and NumPy forward-propagation code.
What you can do after
Apply it via Python and NumPy to compute class scores, probabilities, loss, and predictions.
How it connects
Connects logits with neural networks, tanh, softmax, cross-entropy loss, forward propagation, and argmax prediction.
Statistics Slam DunkLog odds is the natural logarithm of the odds of an event, equal to the logarithm of its probability divided by its failure probability.
Deep Learning with PyTorch, Second EditionUnnormalized model scores produced before applying a probability normalization such as softmax, used here as the direct input to cross-entropy loss.
Statistics Every Programmer NeedsLog odds are the natural logarithm of the odds of an event, calculated as the logarithm of its probability divided by the complement probability.
temperatureused by · A temperature parameter adjusts the concentration of a probability distribution, affecting sampling randomness or the likelihood of accepting less favorable transitions.
softmaxused by · Softmax is a function that exponentiates a vector of scores and normalizes the results into nonnegative values that sum to one.
linear layerproduced by · A linear layer is a neural-network layer that maps input features to outputs by a learned affine transformation, typically connecting each output to every input feature.
language modelproduced by · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.
regression coefficientmeasured by · A regression coefficient is a model parameter that quantifies a predictor’s contribution to a regression model’s output, conditional on other model terms.
sigmoidequivalent to · A sigmoid is a smooth, bounded, S-shaped function that maps real-valued inputs between two limiting values, commonly zero and one.
This explainer was generated by Manning's knowledge graph from the text of Optimization Algorithms: AI techniques for design, planning, and control problems. The section links point to the parts of the book it draws on.
