Error

As explained in Data Without Labels

Prediction error is the discrepancy between a model's predicted output and the corresponding actual value.

Prediction error is the difference between actual values and predicted values. Reducing this discrepancy improves accuracy and guides model optimization. section 8.2.6 section 8.3.4

Error supports several training decisions:

• Use the loss score derived from predictions and actual values as feedback for adjusting weights. section 8.3.4
• Reduce error with gradient descent, which moves iteratively toward a function's global minimum. section 8.2.6
• Choose batch, stochastic, or mini-batch updates according to dataset size, speed, accuracy, and batch-size tuning needs. section 8.2.6

Training feedback

After a network generates predictions, compare them with the actual values to calculate prediction error. Convert that result into a loss score, then use backpropagation to adjust the weights and lower future error. section 8.3.4

Figure 8.8 The concept of gradient descent. It is the mechanism to minimize the loss function.

Use the gradient-descent figure to visualize how iterative updates move toward the minimum of the loss function. section 8.2.6

How Data Without Labels covers it

How deep it goes

Defines error, connects it to loss and backpropagation, and compares update strategies without specifying an error formula.

Examples

Uses predicted-versus-actual values, gradient-descent movement, and batch, stochastic, and mini-batch training updates.

What it assumes

Assumes predicted and actual values, while defining their difference and linking it to model accuracy.

What you can do after

Diagnose training progress by comparing predictions with actual values and use resulting loss feedback to adjust model weights.

How it connects

Connects error with loss functions, gradient descent, backpropagation, weights, biases, and accuracy.

liveBookmanning.com

Read Data Without Labels on liveBook Get the book at manning.com

Related concepts

• scaling lawmeasured by · A scaling law is an empirical relationship describing how a system’s properties change as its scale or resource levels vary.
• accuracycontrasts with · Accuracy is the degree to which an output, prediction, or measurement agrees with a correct or reference outcome under a specified criterion.
• objective functionmeasured by · An objective function assigns values to candidate solutions, predictions, or parameter settings according to a criterion to be minimized or maximized.
• forecast errorincludes · A forecast error is the difference between an observed value and the corresponding forecasted value.
• out-of-bag errorincludes · Out-of-bag error is an estimate of an ensemble’s prediction error computed for each observation using only ensemble members whose bootstrap samples omitted it.

This explainer was generated by Manning's knowledge graph from the text of Data Without Labels. The section links point to the parts of the book it draws on.