A metric representing a model ’s loss against the test set . When building a model, you typically try to minimize test loss. That’s because a low test loss is a stronger quality signal than a low training loss or low validation loss .1
A large gap between test loss and training loss or validation loss sometimes suggests that you need to increase the regularization rate .
목차 목차 관련 글 ML crash course - Linear regression “The goal of training a model is to minimize the loss, reducing it to its lowest possible value.” ML crash course - Logistic regression 관련 절: Loss and regularization Log loss “Squared loss works well for a linear regression where the rate of change of the output values is constant.” Loss curve “A plot of loss as a function of the number of training iterations.” Regularization (machine learning) 관련 절: Training loss vs. real-world performance Validation set “Typically, you evaluate the trained model against the validation set several times before evaluating the model …” 손실 함수 “The goal of training is typically to minimize the loss that a loss function returns.” L1 loss “L1 loss is less sensitive to outliers than L2 loss.” 역전파 “Each of those neurons contribute to the overall loss in different ways.” Early stopping “A method for regularization that involves ending training before training loss finishes decreasing.” L2 loss “Regression models typically use L2 loss as the loss function.” Seeing again: treatment of functional visual loss “There is very little published literature on treatment strategies for functional visual loss.” mini-batch “For example, suppose the entire training set (the full batch) consists of 1,000 examples.” Iteration (machine learning) “A backward pass (backpropagation) to adjust the model's parameters based on the loss and the learning rate.” Convex function “A function in which the region above the graph of the function is a convex set.” Dropout regularization “A form of regularization useful in training neural networks.” Generalization (machine learning) “To encourage generalization, regularization helps a model train less exactly to the peculiarities of the data i…” L2 regularization “A type of regularization that penalizes weights in proportion to the sum of the squares of the weights.” 테스트 안 하기 “Unless you have reason to distrust it, don’t test code from others.” 경사하강법 “A mathematical technique to minimize loss.”