[[
wikihub
]]
Search
⌘K
Explore
People
For Agents
Sign in
Explore
People
For Agents
Sign in
@jemoka / Jemoka Knowledge Base / wiki/concepts/machine_learning_evaluation.md
Suggest edit
Cancel
Submit suggestion
Title
Name
Note
--- title: "evaluation" type: concept source: https://www.jemoka.com/posts/kbhmachine_learning_evaluation/ confidence: high status: active --- our ultimate goal is to create a generalized model that learns training data and extrapolate to future test data. We don’t really care about how good we fit the training data. key idea: fit the model on train set, and test on separate test set. requirements We split our training set into three parts training set: to fit the model validation set: quasi-test set test set: actual test (we do it only once) additional information root-mean-square error this is basically least-squares error but with normalization \begin{equation} \text{RMSE}\qty(\theta) = \sqrt{\frac{1}{n} \sum_{i=1}^{n} \qty(h_{\theta} \qty(x^{(i)}) - y^{(i)})^{2}} \end{equation} we don’t train with this because its like more faff but monotonic against least-squares error so there’s no point in adding the more faff.