Alessandro Morita Enjoying the thermodynamic limit
Linear trees in LightGBM: how to use
What are linear trees? This was originally written as a “Hello world” kind of program aimed at giving my team at the DataLab some help getting started with less noisy variants of GBDTs. From this post: Not everybody knows simple yet effective variations of the Decision Tree algorithm. These are known as Model Trees. They learn an optim... Read more 21 Sep 2022 - 2 minute read
A ROC AUC partial to misclassification cost
Functional implementation Below, as a reference, we leave a function to calculate these metrics following a syntax close to the scikit-learn one. def cost_based_roc_auc(y_true, y_probs, cost_fn=None, cost_fp=None, return_curves=False): from sklearn.metrics import roc_cur... Read more 20 Sep 2022 - 4 minute read
A vector space structure for probabilities
Going to $\mathbb{R}^n$ and back again Let $\Delta_{K}$ be the $K+1$-dimensional probability simplex, which is the natural place for probabilities to live in: \[\Delta_K := \left\{ p \in [0,1]^{K+1}: \sum_{k=1}^{K+1} p_k = 1 \right\}\] Define the logit function as the map $\phi: \Delta_K \to \mathbb R^{K}$ such that, if $p_i$ is the $i$-th co... Read more 18 Sep 2022 - 6 minute read