Machine Learning

In many machine learning and decision-making systems, what we encounter is not a directly measurable quality score, but rather a large number of preference judgments in the form of pairwise comparisons, that is deciding which of two options is better. Although such pairwise comparison data is simple in form, it implicitly contains rich structural information. Starting from a probabilistic semantics perspective, this article will gradually explain how the Bradley–Terry model can transform these preference comparisons into a learnable representation of latent utilities.

In probabilistic modeling and machine learning, entropy is a fundamental concept for quantifying uncertainty. It not only describes the inherent randomness of data, but also implicitly captures the minimum information cost required in prediction and modeling. Many learning objectives that may appear different on the surface, such as maximizing log-likelihood or designing loss functions, can in fact be traced back and understood through the lens of entropy.

When a model has poor performance, it cannot predict the data accurately. The main cause may be overfitting or underfitting. If it is a case of overfitting, we can use regularization to solve model overfitting.

Overfitting and underfitting are the root causes for poor model accuracy. Only by being able to determine whether a model is overfitting or underfitting can we take the correct approach to improve the performance of the model.

In machine learning, before we take the data to train a model, we usually need to perform feature scaling on the data.

The confusion matrix is ​​a tool used to measure the performances of models. This allows data scientists to analyze and optimize models. Therefore, when learning machine learning, we must learn to use confusion matrix. In addition, this article will also introduce accuracy, recall, precision, and F1 score.