Inference 

in Machine Learning

In machine learning, inference is the process of using a trained model to make predictions on new data. After training — where model parameters are learned from a dataset — inference applies these parameters to unseen inputs to produce an output, such as a classification or numerical estimate.
This is closely related to classical statistical regression: once a regression model is fitted to data, inference means applying the model — with coefficients estimated from the training data — to predict outcomes for new observations. Similarly, machine learning inference uses the learned model to generalize beyond the training data.

A real-life example is predicting the composition of a substance in an industrial process based on time series of physical and chemical sensor measurements. By incorporating data from instruments that were previously not considered relevant, or from sensors not expected to correlate with the main measurements, the model can make more accurate predictions and help reduce measurement uncertainty — ultimately improving process control and quality assurance.

This highlights the true strength of machine learning tools: in many cases, the underlying physical or chemical processes are too complex or insufficiently understood to be fully captured by classical models. The combination of model structure and parameters in traditional approaches may fail to reflect reality. Machine learning offers far greater flexibility, allowing the use of more complex models that can better approximate these unknown or poorly defined systems.