Which of the three elements is the most important in machine learning: data, model and algorithm?

Model, data and algorithm can be summarized as the outline of machine learning method. Because of different emphases, it is impossible to compare which is more important.

Model: The determination of the model mainly defines what the user-defined prediction function looks like. There are two forms, P(Y|X) and y=f(x). Because there is more than one unknown parameter ceta in these two formulas, when ceta is uncertain, the formulas are all expressed as various sets. From this, we solve the unknown parameters in the formula and determine the final formula, and use this formula to predict.

Data: After defining the style of the model, we need to use the known data to explore the unknown parameters. Whether it is training data or test data, we finally hope that the formula can predict the same result as the real situation. In fact, it may be exactly the same, and there will be some errors. We can call it a loss, so we introduce a loss function and use the minimum loss to find the unknown parameters.

Algorithm: Find the minimum value, find the first derivative of unknown parameters, and try to make the derivative equal to 0. There are two commonly used algorithms in machine learning: least square method: for linear model; Gradient descent and ascent methods (batch gradient, incremental gradient): any module!

If you want to know more about data, models and algorithms, I suggest you take a look at the course of CDA Data Analyst. The course content takes into account the cultivation of horizontal ability to solve data mining process problems and vertical ability to solve data mining algorithm problems. Explore business problems through digital working methods. Click to make an appointment for a free audition class.