The four steps every Machine Learning application takes – even Google’s
The four steps every Machine Learning application takes - even Google's
To learn about Machine Learning sometimes becomes confusing when you don’t know where to start. To make your way you can use the four fundamental steps of ML as a compass.
The learning process of a machine learning model always passes through a defined process of calculations. This process can be found in all ML models and applications. From social media feeds to the self-driving car. It is the fundamental concept of machine learning:
- data acquisition/hypothesis
- error calculation
- parameter updates/error minimization
First, a hypothesis is calculated on the basis of the available data, which should describe the existing observations as well as possible. In very rare cases, however, the hypothesis coincides exactly with the data points (and if it does, you should become suspicious and question the model or the data). This means that there are distances and errors, which are calculated in the second step using an error function.
The actual learning then takes place in step number three by minimizing the calculated errors until an optimal value is reached. So-called optimization algorithms, such as gradient descent, are used here.
A decisive role is played by the so-called learning rate, which defines the size of the individual optimization steps.
In supervised learning, where we already know about the output, we have step number four in addition. The needed process of evaluation. By previously dividing the data set into a training block and a test block (usually in a ratio of 70/30) it can be checked how accurate the model can make predictions for unknown data.
Was this article helpul for your work? Leave a comment 🙂