Blog 9
What is Machine Learning?
Gizem Baruk I 18.08.2021
Gizem Baruk I 18.08.2021
Machine learning is a subfield of artificial intelligence. With the help of machine learning, IT systems are able to recognize regularities and patterns based on existing data sets and algorithms and to develop solutions for them. In short, artificial knowledge is generated from experience. Insights that have already been gained from data can be generalized and used for new problem solutions or for analyses of previously unknown data.
In order for the software to learn and find solutions independently, the systems must be provided with the data and algorithms relevant for learning. To do this, rules must be established for analyzing the data set and recognizing patterns. If suitable data is available and rules are defined, systems with machine learning can:
• Make predictions based on the analyzed data
• Find, extract and summarize relevant data
• Calculate probabilities for certain events
• Optimize processes based on recognized patterns and adapt independently to developments
In order for the software to learn and find solutions independently, the systems must be provided with the data and algorithms relevant for learning. To do this, rules must be established for analyzing the data set and recognizing patterns. If suitable data is available and rules are defined, systems with machine learning can:
• Make predictions based on the analyzed data
• Find, extract and summarize relevant data
• Calculate probabilities for certain events
• Optimize processes based on recognized patterns and adapt independently to developments
What types of machine learning are there?
• active learning
• reinforcement learning
• unsupervised learning
• semi-supervised learning
The semi-supervised learning
consists of the approaches of unsupervised and supervised learning and is therefore a mixture of both methods.
Algorithms play a central role here. They are responsible for recognizing patterns and generating solutions and can be classified into different categories:
• reinforcement learning
• unsupervised learning
• semi-supervised learning
• supervised learning
At the active learning
The algorithm reacts to input data by using specified questions and thus asking for appropriate results. The questions are selected by the algorithm based on the relevance of the results. The data can be available online or offline, so the origin of the data is not important. The data can also be used multiple times for the learning process.
The reinforcement learning
is based on the principle of rewards and punishments. Negative and positive reactions tell the algorithm how to react to different situations.
At the unsupervised learning
No target values or rewards are defined at the beginning of the learning process. Often the focus is on learning clustering (data point groupings). The algorithm basically tries to differentiate and structure the available data according to independently identified characteristics. The algorithm is thus able to sort individual objects based on their color.
At the supervised learning
In contrast to unsupervised learning, example models are defined in advance. For further assignments, the basic models are then specified, which means that the system learns based on the input and output pairs. During the learning phase, the developer provides the appropriate values for individual inputs and thus contributes to the learning process, so the system has the opportunity to identify relationships in data.
What are the benefits of using this technology?
Machine learning is intended to help people work more efficiently and give them more space to be creative. The technology supports the organization and management of large amounts of data and takes over stupid and repetitive tasks. For example, machine learning can support people in the processing of data by helping to prepare, save and file paper documents. They also have the potential to take on particularly complicated tasks. This includes identifying errors or predicting future damage.
Machine learning is intended to help people work more efficiently and give them more space to be creative. The technology supports the organization and management of large amounts of data and takes over stupid and repetitive tasks. For example, machine learning can support people in the processing of data by helping to prepare, save and file paper documents. They also have the potential to take on particularly complicated tasks. This includes identifying errors or predicting future damage.
