Classes of Data; Labeled and Unlabeled Data

There are two classes of data; labeled and unlabeled data. The difference between data, information, and knowledge, as well as their subsequent organization into a hierarchical structure, are key components of the conventional architecture when creating AI systems:

• Knowledge
• Information
• Data

The abbreviation DIK, which stands for the initial letter of each of its constituents, is used to refer to this structure as the "knowledge pyramid" or " DIK pyramid." Even though this theoretical approach has received a lot of criticism, it is still frequently employed as a common conceptual framework for the creation of AI systems.

According to the model, it is right to assume that AI systems gather data to create information, a process that information to create knowledge, and then apply that knowledge to guide further data acquisition.

Data

Data, the foundation that links the machine learning system to reality, is at the bottom of the pyramid. The collection of measurements or observations made by sensors, which have a raw or unarticulated form, can be understood as the data.

Data examples include:

• A matrix containing numbers
• Strings of text
• A list of categorical values
• Sampled audio frequencies

The value or values included in data structures are referred to as "data" in this context.

Information

To uncover patterns from the data, it may be combined in various ways. By applying mathematical or statistical models to the data, patterns that represent regularities in the distribution of the data can be retrieved.

Information is positioned higher on the pyramid than raw data because it is better able to capture the complexity of the outside world. One approach to express the same concept is to suggest that, unlike data itself, information, or the schemes or patterns of data, allow one to make predictions about the outcomes of upcoming measurements.

Knowledge

After data collection has identified trends, they can be utilized to forecast how the system's activities will affect the state of the world in the future. The predictions made from analyzing data and information can be assumed to be knowledge.

Labeled and Unlabeled Data

All information is unlabeled. It only becomes labeled data when we attribute our past knowledge to it. Data that has not been labeled with information indicating its features, attributes, or categories are referred to as unlabeled data. Unlabeled data is frequently employed in a variety of machine learning techniques. The only pure data that exists is unlabeled data. We get unlabeled data when we activate a sensor or open our eyes with no prior knowledge of the surroundings or how the world works.

Unsupervised machine learning refers to the type of machine learning when the computer program analyzes collections of unlabeled data. The machine learning algorithm distinguishes each piece of data based on its qualities and attributes because the data lacks labels.

Labeled data is data that we have prior knowledge and understanding of. The additional information is imposed on the data by a human or an automated tagger using their prior knowledge. Prior knowledge helps in the understanding and analysis of new data.

Labeled and Unlabeled Data Examples

Example of Unlabeled Data

• A machine learning program presented with pictures of fruits with no data pieces labeled with their names will describe them by their characteristics, i.e., color, shape, etc
• A machine learning program presented with pictures of fruits with no data pieces labeled with their names will describe them by their characteristics, i.e. color, shape, etc
•

Examples of Labelled Data

Examples of data using labels include:

• An image of a cat or dog with the words "cat" or "dog" next to it
• An image of a cat or dog with the words "cat" or "dog" next to it
• a user's score for the product in question together with a brief explanation of the review
• Property characteristics and asking price

Factors to Consider when Selecting Whether to Use Labeled or Unlabeled Data

• The task in question
• The purpose of the task
• The amount of data available
• The amount of general and specialized information needed
• The degree of complexity of the decision function