Lernmaterialien für Big Data & Data Science an der FOM Hochschule für Oekonomie & Management

Qualitative Variable

...is an iterative optimization algorithm which means that in order to get the best result it is necessary to pass over the training set for a multiple of times in a way

• Algorithms from the group of reinforcement learning produce agents rather than classical models that receive information from the environment and react to it by performing an action
• information that the agent receives from the environment in form of numerical data is called state which is stored and then used for choosing the right action
• The result of the action, the agent receives a reward that can be either positive or negative which is then used as a feedback to the agent in order to update its parameters and change its future actions
• process of learning and training is a process of trial and error 
  • agent finds itself in various states and gets punished every time it takes the wrong action and thus starts learning how to take better actions

• Defined as a job a computer program has to learn in order to execute it
• Learning itself is not considered as a discrete task
• eg Teaching a computer in how to play checkers, then playing checkers itself is the Task T

• Characteristic attributes can be distinguished and ranked
• Intervals can be compared
• Metric scale divides further into interval and ratio scale

• Ratio Scale has a naturally given point of origin (eg body weight)
• Interval scale has an artificially given point of origin (eg calculation times)

Gradient Descent