To get to know about data mining techniques we should have a base knowledge of Artificial Intelligence. Here I wrote about AI from the book ‘Artificial intelligence book by Peter Norvig and Stuart J. Russell’ to better understand how AI algorithms or specifically data mining algorithms work. I will write in four sections to understand easier.
This is the second part of the AI section, We would learn what is AI and how it is working.
In the previous part, We found out what is actually artificial intelligence is. In this part, We would explain how environments work and several types of agents.
Before we start the environment types we should know what is an agent in AI. An agent is anything that can sense its environment and give feedback (react) to the environment with its actuators.
Let’s dive into environment types
1. Fully Observable vs. Partially Observable:
If our agent can sense the environment with its sensor and can find out its exact state so our environment is fully observable and partially observable means it cannot get its exact state information. For example, a medical diagnosis’s environment is partially observable because a doctor's diagnosis may not be true.
2. Deterministic vs. Nondeterministic:
If our agent can predict exactly its effect on the environment by its act we can say the environment is deterministic if not the environment is nondeterministic.
3. Episodic vs. Nonepisodic:
If our agent's decision doesn't relate to other decisions in past or doesn't relate to the past state of the environment, our environment is episodic, If not our environment is nonepisodic. For example, the chess game environment is continuous, meaning our act ( agent’s act ) will remain to the end of the game.
4. static vs. dynamic:
While our agent is making decisions the environment changes, Our environment is dynamic, and if not it's static. For example, We can use the taxi driver problem. In a taxi driver environment, the environment can change every second for example a bike may come near the car and the driver needs to change its path.
5. Discrete vs. Continuous :
If the environment’s time is broken down into finite numbers (ex: 1,2,3,… ) the environment is discrete, else the environment is continuous.
for an example of a continuous environment, we can say a taxi driver’s environment. And for a discrete environment, we can say chess game with/without clock ( game turns are broken down into finite numbers ).
6. Single-agent vs. Multi-agent:
Some environments are single-agent like the Crossword game. And an example for multi-agent environments is chess ( multi-agent environments contain more than one agent, here in chess we have two agents ).
Multi-agent environments are two types:
- Competetive: In competitive environments, agents are trying to maximize their score to minimize other agents' score.
- Cooperative: In this type of environment agents together are trying to maximize their score (with an agent getting a higher score other agents would get a higher score too).
So till now, we got a better understanding of environments. It's time to know agent types.
- simple reflex agent: The agent would react to environmental events with its default information. ( no state saving, no conclusion, … )
- model-based agent: This kind of agent would use internal memory to save the previous state. In conclusion, our agent would use the previous state and the present state to react to the environment. ( before the act internal memory would be updated )
- goal-based agents: The agent would react in a way to get to its goal. (state saving is available, no conclusion )
- Utility-based agents: The agent would react in a way to maximize its goals. For example, again a taxi driver has more than one goal, One to minimize fuel consumption and Two to get to his destination sooner, So he has to choose his actions in a way to maximize all his goals.
- Learner agents: Four agents we read didn’t have the ability to learn from their experience. This type of agent can learn and increase their knowledge about the environment.
So we understood the basic terms of environments and agents. If you want to know you can ask a question below or read the textbook Artificial Intelligence A Modern Approach by Stuart J. Russell and Peter Norvig.
In the next section, we will learn how to Solving Problems by Searching. I will post in within two or three days So stay tuned and feel free to ask any questions.