A brief introduction to Machine Learning and its techniques
What is Artificial Intelligence (AI)?
Is the intelligence shown by machines where it can perceive certain data, analyze and then take actions that is most suitable to its environment? For example, as humans when we see a car passing, we take action and avoid it.
Artificial Intelligence is divided into two categories applied AI and general AI. Examples of applied AI, Google search engines uses AI to find the best searches for the words you input. Facebook uses AI in its advertising where it shows ads that are related to your interests from previous searches. General AI is much more complicated as it’s like having robots that think and acts like humans. They plan and perform general tasks, recognize different objects and sounds, learn and apply new jobs.
Machine learning subfield of AI.
Machine learning is a method used to achieve AI. HOW IS THIS DONE? In simple terms, Instead of writing thousands of lines of code, a huge amount of data will be given to the machine, the machine will try to find the correct answers and as more data is given the better the machine will become in realizing the correct from the wrong. For example, a test will be given to see if this picture is human or not. Thousands of images will be shown, humans face and others like cats, dogs or objects. After a lot of trials and errors, the machine will get better at recognizing human faces.
Machine Learning Techniques:
- Supervised learning techniques
- Unsupervised Learning Techniques
- Reinforcement Learning Techniques
Supervised learning techniques
This technique requires human supervision. You set a mapping function and train it using a trained set of data. These trained set of data has inputs and desired outputs. When an error occurs, you alter the mapping function to the desired output. When the performance of the mapping function is achieved, the test set of data are used to test the mapping function. These set of data are not used in the training and this will test the accuracy of the function.
Unsupervised Learning Techniques
In unsupervised learning, the system is presented with unlabeled, uncategorized data and the system’s algorithms act on the data without prior training.This technique is used when the mapping function has a set of data (inputs) and have to categorize them into classes based on attributes within those data. For example, there are 10 patients and we know there attributes blood type, blood sugar, cholesterol and other info. So we can categorize into classes based on their sugar level or cholesterol level.
Reinforcement Learning Techniques
In this technique, the agent has no prior training instead he will have to decide how to perform tasks, then gets a reward function (like a statement this action is good or this action is bad). The agents start learning from those experiences through trial and error with the goal to get maximum reward.
In my opinion, this technique is good for the advancement of Artificial Intelligence.