Machine Learning Series: Machine Learning Debunked
What is Machine Learning?
Machine learning is a subfield of AI that is the science of programming computers so they can learn from data without being explicitly programmed.
In the past, people used to write a lot conditions, or make the computer search through all the possible combinations.
However, with the help of machine learning, we can teach the computer to learn from the data and make predictions.
Types of Machine Learning
There are three types of machine learning:
- Supervised Learning: Given a training dataset and corresponding labels (or a target value), the model learns to predict the labels by learning from the training dataset.
Main problems: Classification and Regression.
- Unsupervised Learning: Given a training dataset without any labels/target values, the model learns to find the patterns in the data.
Main problems: Clustering, Dimensionality Reduction and Anomaly Detection.
- Semi-supervised Learning: Given a training dataset with few labels/target values, the model learns to predict the labels by learning from the training dataset.
Main problems: Classification and Regression, same as supervised learning.
- Reinforcement Learning: The model learns to make a sequence of decisions. The model learns to make decisions by trial and error. (by rewards and penalties).
Main problems: Playing a game (i.e. osu!), Robotics.
This video is an evidence of computers can learn playing games using reinforcement learning and even beat the top players!
Problems that Machine Learning can solve
We have seen some problems in the previous section, but let's see what they actually are:
- Classification: Predicting the category (also called as class or label) of a given data point.
- It has three variations: Binary Classification, Multi-class Classification and Multi-label Classification.
- Binary Classification: Predicting if the given data point belongs to one of the two classes.
- Multi-class Classification: Predicting if the given data point belongs to one of the multiple classes.
- Multi-label Classification: Predicting if the given data point belongs to multiple classes.
- Example: Spam Detection (detecting if the e-mail is spam or not, binary classification), Handwritten Digit Recognition (recognizing the digit from 0 to 9, multi-class classification).
- It has three variations: Binary Classification, Multi-class Classification and Multi-label Classification.
Here is an illustration of binary classification by spam detection: Take the e-mail, digest and determine whether it is spam or not spam.
- Regression: Predicting the value of a given data point.
- Example: Predicting the price of a house based on the features of the house (like number of rooms, area, etc.).
- Clustering: Grouping the data points into clusters based on the similarities, patterns in the data.
- Example: Analyzing the behavior of programmers based on the topics they are interested in.
- Dimensionality Reduction: Reducing the number of features in the data. We reduce the number of the features to reduce the complexity of the data.
- Anomaly Detection: Detecting the outliers in the data.
- Example: Detecting the fraudulent transactions in the bank, a classical example.