What is machine learning ?
- Machine learning is a subset of Ai where machine learns to find pattern from data and make predictions or decisions without being explicitly programmed for every set of rule.
For interview standards - Machine learning is a technique where systems learn form historic data to improve performance on tasks.
For Example
1. Email spam filter
2. Netflix recommendations
3. credit card fraud detection
Core Idea - The core idea of the machine learning is, the data is given to the machine learning algorithm, and the algorithm is trained on a machine learning model, and we get output that predicts the value from the trained data.
" If there is no data, There is no machine learning, period "
AI VS ML VS DL
- Ai stands for ARTIFICIAL INTELLIGENCE, It has a broad concept and the goal is to make machines behaves like humans.
For ex - chess engine with fixed rules.
Chatbots with pre defined logics.
Machine learning
- Machine learning is a subset of artificial intelligence. machines learns from the data and required feature engineering and works well on structured and tabular data.
For ex - House price prediction
Customer churn prediction
Deep learning
- Deep learning is a subset of machine learning. It uses something called neural networks. It automatically learned from features and needs large data + high compute power.
For ex - Image recognition
Speech recoginion
Face detection
Relationship between these 3
AI ---> ML ---> DL
One line comparision
| Term | Depends on data | Featured Engineering | Complexity |
|---|---|---|---|
| AI | Not always | Manual | Low Mid |
| ML | Yes | Required | Mid |
| DL | Huge data | Automatic | High |
Types of machine learning
- There are mainly 3 types of machine learning
- Supervised Machine learning
- Unsupervised Machine learning
- Reinforcement Machine learning
1. Supervised machine learning - In this method the data is labeled and we know the input and output. The goal is to learn mapping between them.
For ex - Predicts salary (Regression)
Email spam / Not spam (Classification)
There are two types -
Regression - For continous outputs
Spam / Not spam
Yes / No
Classification - X(Features) ---> Model ---> Y(Target)
2. Unsupervised machine learning - In this method there are no labels, model finds hidden patterns from data and this method is mostly used for exploration.
For ex - Customer segmentation
Grouping similar products
The common tasks are clustering, dimensionality reduction.
" I don't know the answer, just show me the patterns "
Reinforcement Learning - In this method there are models which learns by doing trail and error method. The agent interacts with environment and user give rewards for doing good work and penalty for bad works.
For ex - Game playing (chess, go)
Self driving cars
Robotics
Key terms
Agent ---> Environment ---> Action ---> Reward
