Introduction to Machine Learning
- What is Machine Learning (ML)
- Machine Learning is study of computer algorithms that improve experience
- Seen as subset of Artificial Intelligence (AI)
- ML algorithms build a mathematical model based on sampe data in order to make predictions or decisions without being explicitly programmed
- Also referred to as Predictive Analytics in its applications business problems.
- Where is ML Used?
- Used in wide variety of applications
- Email Filtering
- Loan Prediction
- Sales Forecasting
- Movie Recommendation
- Why Machine Learning?
- Programs to solve problems like facial recognition are very hard
- Instead of writing a program, collect lots of examples that specify the correct output for a given input
- An ML algorithm then takes these examples and produces a program that does the job
- Program works for new cases & ones it trained on
- Requirements of ML
- Some pattern must exist between the input and output of data (the relationship cannot be completely random)
- Lottery winning numbers cannot be predicted using ML
- Enough data that needs to be present to discover this pattern
- No easily definable set of rules or mathematical formulas should exist that can explain the relationship between input and output of the data
- If it did, we can use traditional programming
- Examples of TASKS BEST SOLVED WITH ML
- Recognizing patterns
- Facial identities
- Handwritten or spoken words
- Recognizing Anamolies
- Unusual sequence of credit card transactions
- Unusual patterns of sensor readings in nuclear power plant
- Unusual sound in the car engine
- Prediction:
- Future Stock prices or currency exchange rates
- Employee attrition
- Loan defaulting
- More Examples:
- Spam filtering, Fraud detection
- Recommendation Systems
- Autonomous cars
- Speech recognition
- Image classification/tagging
- Language translation
- Price Prediction
- Advertising: Ad click prediction
- Playing games (Chess, Go, StarCraft etc.)
- Machine Learning Approaches:
- Traditionally divided into three categories
- Supervised Learning
- Unsupervised Learning
- Reinforcement Learning