2 Machine Learning Foundations
What is Machine learning
A subset of artificial Intelligence that focuses on creating computer systems that can learn and improve from experience. Powered by algorithms that incorporate intelligence into machines.
ML Model: Inputs and Outputs.
Model training: A relationship is established between the input features and the output label y=f(x). Features (X) are relevant information from the available data selected to train the model. Label (y) is the prediction that the model aims to generate based on the input features.
Supervised Learning - Regression
is a Machine Learning model which learns from labeled data Classify data and make predictions
Machine Learning
is creating computer systems to learn from experience instead of specially programed.
Machine Learning example
Example: Showing a puppy the difference between a ball and a book. Show a computer the difference between a cat and a dog. Show pictures of dogs and cats. This is the data. Data Shown is the information.
When is ML NOT the optimal solution?
ML provides good solution where there is a lot of data Simpler Alternatives Rule-based approach used for sorting, validating, and verification. Insufficient Data Self-driving cars data set Is not extensive enough to handle adverse situations. High cost Basic image and text processing ML resizing, filtering, counting Complex Data Handling Feature extraction from unstructured date Scalability Scaling to large data sets and complex problems
ML Applications
ML provides tools to analyze, visualize, and make predictions of data. Online shopping Netflix movie suggestions Spam mail warning Self-driving cars
Data Types Numerical Data - Measurable Data Categorical data - Characteristics Time series data - Number Sequence Text data - Words or paragraph
Numerical Data - Measurable Data Quantitative data Height or weight Whole numbers or continuous %, 12, 82.5 Categorical data - Characteristics Aspects like gender, ethnicity, and so on. Nominal - color of an eye or gender. (Does not have meaningful order) Ordinal - levels of difficultly like easy medium or hard. (Has natural order). Time series data - Number Sequence Sequence of number collected at regular intervals over a period. Average number of book sales / year. Text data - Words or paragraph Type of data leads to type of data & questions you can ask.
ML Examples Reinforcement
Reinforcement Automated robots Autonomous cars Video games
Supervised Learning - Regression Example
Spam detection Input Email content Subject Sender Output Spam or not Binary Disease detection Input Person medical details Output Diseases or not Binary Sentiment Analysis Input Comments Costumer reviews Tweets Output Positive Negative Multi class in nature Categorical output Stock Price prediction Input Volume trading How much Opening price Closing price Output Price Continuous and quantitative in nature
ML Examples Supervised
Supervised Disease detection Weather forecasting Stock price prediction Spam detection Credit scoring
Flavors of Machine Learning. Supervised Unsupervised Reinforcement
Supervised Classify data or make predictions. Unsupervised Understand relationships within data sets Reinforcement Make decisions or make choices
Supervised Model to Identify Fruits
Training Phase Training Data Set Input X Color Size Shape Output T Banana Apple Pineapple Training algorithm Training Loop Correlation between x and t Hypothesis and continues until error is small Trained Model Prediction Trained Model is as good as its data
ML Examples Unsupervised
Unsupervised Fraudulent transaction detection Customer segmentation outlier detection Targeted marketing campaigns