Week 1 Summary
1. Machine Learning
Field of study that gives computers the ability to learn without being explicitly programmed.
2. AI General Intelligence
- Building machines as intellgent as humans.
- Machines try to behave, mimic human brain.
3. Supervised Machine Learning
- Referes to Alogrithms that learn input to output mapping, x -> y mapping
- You provide learning algorithms expamples to learn from lables (right answers).
- Call it Supervised, baceause we try to supervise algorithm to give answers for a given input.
i) Regression
- Type of supervised learning in which a learning algorithm learns to predict a number out of infinte possible values.
ii) Classification
- Type of supervised learning in which we have to predict a category from a fixed set of possible values.
- Classification algorithm predict categories. Categories can be numeric or non-numeric.
4. Unseprvised Machine Learning
- Given data isn’t associated with any output lables.
- We are not asked to diagnose whether a Tumor is
Benign or Malignant. Instead, we have to find some pattern, structure, or something interesting in the data.
- We are not asked to predict output label.
- Call it unsupervised, baceause we are not trying to supervise algorithm to give right answers for a given input. Instead, we figure out what is interesting in the data.
i) Clustering
- Find groups, clusters in the data.
- We divide (un-labelled) data in different clusters based on similar characteristics.
- Group similar data points together.
-
Example <1> group customers into different market segments.
<2> categories of learners, e.g., growing skills, develop career, stay updated with AI.
ii) Anomaly Detection
- Find something unsual in the data.
- Find unsual data points, e.g., unusual transactions.
iii) Dimentionality Reduction
- Compress data using fewer numbers.
- Compress a large dataset into smaller dataset while loosing as little information as possible.
- Usually, resultant dataset has reduced features.
5. Types of Supervised Learning
- Regression
- Classification
6. Types of Unsuprvised Learning
- Clustering
- Anomaly Detection
- Dimentionality Reduction
7. Regression vs Classification Model
- Regression model > predicts numbers out of infinit possible values
- Classification models > predicts categories out of discrete, finite set of ouputs.
Linear Regression Model
- Model fits straight line to data.
i) How do you get a trained Model?
Training Set > Learning Algo > Model f(x)
ii) How Model predicts output?
Features (x) > Model f(x) > Prediction (y^)
iii) How to represent f(x)?
\[f_{w,b}(x^{(i)}) = wx^{(i)} + b \tag{1}\]