* Supervised Learning
Given a set of data points {x(1),.....x(m)} associated to a set of outcomes {y(1),....y(m)}we want to build a classifier that learns how to predict y from x
### Hypothesis:
In the classification problem there exists a “ground truth” function 'f' specifying the correct classification of every possible data point. 'f' is initially unknown and our learners job is to figure out this function. We assume we are given 'n' data points drawn independently and identically (“i.i.d.”) from the distribution 'D' along with corresponding labels that represent ground truth. We need to come up with a learning algorithm that uses these data points as clues and outputs prediction function g, which is our best guess as to what actual 'f' might be.
But we want to generalize beyond the training example so that our learned function can perform well on future unseen examples as well. So what “Learning” really means is that we take the given examples and extrapolating so that g coincides with 'f'.
"The ability to perform well on previously unobserved inputs is called generalization".
### Loss Function:
### Cost Function:
### Optimization using Gradient Descent:
* Unsupervised Learning e.g Clustering, dimensionality reduction
## Data:
(x,y) learn a mapping from x -> y
Learning a Hypothesis function
Machine Learning Types and Applications:
### Applications
Common Problems:
- Image Classification
- Text Classification
- Text Summarization
- Question Answering, Chat bots - NLU
- Clustering: Segmenting Customers for marketing strategy