Learn-Machine-Learning

My adventures in learning machine learning

Machine learning - Subfield of AI.

AlphaGo

Machine Learning is where computers can learn from examples and experience

Classifier - function. takes data as input and asigns label as output. Classify email as spam or not spam. Classify image as apple or orange

Supervised learning - Create a classifier by finding patterns in examples

Unsupervised learning - Unsupervised learning is a type of machine learning algorithm used to draw inferences from datasets consisting of input data without labeled responses. The most common unsupervised learning method is cluster analysis, which is used for exploratory data analysis to find hidden patterns or grouping in data.

helps find previously unknown patterns in data set without pre-existing labels.

Steps:

Collect training data
Train Classifier
Make Predictions

Feature - input to the classifier

Label - output we want

Decisions tree

Classifier - box of rules. many types. input and output always same.

If classifier is a box of rules. A Learning Algorithm is the procedure that creates the rules by finding patterns in the training data. It may notice orange weighs more, so it creates a rule that the heavier the fruit, the more likely it is to be an apple.

Choosing good features is one of your most important jobs.

Model - protoype that define rules of classifier function. has adjustable parameters.

y = mx + b

Learning - using traning data to adjust parameters of a mdoel

Neural network - more sophisticated type of classifier. it is like a decision tree or simple line.

KNN (K-nearest-neighbor) - Look at closes training point. straight line distance between two points.

Euclidean Distance - Uses pythagorem theorem distance formula. Works same no matter dimensions. 2D or 4D (image data?)

k - number of neighbors to consider

Deep learning - branch of machine learning

Generative adversarial Network (GAN) - A generative adversarial network (GAN) is a class of machine learning systems. Given a training set, this technique learns to generate new data with the same statistics as the training set. For example, a GAN trained on photographs can generate new photographs that look at least superficially authentic to human observers, having many realistic characteristics.