Machine Learning - Landscape (1)

I have recently started Data Engineering - Data Science Masters Degree and I have been studying "Machine Learning".  

As part of this subject I have to read the book: "Hands-on Machine Learning with Scikit-Learn, Keras, and TensorFlow, 2nd Edition"

I am going to use my blog as summary of this subject and the other Master ones. Lets go!

How to define "Machine Learning"

Here two approachs:

- Is the science (and art) of programming computers so they can learn from data.

- Is the field of study that gives computers the ability to learn without being explicitly programmed.

Machine Learning is great for...

- Problems for which existing solutions require a lot of fine-tuning or long lists of rules.

- Complex problems for which using a traditional approach yields no good soution.

- Fluctuating environments: ML can adapt to new data.

- Getting insights about complex problems and large amount  of data.

Type of Machine Learning

Supervised

The training set you feed to the algorithm includes the designed solutions, called labels.

a) Classification: i.e. spam / ham.

b) Predictors: predict the target numeric value, given a set of features.

This sort of task is called regression.

Some algorithms: k-nearest neighbour, Linear Regression, Logictics Regressions, Support Vector Machines (SVMs), Decission Trees and Random Forest, Neuronal Networks.

Unsupervised Learning

The training data is unlabeled.

- Dimensionality Reduction: the goal is to simplify the data without losing too much information. I.e. merge several features into one is called Feature Extraction. Is recommended due to improve the performance.

- Amount Detection: if sees a "new" instance can be classified as "anomaly".

- Novelty Detection: it aims to detect new instances that look different from all instances in the training set.

- Association Rule Learning: the goal is to dig into large amounts of data and discover interesting relations between attributes.

Semisupervised learning

Algorithms that can deal with data that is partially labeled.

Reinforcement learning

- Agent: is a learning system. Can observe the environment, select and perform actions, and get rewards in return.

Batch and online learning

- Batch: the system is incapable of learning incrementally: it must be trained using all the available data.

- Online: we train the system incrementally by feeding it data instances secuentially, either on in small groups called mini-batches.

- Out-of-core learning: when the Online learning is used to train systems on huge datasets that cannot fit in one machine´s main memory.

- Learning rate: how fast they should adapt to changing data.

Thats all for today!