Chapter One: machine learning foundation

Foreword

This series of articles derived from reading "machine learning real" book study notes for recording learning content, constantly updated,,

Development in the 21st Century Internet, making it easier for us to collect data in such a large amount premise data, we have to dig out the value of huge amounts of data, and machine learning algorithms favored by many hobbies technicians because these algorithms to some extent, it can help us to apply some regression prediction, classification, identification, and many other scenes.

What data, information, knowledge?

Everything is kind of data, and the information is useful data

For example: by measuring the position of the planets and the corresponding time, we get that data. The trajectory data obtained through these planets is the message, summed up by this information is knowledge of Kepler's three laws.

Structure of the Book

"Machine learning combat" a total of 15 book chapters, mainly introduced the top ten of its data mining algorithms Eight:

C4.5 decision tree, K- mean (K-mean), support vector machine (SVM), Apriori, expectation-maximization algorithm (EM), PageRank algorithm, AdaBoost algorithm, k- nearest neighbor (kNN), Naive Bayes algorithm (NB) classification and regression tree (CART) algorithm. It does not include the expectation-maximization algorithm and PageRank algorithm. This book does not include PageRank algorithm, because the search engine giant Google introduced PageRank algorithm has been fully discussed in many works, there is no need for further tired; and expectation-maximization algorithm was not included, is because it involves too much

The reason for choosing this book

The book's actual combat is relatively strong, and each chapter is illustrated with examples, in line with my classmates to write code like this, if I could look at a watermelon book fell asleep, too much theoretical knowledge.

My plan is to book through a few of these algorithms select watermelon books or courses of study CHANG after a certain understanding.

As the realization python2.x are used in the book, I will use python3.7 to reproduce and display it with jupyter.

What is machine learning it?

Machine learning is a general term for a class of algorithms, these algorithms attempt to dig out from a large number of historical data which implied the law, and to predict or classify, more specifically, machine learning can be seen as looking for a function, the sample data is input the output is the desired result, but this function is too complex, that is not convenient formal expression. It should be noted that the goal of machine learning is to learn the function is well suited for the "new sample", rather than just performing well in training samples. Learn the ability to apply to the new sample function called generalization (Generalization) capability.

In short, machine learning can reveal the true meaning behind the data

Artificial intelligence, machine learning, deep learning

Artificial intelligence is the pursuit of goals, machine learning is a means of achieving deep learning is one way.

AI ** (Artificial Intelligence) **: machine gives wisdom

In 1956, several computer scientists gathered at the Dartmouth meeting, proposed the concept of "artificial intelligence", the dream machine with a computer had just appeared to construct complex, and have the same essential characteristics of human intelligence
after 2012, too benefit from the rise in the amount of data to enhance the computing power and the emergence of new machine learning algorithm (deep learning), the AI outbreak began.

Now the field of artificial intelligence research as long as the expert systems, machine learning, evolutionary computation, fuzzy logic, computer vision, natural language processing, recommendation systems, etc.

Weak AI - Artificial Intelligence strong

We are talking about artificial intelligence are weak artificial intelligence, to achieve a certain kind of specific tasks, such as face recognition, machine translation.
Wall AI probably are like people, right, to learn everything myself

Machine Learning (Machine Learning): means for implementing artificial intelligence

The most basic approach of machine learning, is to use algorithms to parse the data, learn, and then make decisions / predictions of real-world data / event.

According to an algorithm algorithms, machine learning may include: decision trees, clustering, support vector machines, Naive Bayes and so on.

According to learning different machine learning can be divided into: supervised learning, semi-supervised learning, unsupervised learning, integrated learning.

Deep learning (Deep Learning): A method of machine learning implementation

Machine learning, the feature extraction is not simple. Project features often require a lot of time to optimize, but this time, the depth of learning can automatically learn the association between features and tasks, but also complex features extracted from simple features. deep

Learning from the concept of artificial neural networks, multilayer perceptron containing multiple hidden layer structure is a deep study, showing the depth attribute category learning or more abstract features formed by the combination of high-level low-level features, to find the distribution of the data represented by the formula wherein

Three links

Artificial intelligence, machine learning >> >> depth study

supplement

Machine learning tasks personal understanding is that the main job classification and regression. Classification and Regression belong supervised learning, is called a supervised learning, because these algorithms must know to predict what that classified information of the target variable.

Corresponding to the supervised learning unsupervised learning, then there is no data type information, do not give definite target. For discrete traversal

Select the appropriate algorithm:

If you want to predict the value of the target variable, you can select supervised learning algorithm, or you can choose unsupervised learning algorithm. After determining the selected supervised learning algorithm is needed to determine the target variable type, if the target variables are discrete, yes / no, 1/2/3, A / B / C or red / yellow / black and the like, may be selected classifier algorithm ; If the target variable is a continuous value, such as 0.0 ~ 100.00, 999 ~ 999 or the like + ∞ ~ -∞, it is necessary to select the regression algorithm

Discrete: Classifier

Continuity: logistic regression

reference

Introduction to Machine Learning

Know almost - the difference between the depth Study - Artificial Intelligence - Machine Learning