Machine & Deep Learning overview

This section begins the Basic theory system learning phase of machine learning and deep learning, and the blog content is the notes that are collated during the learning process.

1. Machine learning

Concept: Multi-disciplinary interdisciplinary, involving probability theory, statistics, approximation theory, convex analysis, algorithm and complexity theory and other disciplines. Specialized in computer simulation or realization of human learning behavior, in order to acquire new knowledge or skills, reorganize the existing knowledge structure to continuously improve their performance.

Discipline positioning: The core of artificial intelligence, is the fundamental way to make the computer intelligent, its application throughout all fields of artificial intelligence, mainly use induction, synthesis, rather than deduction.

Definition: Explore and develop a series of algorithms to how a computer can learn, model, and use built-in models and new inputs to predict a subject without explicit external instructions.

Learning: For experience E (experience) and a series of task T (tasks) and a certain performance of the measurement p, if with the accumulation of experience e, for the defined task T can improve performance p, said the computer has the ability to learn. For example: Chess, speech recognition, auto-driving cars and so on.

Applications: Speech recognition, autonomous driving, language translation, computer vision, recommender system, UAV, identification of spam ...

2. Deep learning

Deep learning is a new field based on machine learning, which is derived from neural network algorithms inspired by human brain structure, coupled with the increasing development of the depth of the model structure, and a series of new algorithms resulting from the improvement of big data and computing power.

The concept by the famous scientist Geoffrey Hinton and other people in the 2006 and 2007 in the "Sciences" and other articles published by the proposed and rise.

Application: Deep learning, as an extension of machine learning in the field of image processing and computer vision, natural language processing and speech recognition, since 2006, academia and industry cooperation in deep learning research and application in the field has made breakthrough progress. An example of the object recognition competition in the classic image of Imagenet, defeating all the traditional algorithms, has achieved unprecedented accuracy.

Representative academic institutions and companies, the University of Toronto, New York University, Stanford University as the representative of the industry to Google, Facebook and Baidu as the representative, walking in the forefront of deep learning research and application.

In the current Android phone we use, Google's voice recognition, Baidu View, Google's image search, has been used in deep learning technology. In the era of big data, combined with the development of deep learning in the future of the impact of our lives immeasurable, conservative, many of the activities currently engaged in human milk because of deep learning and the development of related technologies are replaced by machines, such as autonomous driving, unmanned aircraft, and more intelligent robots. The development of deep learning allows us to see and approach the ultimate goal of AI for the first time.

3. Some basic concepts in machine learning and deep learning

(1) Basic concepts: Training set, test set, characteristic value, supervised learning, unsupervised learning, semi-supervised learning, classification, regression

(2) Concept Learning: This Boolean function is inferred from the input and output training sample for a Boolean function.

(3) Example

The concept is defined on the instance (instance) collection, which is represented as X (x: All possible days, the value of each day is represented by 6 properties of weather, temperature, humidity, wind, water temperature, forecast, etc.).

The concept or objective function to be learned becomes the concept of the target, which is written in C. When enjoying the movement is C (x) = 1, when not enjoying the movement is x (x) =0,c (x) can also remember to do Y.

X: every instance

X: Sample, a collection of all instances.

Learning Goal: F:x->y

(4) Training set (training Set/data): Used for training, which is the data set that produces the model or algorithm.

Test set: A data set used specifically to test a model or algorithm that has been well learned

Eigenvectors: A collection of attributes, usually identified by a vector, attached to an instance

Tags: tags for instance categories

　　

(5) Classification problem: Target is marked as Category type data

Regression problem: The target is marked as a continuous value.

(6) Supervised learning: The training set has a category tag,

Unsupervised learning: No class marking of training data

Semi-supervised learning: Training data includes training sets with category tags + training set without category tags

(7) Machine learning steps

① split the data into training and test sets

② training algorithm with feature vectors of training set and training set

③ uses the learning algorithm to evaluate the algorithm in the test set, which involves adjusting the parameters (used in the validation set)

The above is a conceptual introduction to machine learning and depth, starting with the next entry into each specific algorithm, including theoretical basis and practical application of the two aspects of learning.

Machine & Deep Learning overview