Machine learning
machine learning is a multi-disciplinary field cross, can be classified from the learning mode and learning methods above, the learning mode will be classified as machine learning, supervised learning, unsupervised learning and reinforcement learning, machine learning methods can be divided into the traditional machine learning learning and deep learning.
Machine learning is divided into supervised learning by learning, unsupervised learning and reinforcement learning.
Machine learning is one involving statistics, system identification, approximation theory, neural networks, optimization theory, interdisciplinary fields of computer science, brain science. It examines how computer simulation or realization of human learning behavior in order to gain new knowledge or skills, re-organize existing knowledge structures and keep them coming. To improve its performance is at the heart of artificial intelligence technology. Based on the data of machine learning is one of the most important modern intelligence methods. Depending on the mode of learning, learning methods and algorithms, machine learning have different classifications.
Mapping knowledge
is knowledge map is essentially a semantic network (Semantic Network) knowledge base. "But a little abstract, so another angle, from the perspective of the practical application of knowledge can in fact be simply understood as a multi-map diagram.
In essence, knowledge Atlas of the semantic knowledge structure is a diagram showing a data structure composed of nodes and edges, the concepts described in symbolic form and their relationships in the physical world, which is a basic unit "entity - relationship - entity" Three group, and its related entities.. "attribute - value" between the different entities linked to each other through relationships, knowledge structure forming a network in the knowledge graph, each node represents the real world "entities", each side is "relationship" between entities and entity. More simply, knowledge is to map all the different kinds of information together and get a network of relationships, provides the ability from the perspective of "relationship" to analyze the problem.
knowledge maps available to fraud, inconsistencies verification, fraud and other groups of public safety and security areas, the need to use exception analysis, Static analysis, dynamic analysis, data mining methods. In particular, knowledge maps have a great advantage in the search engine, visual display and precision marketing has become a popular tool for the industry. However, the development of the knowledge map as well as a great challenge such as noise data problem, namely the existence of redundant data itself or the data error. with the deepening of the knowledge map application, as well as a series of key technologies needed to break through.
Natural language processing
natural language processing is an important direction of computer science and artificial intelligence in the study to achieve a variety of theories and methods of effective communication between people and computers in natural language, more involved in the field, including machine translation, reading comprehension and quiz machine systems.
Machine Translation
Machine translation technology refers to the use of computer technology to the process of translation from one natural language into another natural language. The limitation of prior rule-based translation methods and instance-based statistical machine translation, the translation performance made a huge boost. Based Machine Translation depth of neural networks in everyday language and some scenes of successful applications has shown great potential. With the development of contextual knowledge representation and reasoning ability of context, natural language mapping knowledge expanding, machine translation will make more progress in several rounds of dialogue translated and Text Translation and other fields.
Semantic understanding
semantic understanding technology is the use of computer technology to achieve understanding of the text of the chapter, and answer questions related to the process of the chapter. Semantic understanding more focused on understanding the context as well as answers to precise extent of the control. With the release of MCTest data sets, semantic understanding to receive more attention, and achieved rapid development, and related data sets corresponding neural network model after another. Semantic understanding technology in the smart customer service, product QA and other related fields play an important role, to further improve the accuracy of the Q & A session with the system.
Answering System
Answering System Answering System dialog system is divided into open areas and in specific areas. Q system technology is to make computer technology to communicate in natural language with people like human beings. People can submit questions using natural language to express answering system, the system returns a higher relevance answer. Often used in corporate smart phone robot under development, to find one to one statement based on an index keyword identification and reptiles, and then reply. There has been a lot of question answering system applications appear, mostly in practical application of information systems and services in areas such as smart phones Assistant, which is a relatively mature field of artificial intelligence, the market continues to expand at the same time, opportunities and challenges coexist, a smart phone really easy to use robot also needs four major challenges facing NLP:
Different levels of lexical, syntactic, semantic, pragmatic and voice uncertainty;
2. new vocabulary, terminology, semantics and syntax lead to unpredictability unknown language phenomena;
insufficient resources make it difficult to 3. Data covering the complex linguistic phenomena;
fuzzy correlation and complex semantic knowledge 4. difficult description simple mathematical model, semantic parameters calculated nonlinear calculation requires enormous
Human-computer interaction
human-computer interaction is an important peripheral technology in the field of artificial intelligence. Exchange of information between the main people and computers, including the exchange of information between people and computers, and computer people. HCI is a comprehensive discipline closely related to cognitive psychology, ergonomics, multimedia technology and virtual reality technology. Traditional exchange of information between people and computers primarily rely on interaction device, including a keyboard, mouse, joystick, data clothing, eye tracker, a position tracker, data glove, pressure pen input device, and a printer, plotter, display, HMD, output device such as a speaker. Man-machine interaction technology in addition to traditional basic interaction and graphical interaction, also includes voice interaction, emotional interaction, interaction and somatosensory brain-computer interface.
Computer Vision
Computer vision is the use of computer simulation of the human visual system science. It enables computers to extract, process, and to understand and analyze the image sequence with images similar to the human. Car driving, robotics, intelligent medical fields need to extract and process information from the visual signals by computer vision technology. In recent years, with the development of deep learning, preprocessing, feature extraction, and the gradual integration algorithm processing, a formed end artificial intelligence algorithms technology. The problem solved, computer vision can be divided into five categories: calculating imaging, image understanding, three-dimensional vision, dynamic vision and video codecs.
At present, the rapid development of computer vision technology, with initial industrial scale. Future development of computer vision technology will face the following challenges:
First, how to better integrate different applications combined with other techniques, computer vision can be widely use big data to address some of the issues that has gradually matured, can transcend human but on certain issues can not achieve high precision
degree.
Secondly, how to reduce development time and manpower cost computer vision algorithms. Currently, computer vision algorithms require large amounts of data and manual tagging, and it takes a long period of research and development in order to achieve the required accuracy and time-consuming in the application domain.
How to speed up the design and development of new algorithms. With the emergence of new imaging hardware and artificial intelligence chip, different chips for computer vision algorithms and data acquisition equipment design and development it is also a challenge.
Biometrics
Biometrics refers to recognize and identify the individual identity by physiological or behavioral characteristics of the individual. From the application process, the biometrics registration and identification are usually divided into two phases. In the registration phase, the biometric information acquisition sensor body extraction technology of fingerprint, face and other optical information, microphone acoustic information such as voice, data preprocessing and feature, and stores the corresponding features.
In the identification process, according to the registration process information collection, data preprocessing and feature extraction, and the extracted stored identification with the features as compared to complete identification. From the viewpoint of application tasks, generally divided biometric identification and validation tasks. Recognition is the process of identification of the person to be identified from the repository, which is one to many problems. It refers to identification information stored in the repository specific personal information by comparing the recognition process to be identified person.