In the AI ​​boom, the voice of the neural network was the biggest. However, AI is far more than that.
At present, in the field of AI technology, the most funded research is the study of neural networks. In the eyes of everyone, neural network technology looks like a "programmed brain" (although metaphor is inaccurate).
The concept of neural network was proposed as early as in the 1940s, but until now, people still know very little about the working methods of neurons and brains. In recent years, the research community has become increasingly strong about the innovation of neural network technology. The desire to restart the neural network...
In fact, in addition to neural networks, there are many more interesting, newer, and more promising technologies in the AI ​​field. These technologies are introduced to everyone in the article.
1. Knol extractionKnol refers to information units, that is, keywords, words, etc. Knol extraction technology is the process of extracting key information from text. To give a simple example: For example, "As the name implies, the octopus has 8 legs." After this sentence is extracted, it becomes like this: {"octopus": {"number of legs": 8}}.
Our commonly used Google search engine relies on this technology, and many of the technologies described later include this technology.
2. Ontology constructionOntology construction is based on NLP technology and aims to use software to construct the hierarchical structure of entity nouns. This technology is very helpful for AI conversations. Although the ontology construction surface looks simple, in fact it is not easy to build, mainly because the actual connection between things is more complicated than we think.
For example, use NLP analysis text to create an entity relationship set:
Example: "My labrador has just given birth to a group of puppies. Their father is a poodle, so they are Labrador Poodles (a hybrid dog)." It became: {"Puppy å´½": {"May be": "Labrador Poodle", "Having/Having": "Father"}, "Labrador": {"Owned / "Have": "Puppy å´½"}}.
However, human beings usually do not state all the relations in their language expressions. For example, in this sentence, it is necessary to infer through the fact that “My Labrador is a femaleâ€. This is the fact that Difficulties with ontology construction.
As such, ontology construction technology is currently only applied to top-notch chatbots.
A heuristic is a rule for classification, usually like a conditional statement such as "If this item is red" or "If Bob is at home," these conditional statements are often accompanied by an action or decision, such as:
If the "[component]" attribute of something contains "arsenic": then its ["poison"] attribute is "True."
For each new piece of information, new heuristics and new relationships accompany it. With the establishment of new heuristics, new understandings of related terms can be generated. such as:
Heuristic One: "puppies" (Puppy) shows that they are Babies;
Heuristic II: Babies are very young;
The above two heuristics infer: "puppies" are very young.
The heuristic difficulty is that in most cases the rules are not as simple as "If/Then". Similar phrases such as "some people whose hair is golden" are difficult to express in terms of heuristics. So we have a "cognitive theory" (see below).
4. EpistemologyEpistemology is a combination of ontology construction and custom heuristics, and incorporates probabilistic properties to represent the possibility of a noun being associated with any attribute. For example, with this ontology structure:
{'Person': {'sex': {'male': 0.49, 'female': 0.51}, 'race': {'Asian': 0.6, 'African': 0.14}}
To express the judgment of one's gender and race. At the same time, probability can help identify some “hybrid†phrases with multiple meanings, such as the phrase “plums are like raisins with hormones,†because the phrase “beating hormones†probably means “ "Large volume," and therefore concluded that this sentence is likely to mean "Plums are bigger than raisins."
The realization of epistemology is much more difficult than the construction of ontology. First, it needs more data; and, due to the complexity of its structure, it is difficult to quickly establish a database to find after determining the rules; also, the determination of the rules is usually based on a certain thing in a paragraph of text However, the text may not truly reflect the reality.
Epistemology is similar to the "tensor flow" theory proposed by Asimov. The TensorFlow system of the same name developed by Google is not really based on tensor, but epistemology is based on tensor.
5. Automatic gauge technologyA gauge system must include the corresponding evaluation criteria. Imagine that when choosing a house, factors such as the size, location, price, and style of the house need to be considered. These factors may not necessarily be positive. This requires decisions to be made by measuring trade-offs. For example, if you care more about the size of the house than the price, you would rather spend several times more money to buy a big house.
The self-assessment technique determines the weight of each factor through your emphasis on different factors, thus making recommendations for decision making. Through this process, it is also possible to predict inventory changes, recommend products, and achieve automatic driving. That is to say, most neural networks can achieve the function, auto-gauge technology can do the job, although it takes longer training time, but it has a decision speed of several orders of magnitude faster.
6. Vector differenceVector difference techniques are often used for image analysis and also for processing time-varying data. By constructing an abstract vector image of the target, the candidate object is compared with the target object to be identified, thereby determining whether it is "best dating face type" or "best time to buy".
In general, differences between target objects are accompanied by a quantitative rule that measures the degree of difference. Through the vectorization of features, some “fuzzy†concepts are simply and clearly represented.
For example, for humans, we generally believe that symmetrical faces are more attractive, but for computers, we need accurate calculations to judge, and at this time, facial abstraction is performed through 30 triangles, rather than through complete faces. Comparing the images for calculation can save a lot of computing time and storage space.
The processing of non-image data is also possible. For example, the stock price changes, the ratio of earnings per share to margin, and so on, through the vectorization of these data, and compare it with the ideal value, you can determine the degree of positive investment or risk.
7. Matrix ConvolutionConvolution matrices are commonly used in edge detection and contrast enhancement in the image processing field. For example, many of the filters in PhotoShop are based on convolution matrices or superposition convolutions where multiple convolutions are performed in a specific order.
At the same time, the convolution matrix can also be used to process non-image data. For example, when the convolutional matrix is ​​used to process the timing vector, the pattern can be found as quickly as the edge detection, and then a specific value or range can be found at the minimum or maximum value to make a judgment.
8. Multi-view decision systemMaking a decision is not simple. The multi-view decision system makes decisions in a more democratic manner and in many ways.
For example, in the case of the house just now, your preference for a house may be based on factors that are not comprehensive, followed by the fact that “this house is built on a cliff†(of course, this overwhelming factor may come from Knol extraction) will eliminate all your previous goodwill and allow you to make decisions again.
Therefore, decision-making needs to be considered through more comprehensive factors, and a multi-view decision system can use two sets of criteria (such as you and your spouse) to measure decisions. The multi-viewpoint decision system can also be applied to the field of automated driving, for example, collecting the views of 10,000 vehicle owners to formulate new standards.
Written at the end - to believe that technology does not weigh
Many people have only one tool in their eyes and fall into the pit of “I have a hammer, so everything is a nailâ€. Companies such as Recognant, while applying neural networks, also apply these relatively unpopular techniques in the article, after all, compared to neural network hardware systems.
The advantage of these software technologies is that they can be adjusted and developed at any time for different situations without additional costs. Therefore, if the technology is narrow, it may be trapped by some situations. The wider the technology, the easier it will be to face problems.
High Voltage Power Resistor are designed to withstand and dissipate large amounts of power. Usually the releated power is at least 5watts and up to 500watts. It has Excellent high-frequency characteristics and suited for Radar system and Broadcasting transmit devices, .etc.
High Voltage Polypropylene Capacitor,High Voltage Tubular Resistor,HV thick-film Resistor,Thick Film Power Resistor
XIAN STATE IMPORT & EXPORT CORP. , https://www.shvcomponents.com