Fifth week (web, machine learning notes)

2019/11/2

1, the performance state transition layer (REST, representational state transfer.) One kind of web software architecture style object is to facilitate different software / program over a network (e.g. the Internet) and to transmit information to each other. Presentation layer state transition is based on the root of the Hypertext Transfer Protocol (HTTP) a set of constraints and attributes determined above, designed to provide a World Wide Web services software build style . Meets or compatible with this architecture style (referred to as REST or RESTful) web services, allowing clients to issue uniform resource identifier to access and manipulate network resource request, with the pre-defined set of operations stateless harmonization. Thus the performance state transition layer provides coordination properties (Interoperability) between network computing system, the resource may be used interchangeably with each other.

REST is a style rather than a standard. REST is usually based on the use of HTTP , URI of the , and XML and HTML protocols and standards these existing widely popular.

  • Resources are specified by the URI.
  • Operation of resources, including access, create, modify, and delete resources, these operations correspond exactly GET, POST, PUT and DELETE method of the HTTP protocol provides.
  • Resources to operate through forms of operating resources.
  • Manifestation of the resource is XML or HTML, depending on the reader is a machine or person, the consumer web services client software or web browser. Of course, it could be any other format, such as JSON.

REST architectural style of the most important architectural constraints there are six [2] :

  • Client - Server (Client-Server)

The client - server architecture constraints purpose is to separate concerns the client and server-side logic will be separated from the user interface of interest logic and data storage concerns to help improve cross-platform portability of the user interface. by simplifying the server module also contributes to scalability of the server module.

  • Stateless (Stateless)

The server can not save the information of the client, every request from a client sent, to include the status information for all of the necessary session information stored by the client, the server to process the request according to the status information server session status information transmission additional services, such as database services, such state information may be maintained for some time, thereby realizing the authentication function when the client can switch to a new state of the transmission request information. when one or more requests to be sent, the client in a state transition process. each state of the application description may be used to initialize the client the next state transition.

  • Cache (Cacheability)

As the World Wide Web, the client machine and the middle of the communications transmitters response can be cached response must be clear or indirectly indicate whether itself can be cached, which can prevent the client to get stale in time in the future the requesting or inappropriate data good management can reduce the client caching mechanism - the interaction between the server, or even completely avoid client - server interaction, which further improve the performance and scalability.

  • Unified interface (Uniform Interface)

Unified interface is the basic starting point RESTful system design that simplifies the architecture of the system, reducing the coupling allows all modules to improve four separate bound for the uniform interface is..:

  • ID (Resource identification in requests) containing the resource request

Request includes an identification of various independent resources, for example, URIs in a Web service resource itself and the identity sent to the client is independent. For example, the server may own database information is sent to the customer to HTML XML or JSON way end, but these may not be the internal recording server.

  • Operating resources (Resource manipulation through representations) by identifying

When the client has to identify a resource, including metadata that comes, it will have enough information to delete the resource.

  • Self descriptive message (Self-descriptive messages)

Each message contains enough information to describe how to handle this information. For example, media type (midia-type) can determine what is necessary to analyze the media data analyzer.

  • A hypermedia application state drive (Hypermedia as the engine of application state (HATEOAS))

Users with access to the Web server Home page is similar to that when a REST client access to the URI after the initial application of REST, REST client should be able to use the link provided by the server, dynamic resource discovery and perform all actions available with access to the server provides in the response text hyperlinks, so that clients can obtain the operation currently available client applications without dynamic configuration information provided by the server to record the determined coding manner.

  • Layered system (Layered System)

The client usually does not know whether to connect directly to the final server, or a server on the middle path intermediate server can improve system scalability through load balancing and shared cache mechanisms, which can also facilitate the deployment of security policies.

  • Code on demand (Code-On-Demand, optional)

The server can send executable code to customers by the end of the way temporary extensions or customization capabilities such as Java Applet, Flash or JavaScript.

2、  Building an Ensemble Learning Model Using Scikit-learnEnsemble learning uses multiple machine learning models to try to make better predictions on a dataset. An ensemble model works by training different models on a dataset and having each model make predictions individually.

3,  Gradient Boosting  IS A Machine Learning Technique for Classification and Regression Problems, Which Produces A Prediction Model in form of AN at The Ensemble of weak Prediction Models, typically Decision Trees. Gradient lifting is a machine learning technique for regression and classification problems, it is set to generate a prediction model weak prediction model (usually tree) form. Like other enhancement methods, which build the model in a phased manner, and to allow them to summarize the loss of any differentiable function optimized by. Thought observed enhanced gradient from Leo Breiman, i.e., may be interpreted as a reinforcement for the appropriate optimization of the cost function.

4,  lifting method (on Boosting), can be used to reduce a supervised learning of the deviation of the machine learning algorithm.

5, AdaBoost, English "Adaptive Boosting" (Adaptive Boosting) acronym. AdaBoost method is adaptive in that: a sample before classifier misclassified will be used to train a classifier next. Data is sensitive to noise and data anomalies. AdaBoost method is an iterative algorithm, adding a new weak classifiers in each round, until it reaches a certain predetermined sufficiently small error rate.

6, due to the recursive cause a series of function calls, and there may be a series of double-counting, a recursive implementation of the algorithm efficiency is relatively low.  Iterations : the original value calculated using the variable a new value of a variable if recursion is to call their own own words, iteration is constantly call B. a so-called recursively , in a nutshell is the application itself calls itself, in order to achieve the level of access and query data structures.

7,  enumeration (enumeration), in mathematics and computer science theory, an enumeration set is to list some programs for all members of a finite set of sequences, or a particular type of object count. Both types often (but not always) overlap. Enumeration is a collection of an integer constant is a named! Traversing (Traversal), is an important operation of the tree structure, the tree structure accessing means that according to certain rules in each node, and each node is visited only once. Iteration (iterate), refers to each one by one in a certain order to access the list. Iterative repetition active feedback process, the purpose of which is usually close to and reach the target or desired results. Each repetition of the process is referred to as an " iteration ", and each time iteration results obtained will be used as the next iteration initial value. Cycle (loop): refers to the condition is met, the same code is repeatedly performed. General language, there will be three types of loop: for statements, while statements and do While statements. Recursive function (Recursion function): calling at their own situation constituted a function by name. Recursion (recursion) in mathematics and computer science, it refers to the use of the method in the definition of the function itself functions. The term is frequently used to further recursive self-similar procedure described process is repeated things.

8,  Recursion: recursion , problems typical example, factorial, Norway sized dish (Hanoi Tower?)

Guess you like

Origin www.cnblogs.com/xkeepgoing/p/11785112.html