JP2003006171A - System for entrusting parallel processing calculation via the internet - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a system, capable of a large scale parallel calculation by distributedly processing parallel processing calculation tasks requested by a plurality of customers with many client computers in the background processing during service processing in accepting an order for the calculation tasks from customer computers of the plurality of customers entrusting the calculation tasks via the Internet and respectively providing information via the Internet, in response to service requests from the client computers of many users utilizing an information providing service in a host computer of an information providing service provider via the Internet. SOLUTION: This system is provided with a means, which randomly generates programs with a short operation time in solving calculation tasks, independently performs an individual calculation, on the basis of an initial value and can perform calculation processing of Monte Carlo type optimal solution search, that performs a large number of execution times as a whole, so that a client computer can cope with the situation where the communication time the client computer receives the information service is indeterminate.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention, in a host computer of an information providing service provider via the Internet, accepts via the Internet from customer computers of a plurality of companies / individual customers, etc. entrusting parallel processing calculation tasks, In response to a service request from each client computer of many users who use the information providing service, when the respective information is provided via the Internet, the calculation requested by the customer in the background processing during the service processing. The present invention relates to a meta-computation entrustment system capable of performing a large-scale parallel processing calculation by performing distributed processing on a large number of client computers.

[0002]

2. Description of the Related Art In metacomputing (distributed computing utilizing computer resources such as PCs distributed all over the world) that performs distributed processing by a plurality of computers to perform large-scale parallel processing, how is the calculation performed? Many clients (P used by users
A big issue is whether or not to connect to C). To date, there have been cases where millions of computers have been secured for themes such as cryptanalysis and data analysis for extraterrestrial intelligence search with the cooperation of volunteers. However, it has been extremely difficult to secure such a large number of volunteers in a large-scale calculation of normal work.

Although there are already services that try to make a profit by using meta-computing, the consideration for the computing resource provider is direct payment in cash. However, since payment in cash requires remittance and the like, the amount of money to be paid to the computing resource provider is small with a small amount of settlement, and it is difficult to secure enough persons who will provide the computing resource. If sufficient computing resources cannot be secured, this leads to a vicious circle in which people (customers) who want to use the computing resources to perform huge calculations do not gather.

In addition to the conventional information system in which a large number of clients are connected to one host to receive information services, in recent years, peers have exchanged information with each other.
Two-peer type information systems are widespread. Since a peer-to-peer type information system does not require a specific host, various information services such as information exchanges that meet the diverse needs of individuals occur spontaneously, and a large number of computing resources can be secured as a whole. There was a possibility. However, it was not easy to pay the price among individuals because there was no specific host. Therefore, at present, the use is limited to the exchange and sharing of music files and the like. Furthermore, with this method of file exchange, the file that you want is not provided to the other party, and you will be provided with a one-sided file. There has occurred. Therefore, the file provider often abruptly interrupts the connection immediately before the completion of downloading the file, which unnecessarily increases the network traffic.

As a general method for resolving such dissatisfaction of the information service provider, there is widely used a technique of presenting a banner advertisement to a client, and the information service provider obtains a compensation from the advertisement revenue. But,
An information service provider can earn sufficient advertising income only when a certain number of clients who have seen the banner advertisement actually purchase the sponsor's product and the sponsor can generate sufficient revenue. Therefore, it was not a fundamental solution.

[0006]

Recently, the number of computers connected to the Internet has increased exponentially, but most of them are idling. Therefore, we have devised a method of efficiently utilizing this unused resource for metacomputing. This method can be applied to ecosystem model construction and various applications that require large-scale calculation. In addition, in information service providing sites such as current search sites and video distribution sites, the business model of "delivering banner advertisements to users in return for providing information services to users" is the mainstream. Aims to realize a new business model of "obtaining computing resources of idle computers from users in return for providing information services to users", and to maintain a system for large-scale computation and its utilization services almost semipermanently. It is technically possible. Therefore, there are the following problems to be solved.

That is, in order to secure a large number of computers without expecting cooperation from volunteers, a method of collecting computing resources from clients as a consideration of the information service provided by the host and an incentive for connecting as many clients as possible are provided. A mechanism for giving is necessary.

The computational resources provided by the client are
Since there are great variations in terms of calculation speed and connection time,
There is a need for technology that can effectively use this uneven computing resource.

In particular, the time during which the client receives the information service is indefinite, and it is totally impossible to predict when the connection with the host will be disconnected. In order to effectively use this indefinite connection time, it is necessary to utilize the calculation result even if the calculation time of the program executed by the client is short. In addition, it is necessary to prevent the calculation made by the client from being wasted under the constraint that it cannot predict when the connection from the client will be disconnected.

When the client is downloading a huge file such as an image, the user may have time to finish downloading. It is possible to regard that having the user perform work such as information processing by using the free time can be regarded as the same calculation resource as the calculation using the free time of the CPU.

For example, pattern recognition performed by the user looking at the information displayed on the screen is a computational resource that the client can provide from the host side. Further, in the search calculation of the optimum solution, the search efficiency may be remarkably improved by appropriately checking by humans.

However, in order to secure a large number of users who participate in such work, it is necessary to provide a mechanism for giving an incentive to encourage the participation of the users.

It is necessary to prevent fraud between the host and the client and the client and the client as much as possible while maintaining the privacy of the client connected to the host and the anonymity when exchanging points between the clients. Also, in metacomputing, since the risk of cracking increases because it connects to the Internet for a long time, it is necessary to take measures to prevent it.

When the client cannot accumulate points and exchange the points only for services obtained from the host, the connection incentive of the large client will be impaired if the points cannot be used effectively. It is necessary to allow the host to purchase excess points, and to exchange points for data and goods between users.

[0015]

In order to solve the above-mentioned problems, a parallel processing calculation entrusting system via the Internet of the present invention is a parallel processing in a host computer equipped with at least a Web server of an information providing service provider via the Internet. In response to various service requests from each client computer of a plurality of users who use the information providing service, the calculation task is entrusted from a customer computer including a company or an individual who entrusts the calculation task via the Internet. It is a meta-computation commissioning system that performs parallel processing calculation by performing distributed processing of the above-mentioned calculation tasks by the plurality of client computers in background processing during the service processing when transmitting corresponding information provision to each via the Internet. hand The amount of time the client computer is communicating receives the information service in response to being indefinite, upon solving the computing problems, a short program operation time based on the initial value generated randomly,
The programs are sequentially transmitted to a plurality of connected client computers at random so that the calculation processing of the Monte Carlo optimal solution search, which individually executes each calculation and executes a huge number of times as a whole, is performed randomly. Based on the generated initial value, it is executed independently and concurrently, and the client side sends back a solution with a good evaluation function value to the host computer, and points according to the contribution amount on the client side are stored in the database on the host side. It is characterized by recording in.

Further, the host computer responds to the calculation task entrustment requests from the plurality of customer computers, and displays a calculation task application screen including at least an item column of a calculation program / calculation condition / calculation resource to each customer computer. Calculation task application screen transmitting / input task receiving / storing means for returning and receiving those data input to each customer and accumulating in the calculation task database, and various in response to access signals from the plurality of client computers Service input screen transmission / requested service storage means for transmitting a service input screen to which a service request can be input to a client computer, receiving the data input by each user, and storing the data in a requested service reception database; Received the data entered on the input screen In this case, a meta-calculation applet generation / transmission means for generating a meta-calculation applet for solving the calculation task on each of the client computers, transmitting the meta-calculation applet to the client computer, and starting the meta-calculation applet on the client computer, When the host computer receives a signal from the host computer indicating that the meta-calculation applet has started on the client computer, the information service corresponding to the requested service starts to be sent to the client and background processing is performed. The applet communicates with a host computer to obtain a seed value that serves as a seed for a pseudo-random number that randomly generates the initial value, and obtains information data regarding a problem to be executed, and determines the initial value to search for a Monte Carlo optimal solution. Line Information service transmission / optimum solution search means for making a search for the optimum solution, and when a solution of a better evaluation function is found in the search for the optimum solution, information about the solution is received from the client computer, recorded in a database, and the optimum solution search and The information service process to the client terminates their communication by disconnecting the client computer, and if the communication is not disconnected, the host computer sends the next seed value to the client computer to set a new initial value, It is characterized in that it comprises evaluation function solution reception / repetition calculation reception means for repeating the following Monte Carlo type optimum solution search.

The information service transmitting / optimum solution searching means of the host computer, when the meta-calculation applet starts to be activated on the client computer,
Before acquiring the information data and the seed value regarding the problem to be executed from the host computer, the client computing resource measuring means for executing the benchmark test program for measuring the size of the computing resource of the client computer for the applet is further added. In order to match the size of the calculation request of the customer and the size of the calculation resource provided by the client, whether the iterative calculation of the Monte Carlo optimal solution search is possible a predetermined number of times is compared with the result of the benchmark test. It is characterized by increasing the processing efficiency of the evaluated computing system as a whole.

Further, when the calculation problem of the optimum solution search by the Monte Carlo method is a neural net type problem, the initial values are synapse weights and threshold values.

Further, when the calculation problem of the optimum solution search by the Monte Carlo method is the calculation by the genetic algorithm, the meta-calculation applet performs the work of instructing the user of the client computer to select the optimum individual as appropriate. It is characterized in that a bad solution far from the optimal solution is prevented from surviving as an optimal solution, and that the user who has contributed to the calculation transmits the amount of promotion calculation as a point to the host computer.

Further, points and services / goods / monetary can be safely maintained while maintaining anonymity between the client computer of the user who has accumulated the points, or between the user and the host computer of the information service provider, and the customer computer entrusted with the parallel calculation task. It is characterized by comprising point exchange service means for exchanging with.

Further, the client computer is
CPUs and related devices built into personal computers, TV game devices, mobile phones, home appliances, private cars, etc. that are commonly used in offices and homes.
Computers including Web servers, FTP servers, file servers, printer servers that operate continuously for 24 hours, vending machines, facsimiles, meteorological observation devices, data loggers, etc. The present invention is characterized by including a CPU built in a device and its related device.

[0022]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

FIG. 2 is a diagram showing the configuration of a parallel processing calculation entrustment system 1 via the Internet of the present invention in FIG.

As shown in the figure, this calculation entrustment system 1
Is a customer computer 30 of a customer including a company / individual who entrusts a parallel processing calculation task, a client computer 40 of a user who uses the information providing service, and an information providing service connected to the respective computers 30, 40 via the Internet 20. The host computer 10 of the business operator. Here, 30a, 40
Reference numeral a denotes a Web browser, and reference numeral 40b denotes a meta calculation applet program described later.

The client computer 40 is a CPU built into a personal computer, a TV game device, a mobile phone, a home electric appliance, a private car, etc., which are generally used in offices and homes, and related devices, and 2
Web server, FTP server, which operates continuously for 4 hours
Computers including file servers and printer servers, vending machines, facsimiles, meteorological observation devices, data loggers, etc. are not the original purpose, but the built-in CPU and related devices of equipment operating 24 hours a day Either can be used.

The host computer 10 is constructed as follows. That is, 11 is a Web server including a firewall, 12 is an application server having processing means such as information provision service business and parallel processing calculation contract business, 13 is a database server that manages a database 14, and 15 is a usage history database of customers and users. 14z is a business processing terminal that performs at least charging processing.

The outline of the parallel processing calculation entrustment system 1 configured as described above is shown below. That is, in the host computer 10 provided with at least the Web server of the information providing service provider via the Internet 20, the customer's customer computer 30 including the company / individual who entrusts the parallel processing computation task with In the background of processing the service when the corresponding information provision is transmitted via the Internet 20 in response to various service requests from the respective client computers 40 of the plurality of users who use the information provision service by consignment. A meta-computation entrusting system for performing parallel processing calculation by performing distributed processing of the above-mentioned calculation task by the plurality of client computers 40 in processing, and more specifically, the client computer 40 receives and communicates information services. Corresponding to the fact that the time is indefinite, when solving the calculation task, each calculation is independently executed based on the initial value that randomly generated a program with a relatively short calculation time The plurality of programs are sequentially transmitted to the plurality of connected client computers 40 so that the calculation processing of the Monte Carlo optimal solution search for executing the number of times can be performed, and independently, based on the randomly generated initial values, Also, a solution having a good evaluation function value including a function for outputting an error to the client side is returned to the host computer 10, and points according to the contribution amount on the client side are sent to the database 14 on the host side.
It is a recording system.

Here, the application server 12 of the host computer 10 has means 12a, 12 shown below.
b, 12c, 12d, 12e, 12f. Customers and users of the customer computer 30 and the client computer 40 who access the host computer 10 authenticate by verifying with the customer / user member database 14x by using a password, and the processed data result is stored in the database 14y. The usage history database 14z collects and collects points, respectively.

The means 12a responds to the calculation task entrustment requests from a plurality of customer computers 30 and transmits a calculation task application screen having at least item fields such as a calculation program, calculation conditions, and calculation resources to each customer computer. Then, the calculation task application screen transmitting / input task receiving and accumulating means 12a receives the data inputted by each customer and accumulates them in the calculation task database 14a.

The means 12b also sends a service input screen to the client computer 40, in which various service requests can be input in response to access signals from the plurality of client computers 40, and receives the data input by each user. The service input screen transmission / requested service storage means 12b is stored in the requested service reception database 14b.

Further, when the means 12c receives the data input on the service input screen, the means 12c generates a meta-calculation applet for solving the calculation task on each of the client computers 40 and sends it to the client computer 40. The computer 40
It is a meta-calculation applet generation / transmission means 12c to be activated above.

When the host computer 10 receives a signal indicating that the meta-calculation applet 40b transmitted from the host computer 10 has been activated on the client computer 40, the means 12d sends an information service corresponding to the request service to the client. Of the applet 40b in the background processing when the transmission of
Communicates with the host computer 10 to obtain a seed value that serves as a seed for a pseudo random number that randomly generates the initial value and obtains information data, and performs the Monte Carlo optimal solution search by setting the initial value. This is an information service transmitting / optimum solution searching means 12d.

The means 12e, when searching for an optimal solution,
When a solution of a better evaluation function is found, information about the solution is received from the client computer 40, recorded in the database 14y of the host computer, and the optimum solution search and information service processing to the client are disconnected by the client computer 40. Will terminate their communication, but if the connection is not yet terminated after receiving the solution, the host computer 10 sends the next seed value to the client computer 40,
It is an evaluation function solution reception / repetitive calculation reception means 12e for determining a new initial value and repeating the next Monte Carlo type optimum solution search.

A flow chart of the processing of the parallel processing calculation entrustment system 1 of the means 12c, 12d, and 12e described above is shown in FIG.

Next, the information service transmitting / optimum solution searching means 12d provided in the application server 12 of the host computer 10 further comprises the client computing resource measuring means 12f shown above.

That is, when the meta-calculation applet 40a transmitted from the host computer 10 starts to be activated on the client computer 40, the applet 40a is acquired before the information data regarding the problem to be executed and the seed value are acquired from the host computer 10. And a means 12f for executing a benchmark test program for measuring the size of the computational resources of the client computer 40.

In order to match the size of the calculation request of the customer with the size of the calculation resource provided by the client by this means 12f, whether or not the Monte Carlo type optimum solution search can be repeatedly calculated a predetermined number of times is determined by the benchmark test. The result can be compared and evaluated, and the processing efficiency of the entire calculation system can be improved.

FIG. 3 shows a flowchart of the processing in the client computer 40 of the above-mentioned means 12f.

Next, the parallel processing calculation entrustment system 1 via the Internet of the present invention will be described more specifically.

Thereafter, the host computer is the host,
The client computer is abbreviated as a client, and the customer computer is abbreviated as a customer.

1. A specific embodiment of a method of paying the amount of computational resources on the client side as a consideration for information services. Most clients in the middle of receiving information services such as information retrieval and image distribution (when a specific host exists) are in an aligned state, but there is often a large free space in the hard disk and memory. Therefore, the host side sends the applet to the client immediately before the start of the information service, and while the information service is provided, the applet is operated on the client to perform calculation, information storage,
The input of the recognition / judgment of the user on the client side (hereinafter collectively referred to as a computational resource) is accepted. The host side uses the computational resources of the client by appropriately transmitting the obtained calculation or recognition / judgment results to the host.

At this time, the amount of computing resources provided to the host varies greatly depending on the difference in the hardware configuration (CPU and memory amount) of the client computer and the difference in user correspondence. That is, the contribution amount of each client seen from the host side is very different. Therefore, points (points) corresponding to the amount of contribution of the client are recorded on the host side so that the points can be exchanged for information services, computational resources accumulated by this method, goods, cash, and the like.

Since the time when the client receives the information service is indefinite, the problem to be calculated is limited to the Monte Carlo type optimum solution search, and the individual calculation results gradually contribute to the whole calculation. In the calculation of the Monte Carlo type optimal solution search, it is necessary to execute a program having a relatively short calculation time a large number of times based on an initial value generated at random, but each calculation can be executed independently. Therefore, the program is sent to a large number of clients in parallel and executed independently based on randomly generated initial values. The calculation result will be returned only when the client obtains a good result.

Here, specific examples of services provided by the host include search services for websites, literature information, town information, etc., download services for still images, movies, music, free software, etc., chess, Othello, etc. Competitive games can be considered. In a download service for still images, moving images, music, free software, etc., if patrol software that automatically downloads only files is used, files are provided to the client without the applet being executed on the client. The applet sent from the host to the client is written in Java (R), and the client is Java.
Even if the browser or the like is set not to execute a (R), the applet sent from the host is not executed. In such a case, even if the information service is provided from the host, the amount of calculation made on the client side cannot be received. Therefore, encrypted communication is performed between the applet sent from the host and the host, and after confirming the execution of the applet, the information service of the host is provided.

2. Example of resale of computing resources collected from clients to customers (how to use points) Whether or not there is a host, the client and host receive points according to the provision of services and computing resources. It will accumulate. By collecting these points, the following merits are combined and given.

(1) Have another client execute its own job. At this time, points are paid according to the amount of jobs executed. (2) Receive information services from the host or other clients. At this time, pay points according to the obtained information service. (3) Exchange points for valuables. A third party (i.e., a customer) who wants to perform a large number of jobs that does not have a sufficient amount of points will exchange points for valuables from the host or a client that has enough points.

3. Specific Example of Performing Authentication and Point Movement While Maintaining Anonymity At the time of point exchange, authentication is performed while maintaining anonymity by the following method. (1) All persons involved in transactions have the following ID numbers that have been converted into irregular numbers by n different types of encryption (see the figure below). The host has personal information such as the names and addresses of all parties involved in the transaction, but will not disclose this information unless requested by the courts or other due legal procedure. As a result, anonymity between traders is completely guaranteed as long as the trader conducts a legitimate trade. ID of user 1 ID of user 2 ID of user i ...

(2) Assume that point X is sent from trader A to trader B in return for sending goods or data from trader B to trader A. When the transaction agreement is reached between the two parties, A and B mutually own their own ID numbers (n types are held)
Select one from the following and send it to the other party and the host respectively (for example, underlined below). The ID number is used only once in one transaction. By making the ID number disposable, it is not recognized that they are the same person no matter how many times they deal with the same person. User A's ID User B's ID

(3) When the host confirms that A holds the point X, the host notifies both of them and A
The point X is reduced and the host temporarily keeps it. (4) B immediately sends the article or data to the host. (5) Upon confirming receipt of the article or data from B, the host notifies A and B of the fact and simultaneously increases the number of points of B by X. (6) If B finds that it did not perform this transaction as contracted, A may report B's ID number to the host and the police or the court and take legal action.

In this method, the anonymity of both parties is guaranteed as long as A and B do not commit fraud. The number n of ID numbers that each person has can be increased as much as possible. For example, if 2100 persons each have 2100 types of ID numbers, the total number is 2.
Although 200 kinds of ID numbers are required, this can be represented by 200-bit data (Note 1).

In this method, goods or data are sent via the host, but when it is not necessary to go through the host, transactions are carried out by the following simplified method.

(1) Persons involved in transactions are n kinds of I
I have a D number. How many IDs are there for each person?
You may record the number on a card and carry it around. in this case,
By giving the card to the trading partner each time a transaction is made, it is possible to save the trouble of recording the ID number of the trading partner and to avoid using the same ID number by mistake.

(2) When the trader A can receive the goods or data from the trader B on the spot, or when it is certain that the goods or the data can be received, the point X is paid on the spot.
Want to be sent from trader A to trader B. When a transaction agreement is established between the two parties, A and B mutually select one of their own ID numbers and send their ID numbers and the number of points they wish to move to the host.

(3) When the host confirms that A holds the point X, the host moves the point X from A to B and informs both of them to that effect. (4) Upon confirming the movement of points from the host, A immediately receives the article or data from B. (5) At a later date, if it is found that this transaction is fraudulent, the victim can report the ID number of the other party to the host and the police or the court and take legal action.

(Note 1) Specific Implementation Method for Generating 2n Types of ID Numbers for 2m Persons Each person's ID number is represented by the following m + n + k bit numbers.

The bit string of is a true ID number of each person who exists only in this world, and is a number that only the person and the host know. Is a bit string for giving each person 2n kinds of ID numbers. Is a bit string for the host to strengthen the encryption strength and embed information about points.

The host rearranges the m + n + k bits and performs operations such as bit inversion and rotation to perform encryption. The host issues the ID number for this operation in response to the user's request. Since the user does not know the encryption method and the generation rule of this part, it is almost impossible to arbitrarily create another person's ID.

(Reasonable Evaluation of Points by Benchmark Test) The size of the computing resources of the client does not always match the size of the computing resources required by the customer (customer) having the problem to be calculated by the metacomputing. Therefore, the problems of a plurality of customers are accumulated on the host side that is providing the information service, and the costs corresponding to the calculation amount are collected from the customers. In order to handle the customer's problem most efficiently, the host measures the client's computing resources with a benchmark test, and executes the problem suitable for each client's computing resource. As a result, the size of the calculation request of the customer is matched with the size of the calculation resource provided by the client, and the efficiency of the entire processing is improved.

At this time, a large difference occurs in the amount of calculation executed in the same time between a client having a high speed computer and a client having a low speed computer. Therefore, points (points) according to the amount of calculation performed on the client
Is provided to the client, and the host adds and records points for each client. The client is given the right to use the computing resources provided by the host as a customer according to the points earned.

By this point system, the calculation resources, the information service, and the cash (the currency used in the real world) can be exchanged with each other, so that it becomes possible to smoothly perform the distribution in the transaction on the Internet.

4. Specific example of using tasks such as user recognition / judgment in metacomputing If the service desired to be obtained from the host is unattended, such as downloading a huge file, the user is receiving the service Be bored during the period. This user is presented with problems such as relatively simple pattern recognition that anyone can do and problems in the specialized field that the user is good at, and points corresponding to the amount of work are added to the amount of computational resources provided by the client computer. The host provides.
This work that can be performed by the user has the following specific implementation methods.

(1) Method used for Monte Carlo type calculation such as genetic algorithm In the case of calculation by genetic algorithm, even if there are very few evaluation errors in the evaluation function that evaluates the best individual, the optimum solution does not exist. Bad solutions that are far apart may survive as optimal solutions. In order to avoid this problem, it is possible to effectively search for a good solution by merely instructing a person to select an optimum individual as appropriate. However, while this work is very monotonous, it is a work that requires mental concentration.

Therefore, when there is human contribution to the calculation performed by the applet running on the client side, and the calculation is promoted thereby, the accelerated calculation amount is provided to the client as a point. By this point, we will secure human contributions and utilize the time that humans have to cut into valuable work.

(Example of Specific Realization Method) For example, an artificial neural network (hereinafter, NN) for recognizing a human face pattern by a genetic algorithm is generated. Then, parameters such as synapse weights inside the NN that correctly perform pattern recognition are optimized by a genetic algorithm. In the first step of the genetic algorithm, m by processing such as parameter variation (randomization).
Assuming that a new NN of a seed child is generated, each recognition result is tested and the one with the highest evaluation value is automatically selected. Then, this is mutated a little to generate more children. Obtaining the optimum NN by this repetition is optimization by the genetic algorithm. However, a program is needed to evaluate which NN recognizes a human face most correctly, but this software must be prepared separately. This is the "human face pattern" that we want to generate here. It can be more difficult than the artificial neural network that does the recognition.

Therefore, the above-mentioned work is performed by the user making a judgment instead of the evaluation program. For example, suppose that the m types of NNs each display the result of face recognition on the screen. A comparison will be made between a program that determines the recognition result of the NN although it is incomplete (hereinafter abbreviated as a determination program) and the case of performing human determination. If the judgment program and the user present different judgment results,
It is assumed that the high reliability is a human judgment result. That is, as described below, when the user determines that the second NN has the best face recognition, while the determination program determines that the third NN has the best face recognition,
The evaluation of the judgment program is incorrect. Evaluation of m kinds of NNs by computer User evaluation f (1) = 0.7 f (1) f (2) = 0.5 f (2) f (3) = 0.71 f (3) f (4 ) = 0.69 f (4) :: f (m) = 0.4 f (m)

If the determination program continues to determine that f (3) is optimal, the second NN is superior to the human determination, but the descendants of the third NN continue to proliferate. Optimization progresses in a direction unfavorable to software developers. If the user finds an erroneous determination in the determination program, he or she points out the error and inputs the correct determination result, and the evolution is corrected in the correct direction. As described above, even when a highly accurate judgment program as high as human pattern recognition ability is required, the user can directly perform evaluation work such as judgment, so that it can be developed by conventional programming like "face recognition software". Difficult software development can be done efficiently.

The determination work by the user is easy to be incorporated in the Monte Carlo type calculation such as the genetic algorithm as in this example, and is easy to use for accelerating the large-scale optimization calculation by the metacomputing according to the present invention.

In addition to the above, there are the following problems to which the user's recognition work can contribute. -Problem that finds the shortest distance, such as the salesman traveling problem. The user presents a solution candidate that seems to be the shortest distance.・ Problem to find the minimum energy state in the tertiary structure of protein The user selects and presents the structure that seems to be the most stable ・ Problem to build the software that predicts the stock price most accurately The user is the prediction result and the actual stock price trend Select the best predicted software from the following and present it ・ Optimization problem of product design with many choices such as cars Users select the product design that best matches their sensitivity from products with different colors, patterns, size of each part, etc. Presented

(2) Method of directly performing pattern recognition In pattern recognition by a pattern recognition device such as character recognition, there is always an erroneous recognition. The correct answer information of the pattern that the pattern recognition device cannot recognize is very important information for improving the pattern recognition device. Since the task of inputting the correct answer of misjudged information is often a relatively simple problem for human beings, this correct answer question is presented to the user, and the user receives points according to the amount of work. .

However, the user may make an erroneous decision. When it is desired to reduce the probability of this erroneous determination, the same work is requested to a plurality of users. That is, the host presents the same problem to a plurality of users, and seeks a true correct answer by majority voting. Bonus points will be increased and provided to the user who presented the correct answer. Alternatively, the number of points provided as a penalty from the user who presented the incorrect answer is reduced. By these, the incentive for the user's work can be improved.

There are many other pattern recognition operations such as the following.・ The presence or absence of abnormalities is determined from information such as power on / off, heart rate, respiration rate, acceleration of movements and movements of home appliances emitted by elderly people and sick people who live alone. Number recognition ・ Reading of temperature and humidity meters and other instruments installed in specific places such as show windows, factories, greenhouses, and livestock houses

(3) Method of utilizing specialized knowledge Although specialized knowledge is required for species identification problems such as species determination of plants and animals, it is relatively simple for those who have specialized knowledge and experience. It is work. A user who can perform such work receives points according to the amount of this work.

(4) Method of answering questionnaire Impressions, problems and advantages of product design and functions, etc.
Answering a questionnaire to point out points for improvement is a relatively simple task in which an unspecified number of users can participate. The user receives points according to the amount of answers to the questionnaire.

(5) Method of inputting original contents When the user inputs information about contents with specific needs such as recipes for dishes, locations of delicious restaurants, and special sale information, the user receives points according to the amount of information. At this time, it is necessary to rationally determine how many points the user who provided the information receives. There are the following methods for implementing the method. It is assumed that a user who wants to view the content actually provides the host with computing resources by the metacomputing mechanism according to the present invention while actually accessing the host and viewing the content. At that time, the user will provide computing resources in return for viewing the content. A part of this calculation resource is converted into points, and the points are received by the content information providing user.

(6) Work that can be done only with fingertips by utilizing the finely divided time A user who connects by using the i-mode or mobile terminal,
The above-mentioned work can be performed by using the finely cut time such as the waiting time of the train. However, as a simple job that can be performed even when it is difficult to see the display while walking, etc.-Notification of occurrence of incidents such as accidents and injuries-Specific place, specific time, specific season required for store planning and marketing Information about the number of people you see and what you are wearing

(Method of securing a large number of clients necessary for implementing the present invention) In the present invention, a large number of clients need to be connected. Further, in the case of realizing with Java (R) language and a normal browser, the browser executed by the client needs to be set so that a Java (R) applet (hereinafter, applet) can operate.
Although it is possible to prohibit the execution of Java (R) applets and Java (R) scripts in the security settings of many browsers, many users have prohibited such execution because of security measures. Therefore, the following method is adopted.

(1) Confirmation of applet activation After the applet sent from the host to the client is activated, it is authenticated by encrypted communication with the host to confirm that the applet has been activated, and the service to the client is confirmed. Start offering. Therefore, if your browser settings prevent the applet from running,
The user cannot receive the service.

(2) Ensuring Applet Execution Environment by Combining with Applet Type Service The reason why the user does not execute applet is that it takes some time before the applet is activated in addition to the security problem. Therefore, configure the host so that using the applet can be used as a service. Services that are easier to implement or provide with applets include games, chats, electronic bulletin boards, and
Since there are so many electronic conferences, music / image distribution, videophone, news distribution, office suites (set of spreadsheet software and word processing software), mailers, etc., the present invention can be implemented by combining these applet type services. The environment can be adjusted.

(3) Standardization of Points with Host-Provided Service In the service exemplified in (2), in a service such as a game, points are added according to the score obtained in the game. That is, the user should be able to obtain points according to the score of the game (win or loss in a competitive game such as chess or Othello) as a bonus, in addition to the computational resources provided to the host during execution of the game. Thus, more users can be secured for a longer time.

(4) Bonus points when a good solution is obtained In the case of an optimization problem in which the calculation made by the client optimizes a certain target variable, a solution better than the best solution stored in the host is obtained. When given, bonus points are provided to the user of the client to improve the user's willingness to connect.

(5) Method of paying points only when a good solution is obtained Instead of paying points according to the computational resources provided by the client, a better solution than the optimal solution recorded in the host in the dictionary is found. A method is also used in which bonus points are paid only when Since the probability that an optimal solution will be obtained increases as the total calculation amount of the client increases, the expected value for obtaining points becomes closer to the case where the calculation resources are carefully measured. This method can simplify the operation of the entire metacomputing. Also, the client only needs to connect to the host or send the result by a method such as e-mail to get the points only when a good solution is actually obtained, so the client disconnects the connection line to the network and stays offline for a long time. You can participate in metacomputing even if you are in a state.

(6) Point exchange between users By allowing points to be exchanged between users, the user's motivation for collecting points is improved. The points will be exchanged by the following methods to prevent fraud such as fabrication of points. Note that this method may be applicable to similar point exchanges other than the present invention. For each user, register the ID and password in the host, authenticate both parties who want to exchange points, and then agree on the number of points to be exchanged and the transfer method. To move.

The following moving methods are available. 1) Simply move the number of points recognized by both mobile users in the direction agreed by both users. In this case, the host may set such that the host collects some points as a commission.

2) Move points at the same time as data exchange between users Let users A and B want to exchange points. At this time, A wants to make this transaction under the condition that points are moved to B after confirming that certain data (document, personal identification number, music / video file, program, etc.) has been surely received from B. . In order to prevent the malicious transaction B from being completed without sending the data, the host provides the following services, and the user uses them according to the purpose.

A. A simple sending host receives the data from B in advance, and sends the data to A at the same time when the points are moved.

B. The host encrypts the data received from the encrypted sending B and A
Sent to. At the same time as moving the point, send a decryption key to A to decrypt the data.

C. In both the sending methods a and b in the case where both parties do not have trust, the transaction is completed even when B sends false data. Therefore, these are
This is effective when is a trusted corporation. As a trading method when B is not credible, when the data sent by B is not satisfactory, the host takes all or part of the points B gained from A as a penalty. The percentage of this penalty is decided by agreement between A and B. Also, A is B
In some cases, the data obtained from the company may be satisfactory but may be unsatisfactory. As a method of preventing this, the host deposits a certain number of points, which is determined in advance according to the number of transaction points, with the point A as a deposit. If B complains of A's unfair complaint, the host will automatically take this deposit from A as a penalty.
That is, when the worst trouble occurs in the transaction, both A and B receive a penalty according to the number of transaction points. This psychologically prevents the trouble, and the user who frequently causes the trouble becomes less likely to participate in the transaction because the number of points held is reduced by the penalty.

5. Specific Examples of Monte Carlo Calculations such as Neural Networks and Genetic Algorithms (1) Dynamic Evaluation of Problems to be Calculated Neural nets and genetic algorithms are one of the calculation methods for optimal solution search by Monte Carlo method, but they are solved. It has the versatility that the types of problems that can be addressed are very wide. However, at the same time, in the Monte Carlo type calculation, it is necessary to perform a simple calculation by executing a large number of times, and it is not possible to obtain a useful calculation result without repeating the calculation to some extent. Therefore, the size of the problem to be solved is determined according to the memory size of the client, and whether or not the calculation can be repeated a specified number of times is evaluated by a benchmark test, and the calculation resources of the client are used to the maximum extent.

(2) Reduction of communication traffic in neural net type problem In the calculation by the neural net, the calculation results to be transmitted to the host as the deliverables are the synapse weight and the threshold value, but this information amount is very large. Even if a certain client gets a good calculation result, if the client disconnects (logs off or moves to another host) while sending the result, the calculation result will be returned to nothing.
Even if the information is compressed using the conventional information compression technology, the amount of information is still large, and the risk of disconnection during data return is large.

In a problem such as a neural network, the problems to be solved are the data set of learning samples and the initial values of synapse weights and thresholds. The learning sample is sent from the server side, but the initial values of the synapse weight and the threshold value are pseudo-random numbers in which a specific numerical value is used as a seed (a numerical value used as a seed for random number generation), and can be generated on the client side. Therefore, if only the seeds for pseudo-random number generation are sent from the host to the client and the client finds a better solution (a solution with a better evaluation function value such as error) than the existing solution, a good solution is obtained. It informs the host that it has been found, the host sends a new seed to the client, and the client proceeds to the next calculation without wasting time.

When the customer requests all information (the synaptic weights and thresholds of the neural net obtained by learning) of the best solution from the host, the host uses the seed on the host to perform the same calculation as the client. Is reproduced,
Present all requested information. In some cases, only the seed is provided to the customer to simplify communication, or the customer is provided with a service with added value such as recognition processing using a neural network obtained for unknown data. .

6. Java (R) instead of banner advertisement
A concrete example of sending an applet to a client and using a client resource as a calculation resource. In a web server that provides content such as a search service and image distribution for free, the main revenue source of a web server that provides free content service is a banner. Although it is an advertisement, the Java (R) applet according to the methods of the preceding paragraphs 4 and 5 can be stored in a file size which is not much different from the banner advertisement.
As a result, the client user can receive the service on a neat screen without the banner advertisement without increasing the waiting time before starting the service.

Java sent to the client side
(R) At the same time as executing the calculation, the applet sends to the host that it is in the execution start state. The host starts servicing the client after knowing that the client has executed the Java (R) applet. If the browser of the client does not permit execution of the Java (R) applet, the client user cannot obtain the service, so that the service and the calculation amount can be reliably exchanged.

7. Example Customer-Client-Client Flow The host provides information services to multiple clients. While connected to the host, the client provides the host with part of its computational resources as a price for the information service. When the client provides computing resources more than the price of the information service, points corresponding to the amount of contribution are accumulated in the host. This point is the right to use other computing resources and the right to use information services, and can be exchanged for cash (money) at an appropriate rate.

The customer pays the host for cash,
Request Monte Carlo calculation. The host uses a portion of the revenue collected from the customer to pay the client.
FIG. 4A shows the flow of computational resources. The computing resource of the client is indirectly provided to the customer by the host accumulating the computing resource of the client and transmitting the calculation result to the customer.
When the problem executed by the client is the neural net type, the calculation result sent from the client to the host is only the seed of the random number that generated the initial condition and the value of the evaluation function obtained at that time.

FIG. 4B shows the flow of information service and cache. The client connects to the host as an incentive for payment of information services or cash provided by the host. The host receives payment for cash or points, or an information service equivalent thereto by performing calculation on behalf of the customer.

[0097]

The parallel processing calculation entrusting system via the Internet of the present invention has the following effects. That is, almost all computers connected to the Internet are idle, but this computing resource can be effectively used.

By limiting the calculation executed on the client to the Monte Carlo type calculation problem with high versatility, it is only necessary to send an applet of a small program size to the client, and the client can receive the information service at irregular intervals. It can be used effectively.

Examples of Monte Carlo type calculation include optimization problems by genetic algorithms and pattern recognition by neural nets, which can be used for stock price prediction / exchange rate prediction for predicting the future from fluctuation patterns of time series data. . For example, when a stock price forecasting consulting business is desired, an enormous amount of calculation is required to create a forecasting model for highly accurate stock price forecasting in a short time with a genetic algorithm or a neural network. When securing this calculation amount using an existing supercomputer, the depreciation cost of the supercomputer corresponding to the calculation time required for one prediction is the cost required for the prediction. According to the present invention, a customer who wants to carry out the stock price forecasting consulting business can be secured as a customer by forecasting at a cost lower than this. In addition, the host operator can conduct the stock price forecasting consulting business by himself.

Businesses to which the present invention can be applied include, in addition to stock price prediction, analysis of DNA and amino acid sequences, protein structure calculation, dynamic analysis and prediction of biological individuals and biological populations, and ecosystem, international management flow. The dynamic analysis and prediction of

Conventionally, in order to obtain a compensation for the provision of the information service, the user has only been charged directly or indirectly by a banner advertisement. The present invention technically enables, as a third method, to collect the computing resources of the computer used by the user from the user.

The point system makes it possible to exchange small calculation resources, small information services, small labor, and cash (money) with each other, and to smoothly distribute them. In conventional electronic money, the only backing as money is backing by credit, and
If a company that provides a settlement service using its own electronic money is upset, it will be difficult for them to exchange each other as money, and it has not spread. In the method according to the present invention, the amount of calculation performed by the client is expressed by points, which can be used as a kind of electronic money. In addition, since the customer can always use the computing resources corresponding to the points he / she owns, it is possible to secure the proof that the points can perform a specific calculation at any time depending on the size of the points. Further, even when a plurality of vendors perform the host service according to the present invention with different information services and points with different names, the points of each host can be objectively measured according to the amount of calculation, and therefore the hosts exchange each other. It is easy to do. Therefore, this point has a function as convertible currency, and can be widely used as convertible currency in electronic payment at the time of purchasing an article on the Internet.

[0103] The user can contribute to the depreciation of purchased personal computers by gaining points. There is also an advantage that points can be earned and boredom can be distracted by doing the work directly at the time of fine cut. In addition, it can provide an opportunity for people with disabilities to work without moving at home or on the bed at all.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a parallel processing calculation entrustment system via the Internet of the present invention.

FIG. 2 is a diagram showing an operational flow of the entire parallel processing calculation entrustment system of the present invention.

FIG. 3 is a diagram showing a flow of processing of an applet in a client computer of the entrustment system of the present invention.

FIG. 4 is an explanatory diagram of flows of a computational resource / information service and a cache according to the present invention.

[Explanation of symbols]

1 Parallel processing calculation consignment system via the Internet 10 Host computer (host) of information service provider 11 Web server including firewall 12 Application server 12a Calculation task application screen transmission / input task reception / accumulation means 12b Service input screen transmission / request service Storage means 12c Meta calculation applet generation / transmission means 12d Information service transmission / optimum solution search means 12e Evaluation function solution reception / repetitive calculation reception means 12f Client calculation resource measurement means 12x Information provision service business software 13 Database server 14 Database 14a Calculation task database 14b Request service reception database 14c Information provision service database 14x Customer / user member database 14y Customer / user processing result Database 14z Customer / user usage history database 15 Business processing terminal 20 Internet line 30 Customer customer computer (customer) 30a Web browser 40 entrusting calculation tasks Client computer (client) 40a Web browser 40b host of user who uses information providing service Meta calculation applet program sent from the computer

Claims (7)

[Claims] 1. A company that entrusts a parallel processing calculation task in a host computer equipped with at least a Web server of an information service provider via the Internet,
The customer's computer including an individual accepts the above-mentioned calculation task via the Internet, and responds to various service requests from each client computer of a plurality of users who use the information providing service via the Internet. A meta-computation entrustment system for performing parallel processing calculation by performing distributed processing of the calculation task by the plurality of client computers in background processing during the service processing when the information provision is transmitted. Corresponding to the fact that the time during which the service is received and communicated is indefinite, when solving the above-mentioned calculation task, each calculation is executed independently based on the initial value randomly generated by a program with a short calculation time. As a whole, Monte has a huge number of executions. In order to perform the calculation process of the Carlo type optimal solution, the plurality of programs are sequentially transmitted to a plurality of connected client computers, and are independently and simultaneously executed based on randomly generated initial values. , A parallel processing calculation consignment system via the Internet characterized in that a solution having a good evaluation function value on the client side is returned to the host computer, and points according to the contribution amount on the client side are recorded in a database on the host side. 2. The host computer responds to the calculation task entrustment request from the plurality of customer computers, and displays a calculation task application screen including at least a calculation program / calculation condition / calculation resource item column to each customer computer. Calculation task application screen sending / input task receiving / storing means for sending back and receiving those data input to each customer and accumulating them in the calculation task database, and various in response to access signals from the plurality of client computers Send a service input screen that allows you to enter a service request to the client computer, receive the data that each user has entered, and save it in the requested service reception database Send service input screen /
When the request service storage means and the data input on the service input screen are received, they are generated as a meta calculation applet for solving the calculation task on each of these client computers and transmitted to the client computer, and the client computer Meta calculation applet generation / transmission means to be started on the computer, and request to the client when the host computer receives a signal indicating that the meta calculation applet transmitted from the host computer has started on the client computer. While the information service corresponding to the service is started to be transmitted, the background process causes the applet to generate a seed value for a pseudo-random number that randomly generates the initial value, and information about a problem to be executed by the host computer. Information service transmission / optimum solution search means for communicating with a computer and obtaining the same and performing the Monte Carlo type optimum solution search by defining the initial value, and when a solution of a better evaluation function is found during the optimum solution search, Information about the solution is received from the client computer and recorded in a database, and the optimum solution search and information service processing to the client terminate their communication by disconnection of the client computer, and if the communication is not disconnected, the host The computer further comprises an evaluation function solution reception / repetitive calculation reception means for transmitting the next seed value to the client computer, causing it to define a new initial value, and repeating the next Monte Carlo type optimum solution search. The parallel processing calculation entrustment system via the Internet described in 1. 3. The information service transmitting / optimum solution searching means of the host computer acquires information data and a seed value regarding a problem to be executed from the host computer when the meta-calculation applet starts to be activated on the client computer. Before that, the applet further comprises a client computing resource measuring means for executing a benchmark test program for measuring the size of the computing resource of the client computer, the size of the customer's computing request and the computing resource provided by the client. In order to match the sizes, the processing efficiency of the entire calculation system evaluated by comparing whether the iterative calculation of the Monte Carlo type optimum solution search is possible a predetermined number of times with the result of the benchmark test is improved. Item 2 interface -Parallel processing calculation consignment system via net. 4. The Internet according to claim 2 or 3, wherein when the calculation task of the optimum solution search by the Monte Carlo method is a neural net type problem, the initial value is a synapse weight and a threshold value. A parallel processing entrustment system via. 5. When the calculation task of the optimum solution search by the Monte Carlo method is the calculation by a genetic algorithm, the meta-calculation applet performs a task of instructing the user of the client computer to select the optimum individual as appropriate. 3. A bad solution far from the optimum solution is prevented from surviving as an optimum solution, and the user who has contributed to the calculation is sent the amount of promotion calculation as a point to the host computer.
Alternatively, the parallel processing calculation entrustment system via the Internet described in 3. 6. The points and services / goods / monetary can be safely maintained while maintaining anonymity between the user's client computers accumulating the points, or between the user and the host computer of the information service provider, and the customer computer entrusting the parallel calculation task. 6. The parallel processing calculation entrustment system via the Internet according to claim 1, further comprising point exchange service means for exchanging with. 7. The client computer is a personal computer, a TV game device, a mobile phone, a home appliance, or the like, which is commonly used in offices and homes.
CPU and its related devices built in private cars, 24
Computers including web servers, FTP servers, file servers, printer servers that operate continuously for hours, vending machines, facsimiles, meteorological observation devices, data loggers, etc. are not the original purpose, but they are operating 24 hours a day. The parallel processing calculation commissioning system via the Internet according to any one of claims 1 to 6, further comprising: a CPU built in the device and a device associated therewith.