Patent Application
20180197106
The present invention embodiments relate to a determination of a training data set used for machine learning.
Estimation models have been used to predict various parameters that may be difficult to directly measure. The estimation models are learned from a number of training data sets. Excess training data demands many computing resources to generate the estimation models, and in some cases, reduces the accuracy of the predictions made by the estimation models, for example, by overtraining.
According to a first aspect of the present invention, provided is a computer readable medium having instructions embodied therewith, the instructions executable by a processor or programmable circuitry of a computer to cause the processor or programmable circuitry to perform operations including selecting a first reference training data from a training data set based on a degree of difference between each reference training data in the training data set and a target training data, determining whether a degree of difference between the target training data and the first reference training data is below a first threshold, and determining whether to include the target training data in the training data set using at least a degree of difference between the target training data and at least one second reference training data, wherein the at least one second reference training data is included in the training data set and does not include the first reference training data. The first aspect may also include a method performed by a computer comprising the foregoing computer readable medium, and an apparatus having the above computer readable medium.
The summary clause does not necessarily describe all features of the embodiments of the present invention. Embodiments of the present invention may also include sub-combinations of the features described above.
Hereinafter, example embodiments of the present invention will be described. The example embodiments shall not limit the invention according to the claims, and the combinations of the features described in the embodiments are not necessarily essential to the invention.
The storing section 100 may store a variety of data used for operations of the apparatus 10. The storing section 100 may comprise a volatile or non-volatile memory. One or more other elements in the apparatus 10 (e.g., the obtaining section 102, the calculating section 104, the selecting section 106, the determining section 108, the adding section 112, the training section 114, etc.) may communicate data directly or via the storing section 100.
The obtaining section 102 may obtain a training data set including a plurality of reference training data. The obtaining section 102 may also obtain target training data. The plurality of reference training data in the training data set may have been used for training a current model, and the target training data is not necessarily used for training a current model, but may be used for training a new model. The obtaining section 102 may obtain the training data set and/or the target training data from a database 20.
The calculating section 104 may calculate a degree of difference between one training data and another training data. The calculating section 104 may calculate a degree of difference between each reference training data in the training data set and the target training data. The calculating section 104 may calculate a degree of difference between two reference training data in the training data set.
The selecting section 106 may select a first reference training data from the plurality of the reference training data in the training data set, based on the degree of difference between each reference training data in the training data set and target training data. The selecting section 106 may select the first reference training data, which yields the smallest degree of difference from the target training data among all of the reference training data in the training data set.
The selecting section 106 may also select at least one second reference training data (which may be simply referred to as “second reference training data”) from the plurality of the reference training data in the training data set. The second reference training data may be included in the training data set and may not include the first reference training data.
The selecting section 106 may select the second reference training data such that the degree of difference between the target training data and the first reference training data is smaller than the degree of difference between the target training data and the second reference training data. The selecting section 106 may select two second reference training data including a second-closest-to-target training data and a closest-to-first-reference training data.
The second-closest-to-target training data may yield the second smallest degree of difference from the target training data among all of the reference training data in the training data set. The closest-to-1st-reference training data may yield the smallest degree of difference with the first reference training data among all training data in the training data set.
The determining section 108 may determine whether a degree of difference between the target training data and the first reference training data is below a first threshold. When deciding that the degree of difference between the target training data and the first reference training data is below a first threshold, the determining section 108 may determine whether to include the target training data in the training data set to update the training data set.
The determining section 108 may determine whether to include the target training data in the training data set by using at least a degree of difference between the target training data and at least one second reference training data. The determining section 108 may determine to include the target training data in the training data set in response to determining that the degree of difference between the target training data and the second-closest-to-target training data is larger than a degree of difference between the first reference training data and the closest-to-first-reference training data.
The adding section 112 may update the training data set. The adding section 112 may add the target training data to the training data set in response to the determining section 108 determining to include the target training data in the training data set. The adding section 112 may replace the first reference training data with the target training data in the training data set in response to the determining section 108 determining to include the target training data in the training data set.
The training section 114 may train a model (e.g., estimation model) by using at least a portion of the training data set that is updated by the adding section 112 and that includes the target training data.
At S110, an obtaining section, such as the obtaining section 102, may obtain a training data set including a plurality of reference training data from a storing section, such as the storing section 100, or an external device or storage media, such as the database 20.
At S120, the obtaining section may obtain target training data from the storing section 100 or the external device or the storage media. Training data, such as the reference training data and target training data, may include input data and output data.
The input data of each reference training data and the target training data may be represented by a vector. In an embodiment, the input data may include one or more values that are easily measured by one or more sensors (e.g., temperature, pressure, flow rate, etc. measured in a plant).
The output data of each reference training data and the target training data may be represented by value(s) or label(s). In some embodiments, the output data may include one or more values that cannot be directly measured and/or that cannot be measured in real time (e.g., solution concentration), and/or one or more labels that cannot be directly obtained (e.g., normal operation and/or abnormal operation).
Output data corresponds to input data in each of the reference training data and the training data. In one embodiment, output data and input data in one reference training data or target training data may be derived from one plant at a time. The input data may be used for estimation of corresponding output data by the model. Such estimation may be referred to as a soft sensor.
In some embodiments, the target training data may be newer than the plurality of reference training data in the training data set obtained at S110. In these embodiments, the target training data may be obtained from one plant at a time later than a time when the plurality of reference training data in the training data set are obtained from the plant.
At S130, a calculating section, such as the calculating section 104, may calculate a degree of difference between the target training data and each of the plurality of the reference training data. The calculating section may calculate a database monitoring index (or DMI) between two training data as the degree of difference. DMI is defined by the following formula (1):
where (Xi, yi) and (Xj, yj) represent pairs of (input data, output data) of training data i and training data j, “a” in the formula (1) represents a constant, and “sim (Xi, Xj)” represents a similarity between two input data (Xi and Xj).
In some embodiments, “sim (Xi, Xj)” may be calculated based on Euclidean Distance and/or Mahalanobis Distance between the two input data (Xi, Xj). In these embodiments, “sim (Xi, Xj)” may be defined by the following formula (2):
sim(Xi,Xj)=exp(−γ∥Xi−Xj∥2) (2)
where ∥Xi,−Xj∥ represents Euclidean Distance and/or Mahalanobis Distance between the two input data (Xi, Xj), and γ represents a constant.
According to the formulae (1)-(2), the more similar the two input data (Xi, Xj) and/or the two output data (yi, yj) become, the smaller the degree of difference between two training data becomes. The constant “a” in the formula (1) may amount to a weight of the output data against the input data. For example, when a value of “a” becomes small (e.g., smaller than 1), the value of DMI becomes larger in response to a minor variation of output data y (e.g., a variation giving |yi−yi|<1), such as a drift of a model. On the other hand, when a value of “a” becomes greater than 1, the value of DMI becomes smaller in response to a minor variation of output data y while becoming less vulnerable to noise.
In alternative embodiments, the calculating section 104 may calculate the degree of difference between two training data by other methods. For example, the calculating section 104 may calculate a sum or a product of (i) Euclidean Distance and/or Mahalanobis Distance between the two input data, and (ii) Euclidean Distance and/or Mahalanobis Distance between the two output data, as the degree of difference. The calculating section 104 may calculate a similarity between the two training data by at least partially utilizing the above-explained methods or other methods.
At S140, a selecting section, such as the selecting section 106, may select the first reference training data that yields the smallest degree of difference from the target training data among all of the reference training data in the training data set, based on the degrees of difference calculated at S130.
At S145, a determining section, such as the determining section 108, may determine whether a degree of difference between the target training data and the first reference training data is below a first threshold. The determining section 108 may proceed with an operation of S150 if the decision is positive. The determining section 108 may proceed with an operation of S180 if the decision is negative. In an alternative embodiment, the determining section may end the operations of
At S150, the calculating section 104 may calculate the degree of difference between the first reference training data and the other reference training data in the training data set. The calculating section 104 may perform the calculation in a manner explained in relation to the operation of S130.
At S160, the selecting section 106 may select one or more second reference training data. The selecting section 106 may select second-closest-to-target training data and a closest-to-first-reference training data as the second reference training data. The selecting section 106 may select any reference training data from the training data set as the second reference training data except the first reference training data.
In the embodiment of
In the embodiment of
In some embodiments, the second-closest-to-target training data and the closest-to-first-reference training data can be the same. In other embodiments, the second-closest-to-target training data and the closest-to-first-reference training data cannot be the same.
At S170, the determining section 108 may determine whether the degree of difference between the target training data and the second-closest-to-target training data is larger than the degree of difference between the first reference training data and the closest-to-first-reference training data. In the embodiment of
At S180, an adding section, such as the adding section 112, may update the training data set by adding the target training data to the training data set. In an embodiment, the adding section 112 may replace the first reference training data with the target training data. In other embodiments, the adding section 112 may add the target training data to the training data set without deleting the first reference training data.
In general, the more distant training data is from other training data, the more informative the training data is. In a case where S180 is performed, the first reference training data “A” is less informative than the target training data “T” because the target training data “T” is more distant from the other reference training data than the first reference training data “A”. Thus by the operation of S180, the adding section can replace less informative training data in the training data set (such as the first reference training data “A”) with the more informative training data (such as the target training data “T”).
At S190, a training section, such as the training section 114, may train a model using all or a part of training data in the training data set updated at S180. In some embodiments, the training section 114 may update an adaptive model such as online support vector regression by using the training data set updated by the adding section. In other embodiments, the training section 114 may train another type of regression model using the updated training data set. The training section 114 may train neural networks or other models used for machine learning using the updated training data set. In the embodiment of
In response to obtaining target training data, the apparatus may add the target training data to a training data set if the target training data is more informative than the nearest reference training data in the training data set. In other words, the apparatus may update the training data set with a new training data if it is beneficial. Thereby, the apparatus may limit the number of training data in the training data set, while saving computational resources used for training an estimation model, and maintaining the accuracy of the estimation model by avoiding overtraining.
In some embodiments, the apparatus may repeat operations S120-S190. For example, the apparatus may repeatedly obtain new training data as new target training data from a plant by monitoring operational conditions of the plant. In the example, the apparatus may perform operations of S120-S190 for each new target training data.
The apparatus may treat the training data set updated at S180 as the training data set for the next iteration of S120-S190. Thereby, the apparatus may determine whether to update the training data set with the new target training data based on whether the new target training data is informative. Many embodiments may be applied to adaptive models such as Moving Window (MV) model and/or Just-in-Time (JIT) model, which may keep updating the estimation model.
In some embodiments, the apparatus may eliminate reference training data from the training data set, without adding the target training data. In these embodiments, the selecting section may randomly select one of the reference training data from the training data set instead of obtaining the target training data at S120, and then perform at least some of operations S130-S190 by treating the selected reference training data as the target training data. In these embodiments, the apparatus may eliminate the selected reference training data from the training data set if the decision of S145 or S170 is negative, and may eliminate the first reference training data at S180.
Thereby, the apparatus may reduce the number of training data in the training data set while maintaining a diversity of training data in the training data set to reduce computational resources used for training. The apparatus may repeat the elimination of reference training data from the training data set.
In some embodiments, at S145, the determining section 108 may determine whether a number of the reference training data in the training data set exceeds a threshold number additionally or alternatively to comparing the degree of difference to the first threshold. Thereby, the determining section 108 may proceed with operations of adding the target training data only when a size of the training data set is sufficiently small.
In some embodiments, at S170, the determining section 108 may determine whether to add the target training data based on not only (a) the degree of difference between the target training data and the second-closest-to-target training data, and (b) the degree of difference between the first reference training data and the closest-to-first-reference training data, but also (c) one or more of the degrees of difference between the target training data and each other second reference training data (e.g., third-closest-to-target training data, fourth-closest-to-target training data, etc.) in the training data set, and (d) one or more of the degrees of difference between the first reference training data and each other second reference training data (e.g., second closest-to-first-reference training data, third closest-to-first-reference training data, etc.) in the training data set.
In these embodiments, the determining section 108 may determine to include the target training data in the training data set in response to determining that an average degree of difference between the target training data and a plurality of the second reference training data (e.g., an average of (1) the degree of difference between the target training data and the second-closest-to-target training data, (2) the degree of difference between the target training data and the third-closest-to-target training data, (3) the degree of difference between the target training data and the fourth-closest-to-target training data, . . . (n−1) the degree of difference between the target training data and the nth-closest-to-target training data) is larger than an average degree of difference between the first reference training data and a plurality of second reference training data (e.g., an average of (1) the degree of difference between the first reference training data and the closest-to-first-reference training data, (2) the degree of difference between the first reference training data and the second closest-to-first-reference training data, . . . , (n−1) the degree of difference between the target training data and the (n−1)th-closest-to-reference training data), where n is a number equal to or less than the number of the reference training data in the training data set.
The calculating section may calculate the degrees of difference DT-A between T and A, DT-B1 between T and B1/C2, DT-B2 between T and B2, DA-C2 between A and B1/C2, and DA-C1 between A and C1. The determining section 108 may determine to replace the first reference training data A with the target training data T in response to determining that an average of DT-B1 and DT-B2 is larger than an average of DA-C1 and DA-C2.
In a further alternative embodiment, at S170, the determining section 108 may determine whether to add the target training data based on one second reference training data. In these embodiments, the determining section 108 may determine to add the target training data if the degree of difference between the target training data and the one second reference training data is above a threshold.
In other alternative embodiments, the apparatus may not perform some of operations S110-S190. In these embodiments, the apparatus may not perform S150-S170.
In embodiments explained in relation to
The storing section 600, the obtaining section 602, the calculating section 604, the selecting section 606, the determining section 608, the adding section 612, and the training section 614 may have substantially the same function with the storing section 100, the obtaining section 102, the calculating section 104, the selecting section 106, the determining section 108, the adding section 112, and the training section 114, which are explained in the embodiments in relation to
Hereinafter, the storing section 600, the obtaining section 602, the calculating section 604, the selecting section 606, the determining section 608, the adding section 612, and the training section 614 may be explained in terms of differences from the storing section 100, the obtaining section 102, the calculating section 104, the selecting section 106, the determining section 108, the adding section 112, and the training section 114.
The obtaining section 602 may obtain a target training data and a plurality of training data sets. Each of the plurality of training data sets may include a plurality of reference training data. The plurality of training data sets may be regarded as a plurality of groups of training data, where one or more similar training data may be allocated to the same group.
The calculating section 604 may calculate a degree of difference between each reference training data in the plurality of training data sets, and the target training data. The calculating section 604 may further calculate a degree of difference between two training data in a training data set.
The selecting section 606 may select the first reference training data, which yields the smallest degree of difference from the target training data, among all of the reference training data in the plurality of training data sets. The selecting section 606 may select a training data set that includes the first reference training data, which may be referred to as “target training data set.”
The selecting section 606 may also select the at least one second reference training data, which includes a farthest-from-target training data that gives the largest degree of difference with the target training data among all reference training data in the target training data set. Thereby, the selecting section 606 may select the second reference training data from the same group as the first reference training data.
The determining section 608 may determine whether a degree of difference between the target training data and the first reference training data is below a first threshold. In determining whether the degree of difference between the target training data and the first reference training data is below a first threshold, the determining section 608 may further determine whether to include the target training data in the target training data set.
The determining section 608 may determine whether to include the target training data in the training data set by using at least a degree of difference between the target training data and at least one second reference training data. In some embodiments, the determining section 608 may determine to include the target training data in the target training data set in response to determining that the degree of difference between the target training data and the farthest-from-target training data is below a second threshold.
The adding section 612 may add the target training data to the target training data set that corresponds to one group among the plurality of groups of training data, in response to the determining section 608 determining to include the target training data in the target training data set.
The forming section 613 may form a selected data set for performing machine learning. The forming section 613 may select one or more reference training data from each of the plurality of groups of training data to form the selected data set.
The training section 614 may train a model (e.g., estimation model). The training section 614 may perform machine learning using the selected data set including one or more reference training data selected by the forming section 613.
At S210, an obtaining section, such as the obtaining section 602, may obtain a plurality of training data sets from a storing section such as the storing section 600 or an external device or storage media such as the database 70. The plurality of reference training data in each training data set may be close to each other in terms of the degree of differences between reference training data. In some embodiments, a number of reference training data are grouped into the plurality of training data sets, such that each reference training data has at least one reference training data having a degree of difference from each other reference training data below a threshold, in each training data set.
At S220, the obtaining section 602 may obtain a target training data from the storing section or the external device or storage media. The operations of S210 and S220 may be performed in substantially the same manner as the operations S110 and S120, except as otherwise specified.
At S230, a calculating section, such as the calculating section 604, may calculate a degree of difference between the target training data and each reference training data in the plurality of training data sets. In some embodiments, the calculating section 604 may calculate a database monitoring index (or DMI) between two training data as the degree of difference as explained in relation to the operation of S130.
At S240, a selecting section, such as the selecting section 606, may determine the first reference training data among all reference training data of the plurality of the training data sets, based on the degree of difference between the target training data and each reference training data in each plurality of training data sets. In some embodiments, the selecting section 606 may select the reference training data that yields the smallest degree of difference from the target training data, as the first reference training data.
At S250, the selecting section 606 may select a target training data set among the plurality of training data sets. In other words, the selecting section may determine one group from a plurality of groups corresponding to the plurality of training data sets. The selecting section may select, as the target training data set, a training data set that includes the first reference training data selected at S240.
The target training data is presented as a circle shown as “T.” At S240, the selecting section may select training data “A” from among all fourteen training data as the first reference training data, if the training data “A” yields the smallest degree of difference (shown as “DT-A”) from the target training data “T” among the fourteen training data. At S250, the determining section 608 may select the training data set 800 as the target training data set.
At S260, the selecting section 606 may select one or a plurality of second reference training data. The selecting section 606 may select, as a second reference training data, a farthest-from-target training data that gives the largest degree of difference with the target training data among all reference training data in the target training data set selected at S250.
In the embodiment of
At S270, a determining section, such as the determining section 608, may determine whether a degree of difference between the target training data and the first reference training data is below a first threshold. The determining section 608 may proceed with an operation of S290 if the decision is positive. The determining section 608 may proceed with an operation of S280 if the decision is negative.
At S280, an adding section, such as the adding section 612, may generate a new training data set including the target training data, as a new group of training data.
At S290, the determining section 608 may determine whether the degree of difference between the target training data and the farthest-from-target training data selected at S260 is below a second threshold. The second threshold may be larger than the first threshold. The determining section 608 may proceed with an operation of S300 if the decision is positive. The determining section 608 may proceed with an operation of S310 if the decision is negative.
At S300, the adding section 612 may add the target training data to the target training data set selected at S250. Thereby, the adding section 612 may add the target training data to the one group of the plurality of groups.
At S310, the determining section 608 may determine whether to replace the farthest-from-target training data (i.e., the second reference training data) with the target training data in the target training data set. The calculating section 604 may calculate a degree of difference between the farthest-from-target training data and the first reference training data, and the determining section 608 may determine whether the degree of difference between the target training data and the first reference training data is smaller than the degree of difference between the farthest-from-target training data and the first reference training data.
In the embodiment of
At S320, the adding section 612 may include the target training data in the target training data set and exclude the farthest-from-target training data from the target training data set. In the embodiment of
Thereby, the adding section 612 may perform the replacing in response to the determining section 608 determining (i) that the degree of difference between the target training data and the first reference training data is below the first threshold, (ii) that the degree of difference between the target training data and the farthest-from-target training data is equal to or above a second threshold, and (iii) that the degree of difference between the target training data and the first reference training data is smaller than the degree of difference between the farthest-from-target training data and the first reference training data, during the operations of S210-S320.
As explained above, during the operations of S210-S320, the apparatus may obtain the target training data and the plurality of training data sets, identify one training data set nearest to the target training data as a target training data set, and determine whether to add the target training data to the target training data set. According to the operations of
In addition, the apparatus may generate the new group of training data including the target training data in response to determining not to add the target training data to the one group of the plurality of groups of training data.
In alternative embodiments, the selecting section 606 may determine a target training data set (i.e., one group) and the first reference training data by calculating an average of degrees of difference between the target training data and reference training data of each training data set, instead of performing S240-S250.
In these embodiments, the calculating section 604 may calculate an average degree of difference between the target training data and each of the plurality of reference training data of each training data set. In the example of
The selecting section 606 may identify a target training data set of which the average is the smallest among the plurality of training data sets. Then the selecting section 606 may select the first reference training data, which yields the smallest degree of difference from the target training data, from the identified one training data set.
In alternative embodiments, the apparatus may not perform some of operations S210-S320. In these embodiments, the apparatus may not perform S310-S320 and/or S280.
At S410, an obtaining section, such as the obtaining section 602, may obtain a plurality of target training data. Each of the plurality of target training data may be a candidate of the selected data set later generated and used for machine learning. Each of the plurality of target training data may be the same or similar to the target training data explained in relation to S120 and S220.
At S430, a selecting section, such as the selecting section 606, may select a target training data from the plurality of target training data obtained at S410. The selecting section 606 may select a target training data that has not been selected at previous operation(s) of the S430 in iterations of a loop of S430-S470.
At S450, the apparatus performs operations substantially similar to operations S230-S320 explained in relation to
At S460, the selecting section 606 may determine whether all target training data obtained at S410 have been selected at iterations of S430. If all the target training data have not been selected, then the selecting section 606 may go back to S430. If all the target training data have been selected, then the selecting section 606 may proceed with S470.
At S470, a forming section, such as the forming section 613, may form a selected data set based on a plurality of training data sets (i.e., a plurality of groups) generated at iterations of S450. The forming section 613 may pick up a predetermined number of training data from each of the plurality of training data sets to generate the selected data set. For example, when a training data set X, a training data set Y, and a training data set Z are generated during the iterations of S450 each including one thousand training data, the forming section 613 may randomly pick up one hundred training data from each of the training data set X, the training data set Y, and the training data set Z, and form a selected data set including three hundred training data.
At S480, a training section, such as the training section 614, may train an estimation model using training data of the selected data set formed at S470. In an embodiment, the estimation model may function as a soft sensor.
As explained above, during the operations of S410-S480, the apparatus may obtain a plurality of target training data, group the plurality of target training data into a plurality of training data sets based on the degree of difference, and train the estimation model based on training data selected from each of the plurality of training data sets. Since the training data of the selected data set are selected from different training data sets (i.e., different groups), the selected data set may enable training of the estimation model with balanced training data to improve accuracy of the estimation model.
The embodiments explained in relation to
In embodiments related to the operations of
In some embodiments, the first threshold and the second threshold are different among the plurality of training data sets. For example, a first threshold and a second threshold used for a first training data set are higher than a first threshold and a second threshold used for a second training data set, if a number of training data of the second training data set is larger than a number of training data of the first training data, in order to even the numbers of training data between the plurality of training data sets.
In another example, a first threshold and a second threshold used for a first training data set are higher than a first threshold and a second threshold used for a second training data set if an average of degrees of difference of all combinations of two training data of the second training data set is larger than an average of degrees of difference of all combinations of two training data of the first training data, in order to even the average of the degree of difference among the plurality of training data sets.
In some embodiments, the apparatus may divide one training data set into two or more new training data sets in response to determining that a number of training data in the one training data set exceeds a threshold. In these embodiments, the apparatus may divide one training data set such that the two or more new training data sets have the same or approximately the same number of training data, while a sum of averages of degrees of difference between training data in a set for each of the two or more new training data sets is minimized.
In some embodiments, the apparatus may perform some of the operations of
The apparatus may further perform S110-S190 using the selected data set as a training data set obtained at S110. Thereby, the apparatus may update the training data set used for training when a new target training data is obtained.
The computer 1200 according to the present embodiment includes a CPU 1200-12, a RAM 1200-14, a graphics controller 1200-16, and a display device 1200-18, which are mutually connected by a host controller 1200-10. The computer 1200 also includes input/output units such as a communication interface 1200-22, a hard disk drive 1200-24, a DVD-ROM drive 1200-26 and an IC card drive, which are connected to the host controller 1200-10 via an input/output controller 1200-20. The computer 1200 also includes legacy input/output units such as a ROM 1200-30 and a keyboard 1200-42. The keyboard 800-42 is connected to the input/output controller 1200-20 through an input/output chip 1200-40.
The CPU 1200-12 operates according to programs stored in the ROM 1200-30 and the RAM 1200-14, thereby controlling each unit. The graphics controller 1200-16 obtains image data generated by the CPU 1200-12 on a frame buffer or the like provided in the RAM 1200-14 or in itself, and causes the image data to be displayed on the display device 1200-18.
The communication interface 1200-22 communicates with other electronic devices via a network 1200-50. The hard disk drive 1200-24 stores programs and data used by the CPU 1200-12 within the computer 1200. The DVD-ROM drive 1200-26 reads the programs or the data from the DVD-ROM 1200-01, and provides the hard disk drive 1200-24 with the programs or the data via the RAM 1200-14. The IC card drive reads programs and data from an IC card, and/or writes programs and data into the IC card.
The ROM 1200-30 stores therein a boot program or the like executed by the computer 1200 at the time of activation, and/or a program depending on the hardware of the computer 1200. The input/output chip 1200-40 may also connect various input/output units via a parallel port, a serial port, a keyboard port, a mouse port, and the like to the input/output controller 1200-20.
A program is provided by computer readable media such as the DVD-ROM 1200-01 or the IC card. The program is read from the computer readable media, installed into the hard disk drive 1200-24, RAM 1200-14, or ROM 1200-30, which are also examples of computer readable media, and executed by the CPU 1200-12. The information processing described in these programs is read into the computer 1200, resulting in cooperation between a program and the above-mentioned various types of hardware resources. An apparatus or method may be constituted by realizing the operation or processing of information in accordance with the usage of the computer 1200.
For example, when communication is performed between the computer 1200 and an external device, the CPU 1200-12 may execute a communication program loaded onto the RAM 1200-14 to instruct communication processing to the communication interface 1200-22, based on the processing described in the communication program. The communication interface 1200-22, under control of the CPU 1200-12, reads transmission data stored on a transmission buffering region provided in a recording medium such as the RAM 1200-14, the hard disk drive 1200-24, the DVD-ROM 1200-01, or the IC card, and transmits the read transmission data to network 1200-50 or writes reception data received from network 1200-50 to a reception buffering region or the like provided on the recording medium.
In addition, the CPU 1200-12 may cause all or a necessary portion of a file or a database to be read into the RAM 1200-14, the file or the database having been stored in an external recording medium such as the hard disk drive 1200-24, the DVD-ROM drive 1200-26 (DVD-ROM 1200-01), the IC card, etc., and perform various types of processing on the data on the RAM 1200-14. The CPU 1200-12 may then write back the processed data to the external recording medium.
Various types of information, such as various types of programs, data, tables, and databases, may be stored in the recording medium to undergo information processing. The CPU 1200-12 may perform various types of processing on the data read from the RAM 1200-14, which includes various types of operations, processing of information, condition judging, conditional branch, unconditional branch, search/replace of information, etc., as described throughout this disclosure and designated by an instruction sequence of programs, and writes the result back to the RAM 1200-14.
In addition, the CPU 1200-12 may search for information in a file, a database, etc., in the recording medium. For example, when a plurality of entries, each having an attribute value of a first attribute is associated with an attribute value of a second attribute, are stored in the recording medium, the CPU 1200-12 may search for an entry matching the condition whose attribute value of the first attribute is designated, from among the plurality of entries, and reads the attribute value of the second attribute stored in the entry, thereby obtaining the attribute value of the second attribute associated with the first attribute satisfying the predetermined condition.
The above-explained program or software modules may be stored in the computer readable media on or near the computer 1200. In addition, a recording medium such as a hard disk or a RAM provided in a server system connected to a dedicated communication network or the Internet can be used as the computer readable media, thereby providing the program to the computer 1200 via the network.
The present invention may be a system, a method, and/or a computer program product. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present invention.
The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.
Computer readable program instructions described herein can be downloaded to respective computing/processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and/or a wireless network. The network may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and/or edge servers. A network adapter card or network interface in each computing/processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing/processing device.
Computer readable program instructions for carrying out operations of the present invention may be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) may execute the computer readable program instructions by utilizing state information of the computer readable program instructions to individualize the electronic circuitry, in order to perform aspects of the present invention.
Aspects of the present invention are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer readable program instructions.
These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and/or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function/act specified in the flowchart and/or block diagram block or blocks.
The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions/acts specified in the flowchart and/or block diagram block or blocks.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.
While the embodiments of the present invention have been described, the technical scope of the invention is not limited to the above described embodiments. It is apparent to persons skilled in the art that various alterations and improvements can be added to the above-described embodiments. It is also apparent from the scope of the claims that the embodiments added with such alterations or improvements can be included in the technical scope of the invention.
The operations, procedures, steps, and stages of each process performed by an apparatus, system, program, and method shown in the claims, embodiments, or diagrams can be performed in any order as long as the order is not indicated by “prior to,” “before,” or the like and as long as the output from a previous process is not used in a later process. Even if the process flow is described using phrases such as “first” or “next” in the claims, embodiments, or diagrams, it does not necessarily mean that the process must be performed in this order.
As made clear from the above, the embodiments of the present invention enable a learning apparatus learning a model corresponding to time-series input data to have higher expressive ability and learning ability and to perform the learning operation more simply.
