Classifications

• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
  • G06F17/10—Complex mathematical operations
  • G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q10/00—Administration; Management
  • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q30/00—Commerce
  • G06Q30/02—Marketing; Price estimation or determination; Fundraising

Definitions

• a user-item pair consists of one of m users and one of n items such that the recommendation value for the specific user-item pair is a recommendation value of a specific item for a specific user.
• Such recommendation values may be used in many recommender systems.
• Recommender systems generate recommendations in terms of items that are likely to be of interest to a particular user.
• ratings data is stored in the users-items matrix. In general, only a fraction of this matrix is filled. From this data, the system should be able to predict the remaining ratings in terms of recommendation values.
• Factorisation techniques as a realisation of latent factor models are no stranger to rating based recommender systems.
• matrix factorisation is defined for completely filled matrices. Its extension for partly filled matrices implies the restoration of large parts of missing data, leading to an optimisation problem for which the choice of constraints is a major issue.
• Matrix factorisation techniques as a realisation of latent factor models are used in the development of many recommender systems. For example, see “Learning collaborative information filters” by Billsus and Pazzani (International Conference on Machine Learning, Morgan Kaufmann Publishers, 1998),
• the data may be imagined as a cloud.
• the position and the shape of the cloud are determined by the centre of mass and the principal axes/components.
• a reference frame is attached to the cloud's centre of mass.
• Covariance matrix determines the principal components. If the variation is high in k dimensions and low in others then the data may effectively be viewed as a / -dimensional cloud.
• the cloud is represented by its projections onto the one dimensional spaces determined by the principal components.
• Every rating matrix determines two clouds where corresponding principal components are factors and cofactors and the variations are described by the singular values. Knowing the singular values, principal factors and cofactors, the matrix can be reproduced; this is the essence of singular value decomposition.
• the main advantage of projections is that the system may be described with a smaller number of variables while the main information is still in the database.
• the dimensionality of the problem is reduced by putting to zeros singular values that are less than a certain threshold.
• the rating matrix is represented as a sum of one rank matrices. Each matrix corresponds to one (co)factor. The number of factors is reduced by removing the non-essential factors. The larger the inner product of the
• R is an m ⁇ n rating matrix that is partly filled with user ratings.
• r ti is known only if user / rated item j.
• the set of observed ratings is denoted by K, so
• K ⁇ (/ ' , j, nj) I n j is known ⁇ is relatively small.
• the optimisation can be solved by a stochastic gradient descent or alternating least squares. See, for example, "The BellKor 2009 Solution to the Netflix Grand Prize” by Koren (August 2009, available online at
• the optimisation problem is undetermined because the data set is sparsely filled.
• the existing approach does not distinguish users with a lot of ratings from users with fewer ratings.
• Another aspect of the problem is
• the present invention seeks to provide an alternative way of generating
• a user-item pair corresponds to one of m users and one of n items such that the recommendation value for said specific user-item pair is a recommendation value of a specific item of the n items for a specific user of the m users.
• Each user of the m users is associated with a corresponding user vector.
• Each item of the n items is associated with a corresponding item vector.
• the method comprises the steps of: generating the user vector associated with the specific user; generating the item vector associated with the specific item; and generating the recommendation value for the specific user-item pair based on a dot product of the user vector associated with the specific user and the item vector associated with the specific item.
• the method includes the following steps (a) to (c) for each stage in a series of stages. Each stage has a respective item threshold value and a respective user threshold value. The item threshold value for a stage is less than the item threshold values for any preceding stages, and the user threshold value for a stage is less than the user threshold values for any preceding stages.
• Step (a) involves identifying selected users from the m users and selected items from the n items so as to define a subset of user-item pairs.
• step (b) involves calculating the associated user vector based on known recommendation values for user-item pairs in the subset corresponding to said selected user and based on any corresponding item vectors calculated in any preceding stages.
• step (c) For each selected item that has not been selected in any preceding stages, step (c) involves calculating the associated item vector based on known recommendation values for user-item pairs in the subset corresponding to said selected item and based on any corresponding user vectors calculated in any preceding stages.
• the method of the present invention splits the optimisation problem into a series of stages. Each stage helps to build up a full prediction model that is able to generate recommendation values for specific user-item pairs.
• the first stage uses a block (i.e. subset) of data from an area of user-item space that is more densely populated with known representation values (i.e. ratings).
• Parameters i.e. user vectors and item vectors
• the second stage uses a block of data with a lower threshold for the level of population of known representation values in the user-item space.
• Previously calculated parameters i.e. user vectors and item vectors
• Calculation of parameters in subsequent stages does not affect the parameters which were calculated in any previous stages using more densely populated data.
• Splitting the data into blocks reduces the uncertainty of the problem and provides a new iterative method of solving such optimisation problems.
• An important consideration for any prediction is the quality. Intuitively, more densely filled data should produce a more reliable prediction, and this concept has been used in the present methodology.
• the present methodology also allows the introduction of the concept of reliability in terms of the density of the data.
• the step (a) for at least one stage in the series, in the step (a), for each unselected user, the number of known recommendation values for user- item pairs not in the subset is less than or equal to the user threshold value, and for each unselected item, the number of known recommendation values for user- item pairs not in the subset is less than or equal to the item threshold value.
• the unselected users are the users from the m users that are not selected to define the subset of user-item pairs.
• the unselected items are the items from the n items that are not selected to define the subset of user-item pairs.
• this methodology is used for each stage in the series.
• the first stage of the series uses all of the areas of user-item space that are most densely populated with known recommendation values.
• the second stage additionally encompasses all of the slightly less densely populated areas of user-item space, and so on.
• this embodiment will generally yield very high quality predictions of recommendation values for given user-item pairs.
• the steps (b) and (c) may together comprise minimising the difference between: 1 ) known recommendation values for each user-item pair in the subset, and 2) corresponding dot products of the user vectors for the selected users and the item vectors for the selected items.
• the method requires a minimisation (or optimisation) problem to be solved at one or more stages in the series.
• minimisation problems e.g. least squares, iterative techniques
• this embodiment uses a methodology that is well understood by those skilled in the art.
• the plurality of iterations of the steps (b) and (c) may commence with the step (b).
• the first iteration of the step (b) further comprises calculating said associated user vector based on randomly initialised item vectors for those selected items that have not been selected in any preceding stages.
• the plurality of iterations of the steps (b) and (c) may commence with the step (c).
• the first iteration of the step (c) further comprises calculating said associated item vector based on randomly initialised user vectors for those selected users that have not been selected in any preceding stages. Random initialisation is one possible way of initialising the item vectors or user vectors.
• any iterations of the step (b) which follow a previous iteration of the step (c) may further comprise calculating said associated user vector based on item vectors calculated in the previous iteration of the step (c).
• any iterations of the step (c) which follow a previous iteration of the step (b) further comprise calculating said associated item vector based on any user vectors calculated in the previous iteration of the step (b).
• the most recently calculated item vectors are used to calculate any user vectors
• the most recently calculated user vectors are used to calculate any item vectors.
• the step (b) and/or the step (c) may comprise solving a system of linear equations. Reducing the optimisation problem to a set of linear equations makes it possible to solve accomplish the optimisation in an efficient and easily understandable manner.
• the item threshold value is equal to the user threshold value such that only a single threshold value is required for each stage. This is useful for asymmetric data sets.
• the item threshold value may be different to the user threshold value for a given stage if desired.
• each user vector comprises k elements and each item vector comprises k elements.
• the user vectors associated with the selected users and the item vectors associated with the selected items each have a constant number C of elements set to zero, where C is greater than or equal to 1 . This enables the dimensionality of the problem to be reduced in one or more of the stages in the series.
• the method may further comprise the steps of: receiving a new known recommendation value for a user-item pair, wherein the new known
• recommendation value affects which users and items are selected in step (a) for each of stages s to S in the series, wherein stage S denotes the final stage in the series; and, for each of stages s to S in the series, performing the steps (a) to (c) again so as to update the calculated user vectors and item vectors based on the new known recommendation value.
• the present method can be adapted to the more practical situation of dynamically evolving datasets of recommendation values (i.e. situations where dynamics is involved in the ratings).
• a computer program which, when executed by a processor, causes the processor to carry out the method of the first aspect.
• a data carrying medium carrying the computer program of the second aspect.
• the data carrying medium may be a storage medium or a transmission medium.
• an apparatus comprising a processor, the processor being configured to carry out the method of the first aspect.
• Figure 1 shows an exemplary rating matrix R with some ratings known, and other ratings missing.
• Figure 3 illustrates the components of matrix U that is used in the construction of matrix A in Figure 2.
• Figure 4 illustrates the components of matrix V that is used in the construction of matrix A in Figure 2.
• Figure 5 illustrates the number of variables in the optimisation problem for each block.
• Figure 6 schematically illustrates the system of linear equations solved for every column during one iteration of step 2 of the iterative process.
• Figure 7 schematically illustrates the system of linear equations solved for every row during one iteration of step 3 of the iterative process.
• Figure 9 shows a comparison of the original data with the final results of the iterative process illustrated in Figure 8.
• Figure 10 is a Table showing the minimum number of ratings p s in every block of the second MovieLens data set which was used for evaluation purposes.
• Figure 1 1 shows Mean Average Error of various evaluated predictions for the second MovieLens data set.
• Figure 12 shows Root Mean Square Error of various evaluated predictions for the second MovieLens data set.
• Figure 13 schematically illustrates a recommender system according to an embodiment of the invention.
• Figure 14 schematically illustrates an example computer system which may be used to form a data processing system forming the whole or a part of the recommender system and/or one or more user computer systems.
• the diagonal matrix ⁇ is an m ⁇ n matrix with nonnegative singular values on its diagonal. Assuming decreasing order of the singular values gives a uniquely determined matrix ⁇ .
• the matrix A is approximated by selecting only the k largest singular values and their corresponding vectors from U and V, the so-called / -rank
• A be an m ⁇ n matrix of rank k such that An is a k ⁇ k nonzero minor. Then A may be written in the following four-block structure:
• A- - As 4ii is a k ⁇ k matrix with nonzero determinant, it is an invertible
• the Schur decomposition factor ises the block matrix into a lower triangle, a block-diagonal, and an upper triangle matrix:
• the rank is not known, but suppose that it is known and let it be k. Thus, there exists at least one full rank k ⁇ k block; all the blocks of higher dimension are not full rank. Suppose it is possible to find this block.
• this block may be used as to construct a block matrix (5).
• the sin ular value decomposition is the following:
• the matrix A is first constructed:
• R is the sparsely filled m ⁇ n matrix that contains all known ratings; / is a k x k identity matrix; U and V are m k and n / matrices respectively.
• Corresponding examples of matrices U and V are shown in Figures 3 and 4 respectively.
• K is a set of known ratings in R. This is equivalent to the optimisation problem (1 ). In other words, it is desired to minimise the difference between the known ratings in R and the dot products of the corresponding user vectors and item vectors.
• the matrix R is split into S blocks Si ... B s depending on the number of known values for each user and for each item.
• this splitting procedure is accomplished by permuting the rows and columns of the matrix R . Whilst the permutation of rows and columns aids the mathematical processing of the ratings data, it will be appreciated that this permutation is not essential.
• the data is split into blocks by virtue of data density. Having split matrix R into blocks, all users and all items from block B ⁇ have at least M ⁇ ⁇ ratings, all users and all items from block B 2 have at least M 2 ratings with M 2 ⁇ Mi , and so on.
• the user threshold value is equal to the item threshold value such that only a single threshold value M s is required for each stage s.
• Si is the densest data block
• B 2 is the second densest data block
• ... is the least dense data block:
• the present method involves a series of stages.
• a first stage of the method starts with the densest block Si , and uses the iterative process described below to find U ⁇ ⁇ and ⁇ / ⁇ such that:
• next densest block B 2 is taken, and the block S-i from the previous step is incremented.
• the proposed incremental approach yields the possibility to get dynamically changing recommender systems by updating the recommender system regularly.
• a user-item pair may be able to be moved from the initial block B s to a denser one S s- i .
• L/ s- i ... Us, and V s- i ... V s will be affected.
• the matrices of factors U and cofactors V for higher density data blocks will remain unaffected.
• the users' tastes depend linearly on the tastes of others and, also, that item ratings are linearly dependent as well.
• the set of independent users/items has dimension k.
• the rating for an item is dependent on user behaviour and on how an item is rated in relation to others. This may give predicted values that are outside the rating scale (e.g., higher than 5 or less than 1 for a rating scale of 1 to 5). In fact, this is one of the advantages of the present method because items are really distinguished, e.g., an item with rating 5.9 should be appreciated more by a user than an item with rating 5.2. It is not necessary to use baseline predictions because the method itself takes care of deviations from the average for every user and every item simultaneously.
• Figure 5 shows that the number of variables in the optimisation problem is k (n s - n s -i + m s - m s .- ⁇ ), and the number of equations is N s , where N s is the number of known ratings for all items and users in the s th block.
• N s is the number of known ratings for all items and users in the s th block.
• step 4 Repeat steps 2 and 3 until stop criterion is satisfied. The iterative process converts very fast and after a few iterations it is reasonable to stop.
• the values of U are randomly initialised as per step 1 of the iterative process.
• the randomly initialised values of U are clearly shown in Figure 8b.
• step 2 of the iterative process these random values of U are fixed and a system of linear equations is solved to find the values of V.
• Figure 9 shows a comparison between the original ratings data and the predicted ratings.
• U and V are also shown in Figure 9.
• U s would not need to be randomly initialised in this embodiment.
• k depends on the data set in question, k may be dependent on the number n of items of the number m of users to some extent. In addition, k is dependent on the density of known ratings in the data set. k is chosen experimentally.
• the balance between the number of equations and the number of unknowns may be viewed in a different way.
• the number of equations in each system depends on the number p, of known values in the corresponding row or column.
• the number of variables k may also be made dependent on this number and it may be varied for different rows and columns.
• the value of k for that block i.e. k s
• this is achieved by setting some of the k elements of the user vectors and item vectors to zero so as to reduce the dimensionality of the problem.
• Reliability of prediction is defined as the quantity and quality of the information in a block.
• the block approach allows distinguishing the reliability of the prediction. The denser the block, the more ratings there are for every item and every user, the more factors should be retrieved, and the more reliable and accurate the results of the prediction are in comparison to the less dense blocks.
• the most reliable and accurate predicted block is the first block Si and the less reliable one is the last block B s .
• the second MovieLens data set is used for evaluation here.
• This data set contains ratings as explicit feedback from the users, information about the users and items with different properties and different sizes. Additionally, this data sets contain time stamps, which are not used in the present method. All users in the data set have rated at least 20 movies on a scale of 1 to 5. There are 1 ,000,209 ratings from 6,040 users on 3,593 items in the data set. The results presented here are found using the following predefined parameters:
• the blocks were defined based on minimum numbers of ratings for every user and every item in the particular block (see the Table shown in Figure 10). Thus, for the first block this value is 173, for the second block the value is 153, and so on until the last block where the minimum number of ratings is 13. Predictions are generated for all users but not for all items as the minimum number of ratings for users is initially 20. There are nine blocks in total. The first block is the most densely populated with ratings, and the ninth block is the most sparsely filled. The first column in the Table of Figure 10 gives the rounded number of users in every block and the first row in the table gives the rounded number of items in every block.
• Figure 1 1 shows MAE for all ratings and MAE for positive ratings (ratings where the predicted values are > 3) in three cases: “predicted”, “scaled” and “rounded".
• the "predicted” values are those taken directly from the model. As previously mentioned, the predicted ratings can be higher than 5 or lower than 1 . Thus, the "scaled” values are scaled such that all ratings are between 1 and 5. Of course, this decreases the MAE, as shown in Figure 1 1.
• the "rounded” values have been rounded to the nearest integer. Figure 1 1 shows that rounding decreases MAE even further. This happens because there are only integers in the test data set. Thus, on average, the predicted values are close to the real ones. Positive ratings have been evaluated separately because, for a
• the data set is positive and contains 83% of positive ratings.
• the MAE for the positive ratings (using 156,000 data points) is less than the MAE for all ratings (using 200,000 data points).
• Figure 12 shows RMSE for all ratings and RMSE for positive ratings in two cases: “predicted” and “scaled”.
• the "rounded” predictions are not included because rounding makes certain errors larger due to the nature of the RMSE, and the large errors are weighted more than the small errors.
• Figures 1 1 and 12 show that MAE and RMSE increase as the data density within each block becomes more sparse. This confirms that reliability of the prediction depends on the density of ratings within a given data block.
• Rating-less recommender systems are systems that contain information without ratings.
• a rating based recommender system for films may allow a user to rate films they have watched on a scale of 1 to 5.
• a rating-less recommender system for films may simply indicate whether or not a user has watched particular films by having a "rating" of 1 for films that are known to have been watched, with all other entries being empty. The empty user-item pairs may or may not have been watched by the user.
• a rating-less recommender system concerns transactions. Such a system would contain entries where a user was known to have bought an item or used a specific service.
• the present methodology may also be used for binary recommender systems in which, for example, entries of 0 may indicate user-item pairs where the user did not like the item, and entries of 1 may indicate user-item pairs where the user did like the item.
• a new matrix factorisation methodology has been presented that allows the incremental generation of predictions starting from the densest part of the ratings data matrix. Splitting the ratings data into blocks allows an optimisation problem to be solved for each block using an iterative process. The dimensionality of the problem is also reduced in this way. The method has low complexity.
• the reliability of the system has been linked with the density of the blocks considered at every iteration, making reliability an intrinsic part of the method. As intuitively expected the denser the block the more accurate the prediction.
• the most popular items and the most experienced users contain enough information to define key factors and cofactors, while other users and items can be characterised only by some factors and cofactors reducing the uncertainty of the problem.
• the described embodiment has been evaluated using the publically available MovieLens data set, and shows good performance.
• the obtained results outperform existing methods that were evaluated on the same data set.
• the existing methods used for comparison are described in "Mixed Membership Matrix Factorization" by Mackey et al. (2010, Proceedings of the 27th
• the methodology of the described embodiment copes well and efficiently with dynamic situations. With time, new ratings appear and a user-item pair as entry in the matrix can be moved from an initial block to a denser one. In this case, only factors U and cofactors V for the current block and the blocks that are more sparse than it need be recalculated, while not touching the other (denser) blocks.
• Figure 13 schematically illustrates a system 1300 according to an embodiment of the invention.
• the system 1300 comprises one or more user computer systems 1302 that are communicably connected to a recommender system 1306 via a network 1304. Whilst Figure 13 illustrates the system 1300 as comprising three user computer systems 1302, it will be appreciated that the system 1300 may comprise any other number of user computer systems 1302.
• the user computer systems 1302 and the recommender system 1306 may each comprise one or more computer systems, examples of which shall be described below with reference to Figure 14.
• the network 1304 may be any network capable of communicating data between a user computer system 1302 and the recommender system 1306.
• the network 1304 may comprise one or more of: the internet, a local area network, a wide area network, a metropolitan area network, a broadcast network, a satellite communications network, a telephone network, etc.
• a user may make use of a user computer system 1302 to provide (via the network 1304) a rating (or recommendation value) of an item to the recommender system 1306.
• the recommender system 1306 may process such ratings in the manner described above.
• the recommender system 1306 may generate and provide (via the network 1304) a recommendation value of an item to a user of a user computer system 1302 (again, in the manner described above).
• recommender system 1306 may be arranged to provide such a recommendation value in response to receiving a corresponding request for a recommendation value that the user of the user computer system 1302 has provided (via the network 1304) to the recommender system 1306. For example, the
• recommender system 1306 may host a website comprising one or more webpages via which a user may provide recommendation values of particular items and/or request and receive item recommendation values that the
• Figure 14 schematically illustrates an example computer system 1400 which may be used to form a data processing system forming the whole or a part of the recommender system 1306 and/or one or more of the user computer systems 1302.
• the system 1400 comprises a computer 1402.
• the computer 1402 comprises: a storage medium 1404, a memory 1406, a processor 1408, a storage medium interface 1410, an output interface 1412, an input interface 1414 and a network interface 1416, which are all linked together over one or more communication buses 1418.
• the storage medium 1404 may be any form of non-volatile data storage device such as one or more of a hard disk drive, a magnetic disc, an optical disc, a ROM, etc.
• the storage medium 1404 may store an operating system for the processor 1408 to execute in order for the computer 1402 to function.
• the storage medium 1404 may also store one or more computer programs (or software or instructions or code) that form part of an embodiment of the invention.
• the memory 1406 may be any random access memory (storage unit or volatile storage medium) suitable for storing data and/or computer programs (or software or instructions or code) that form part of an embodiment of the invention.
• the processor 1408 may be any data processing unit suitable for executing one or more computer programs (such as those stored on the storage medium 1404 and/or in the memory 1406) which have instructions that, when executed by the processor 1408, cause the processor 1408 to carry out a method according to an embodiment of the invention and configure the system 1400 to be a system according to an embodiment of the invention.
• the processor 1408 may comprise a single data processing unit or multiple data processing units operating in parallel, in cooperation with each other, or independently of each other.
• the processor 1408, in carrying out data processing operations for embodiments of the invention, may store data to and/or read data from the storage medium 1404 and/or the memory 1406.
• the storage medium interface 1410 may be any unit for providing an interface to a data storage device 1422 external to, or removable from, the computer 1402.
• the data storage device 1422 may be, for example, one or more of an optical disc, a magnetic disc, a solid-state-storage device, etc.
• the storage medium interface 1410 may therefore read data from, or write data to, the data storage device 1422 in accordance with one or more commands that it receives from the processor 1408.
• the input interface 1414 is arranged to receive one or more inputs to the system 1400.
• the input may comprise input received from a user, or operator, of the system 1400; the input may comprise input received from a device external to or forming part of the system 1400.
• a user may provide input via one or more input devices of the system 1400, such as a mouse (or other pointing device) 1426 and/or a keyboard 1424, that are connected to, or in communication with, the input interface 1414.
• a mouse or other pointing device
• keyboard 1424 a keyboard
• the system may comprise a microphone 1425 (or other audio transceiver or audio input device) connected to, or in communication with, the input interface 1414, the microphone 1425 being capable of providing a signal to the input interface 1414 that represents audio data (or an audio signal).
• the computer 1402 may store the input received from the/each input device
• the output interface 1412 may be arranged to provide a graphical/visual output to a user, or operator, of the system 1400.
• the processor 1408 may be arranged to instruct the output interface 1412 to form an image/video signal representing a desired graphical output, and to provide this signal to a monitor (or screen or display unit) 1420 of the system 1400 that is connected to the output interface 1412.
• the output interface 1412 may be arranged to provide an audio output to a user, or operator, of the system 1400.
• the processor 1408 may be arranged to instruct the output interface 1412 to form an audio signal representing a desired audio output, and to provide this signal to one or more speakers 1421 of the system 1400 that is/are connected to the output interface 1412.
• network interface 1416 provides functionality for the computer 1402 to download data from and/or upload data to one or more data
• communication networks such as the Internet or a local area network.
• the architecture of the system 1400 illustrated in Figure 14 and described above is merely exemplary and that other computer systems 1400 with different architectures and additional and/or alternative components may be used in embodiments of the invention, and that not all of the components mentioned above may be present.
• some or all of the input devices e.g. the keyboard 1424, the microphone 1425 and the mouse 1426
• the output devices e.g. the monitor 1420 and the speaker 1421
• the input devices e.g. the keyboard 1424, the microphone 1425 and the mouse 1426
• the output devices e.g. the monitor 1420 and the speaker 1421
• peripheral devices communicatively coupled to the computer 1402 (e.g. via a cable and/or wirelessly).
• Figure 14 and the discussion thereof provide an exemplary computing architecture, this is presented merely to provide a useful reference in discussing various aspects of the invention.
• the description of the architecture has been simplified for purposes of discussion, and it is just one of many different types of architecture that may be used for embodiments of the invention. It will be appreciated that the boundaries between logic blocks are merely illustrative and that alternative embodiments may merge logic blocks or elements, or may impose an alternate decomposition of functionality upon various logic blocks or elements.
• the system 1400 may be any type of computer system, such as one or more of: a games console, a set-top box, a personal computer system, a mainframe, a minicomputer, a server, a workstation, a notepad, a personal digital assistant, and a mobile telephone. It will be appreciated that, insofar as embodiments of the invention are implemented by a computer program, then a storage medium and a transmission medium carrying the computer program form aspects of the invention.
• the computer program may have one or more program instructions, or program code, which, when executed by a computer (or a processor) carries out an embodiment of the invention.
• program may be a sequence of instructions designed for execution on a computer system, and may include a subroutine, a function, a procedure, an object method, an object implementation, an executable application, an applet, a servlet, source code, object code, a shared library, a dynamic linked library, and/or other sequences of instructions designed for execution on a computer system.
• the storage medium may be a magnetic disc (such as a hard drive or a floppy disc), an optical disc (such as a CD-ROM, a DVD-ROM or a BluRay disc), or a memory (such as a ROM, a RAM, EEPROM, EPROM, Flash memory or a portable/removable memory device), etc.
• the transmission medium may be a communications signal, a data broadcast, a communications link between two or more computers, etc.

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• Business, Economics & Management (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Strategic Management (AREA)
• Development Economics (AREA)
• Economics (AREA)
• Mathematical Physics (AREA)
• Entrepreneurship & Innovation (AREA)
• Mathematical Optimization (AREA)
• Finance (AREA)
• Operations Research (AREA)
• General Business, Economics & Management (AREA)
• Marketing (AREA)
• Human Resources & Organizations (AREA)
• Computational Mathematics (AREA)
• Mathematical Analysis (AREA)
• Accounting & Taxation (AREA)
• Game Theory and Decision Science (AREA)
• Pure & Applied Mathematics (AREA)
• Data Mining & Analysis (AREA)
• Algebra (AREA)
• Databases & Information Systems (AREA)
• Software Systems (AREA)
• General Engineering & Computer Science (AREA)
• Quality & Reliability (AREA)
• Tourism & Hospitality (AREA)
• Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Priority Applications (1)

Applications Claiming Priority (3)

Publications (1)

Family

ID=46508014

Family Applications (2)

Family Applications Before (1)

Country Status (4)

Families Citing this family (6)

Family Cites Families (4)

• 2011
  • 2011-07-15 EP EP11174265A patent/EP2546759A1/de not_active Withdrawn
• 2012
  • 2012-07-03 WO PCT/EP2012/062925 patent/WO2013010787A1/en active Application Filing
  • 2012-07-03 EP EP12734881.1A patent/EP2732387A1/de not_active Withdrawn
  • 2012-07-03 CN CN201280044875.0A patent/CN103782290A/zh active Pending
  • 2012-07-12 ZA ZA2012/05185A patent/ZA201205185B/en unknown

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events