Method and system for automatic and interactive model training using domain knowledge in video games

1. A computer-implemented method of allowing a user to automatically transform domain knowledge into a machine learning model to be used in real-time operation of video games, the method comprising:

providing a user interface via a user computing device, the user interface being configured to allow a user to input domain knowledge relating to a video game by specifying-one or more labeling functions;

receiving, from the user computing device, the labeling functions;

transforming the labeling functions into executable code;

labeling raw data relating to the video game using the executable code to obtain labeled data;

applying an automated machine learning module to the labeled data to obtain a machine learning model; and

transmitting, to the user computing device, the obtained machine learning model and metrics associated with the obtained machine learning model,

wherein the user interface is further configured to allow the user to adapt or change the labeling functions based on the metrics for use in obtaining a further machine learning model.

2. The computer-implemented method of claim 1 , wherein the machine learning model obtained by the application of the automated machine learning module is a robust machine learning model in which residual noise in the labeled data has been eliminated.

3. The computer-implemented method of claim 1 , wherein a hyperparameter optimization is performed by the automated machine learning module to find an optimal machine learning model.

4. The computer-implemented method of claim 1 , wherein the machine learning model obtained is used as part of an integrity checking module running on an online games server in order to detect anomalies during real-time operation of the video game, wherein the method further comprises, at the integrity checking module: generating an output of the machine learning model based on input data related to real-time operation of the video game, wherein the output of the machine learning model indicates whether an anomaly is present during real-time operation of the video game.

5. The computer-implemented method of claim 4 , wherein the output of the machine learning model indicates whether a player of the video game is cheating.

6. The computer-implemented method of claim 5 , wherein the labeling function is based on an accuracy in firing.

7. The computer-implemented method of claim 6 , wherein the labeling function defines that if a player's accuracy in firing is greater than 98%, then the player must be cheating.

8. The computer-implemented method of claim 1 , wherein the machine learning model obtained is a trained neural network or a trained random forest.

9. The computer-implemented method of claim 1 , wherein applying the automated machine learning module to the labeled data to obtain a machine learning model comprises:

storing the labeled data in a data store;

distributing the labeled data among compute nodes where the automated machine learning module is applied to the labeled data to obtain the machine learning model; and

storing the obtained machine learning model back in the data store.

10. The computer-implemented method of claim 1 , wherein the labeling functions are specified in the form of rules defined in a visual language.

11. The computer-implemented method of claim 1 , wherein the raw data relating to the video game is derived from or comprises input control data obtained from a game controller.

12. The computer-implemented method of claim 11 , wherein the raw data relating to the video game comprises sequential data that represents actions performed by a player in a gaming environment.

13. The computer-implemented method of claim 1 , further comprising:

outputting, via the user interface, the metrics in the form of a metrics table which ranks individual features associated with the labeling functions according to their importance in the obtained machine learning model; and

allowing, via the user interface, the user to analyse the features and, if required, to reformulate or add further labeling functions for use in obtaining the further machine learning model in an iterative manner.

14. The computer-implemented method of claim 13 , further comprising

using the metrics to determine a level of noise of the labeling function; and

in response to determining that the level of noise of the labeling function is above a threshold level, allowing the user to adapt or change the labeling function in an iterative manner.

15. The computer-implemented method of claim 1 , wherein the user interface is a drag-and-drop menu that allows a user to define the labeling functions in the form of rules by selecting from a list of features, an operator, a threshold value and a conclusion.

16. A non-transitory computer readable medium comprising a trained machine learning model to be used in real-time operation of video games, wherein the model is obtained by performing the following steps:

providing a user interface via a user computing device, the user interface being configured to allow a user to input domain knowledge relating to a video game by specifying one or more labeling functions;

receiving, from the user computing device, the labeling functions;

transforming the labeling functions into executable code;

labeling raw data relating to the video game using the executable code to obtain labeled data;

applying an automated machine learning module to the labeled data to obtain a machine learning model; and

transmitting, to the user computing device, the obtained machine learning model and metrics associated with the obtained machine learning model;

wherein the user interface is further configured to allow the user to adapt or change the labeling functions based on the metrics for use in obtaining a further machine learning model.

17. A distributed computing system, comprising

a non-transitory data store on which a program resides that when executed is arranged to:

receive labeling functions input, via a user interface on a user computing device, by a user based on his domain knowledge; and to transform the received labeling functions into executable code;

label raw data relating to a video game using the executable code to obtain labeled data; and

distribute the labeled data to a plurality of compute nodes; and

the plurality of compute nodes on which a program resides that when executed is arranged to:

receive the labeled data; apply the automated machine learning algorithm to the labeled data to obtain a machine learning model;

transmit, to the user computing device, the obtained machine learning model and metrics associated therewith, wherein the user interface is further configured to allow the user to adapt or change the labeling functions based on the metrics; and

the program further arranged to receive the adapted or changed labeling functions for use in obtaining a further machine learning model; and

transmit, to the user computing device, the obtained machine learning model and metrics associated with the obtained machine learning model, wherein the user interface is further configured to allow the user to adapt or change the labeling functions based on the metrics for use in obtaining the further machine learning model; and

the program further arranged to receive the adapted or changed labeling functions for use in obtaining the further machine learning model.

18. The distributed computing system of claim 17 , further comprising the user computing device on which a program resides that when executed is arranged to enable a user to input the labeling functions based on his domain knowledge in the user interface.

19. The distributed computing system of claim 17 , further comprising an online games server comprising one or more processors configured to perform an integrity checking procedure to detect anomalies during real-time operation of the video game, the integrity checking procedure comprising generating an output of the machine learning model based on input data related to real-time operation of the video game, wherein the output of the machine learning model indicates whether an anomaly is present during real-time operation of the video game.

20. The distributed computing system of claim 19 , wherein the output of the machine learning model indicates whether a player of the video game is cheating.