Recursive neural networks on future event prediction

1. A computer-implemented method for training a recursive neural network (RNN), the method comprising:

generating, by a processor using the RNN, a plurality of embedding vectors based on a plurality of observations, wherein the observations include (i) a subject, (ii) an action taken by the subject, and (iii) an object on which the subject is taking the action on, wherein the subject and object are constant;

generating, by the processor, based on a prediction score, predictions of one or more future events based on one or more comparisons of at least some of the plurality of embedding vectors, the RNN being trained to distinguish between likely and unlikely future events and to rank and score future events determined to be the likely events, wherein the prediction score represents a probability that an event Y will occur if an event X occurs first, a rank loss is calculated based on a determined accuracy of the predictions, and margin loss is minimized during training of the RNN; and

initiating, by the processor, based on the predictions, an action to a hardware device to mitigate expected harm to at least one item selected from the group consisting of the hardware device, another hardware device related to the hardware device, and a person related to the hardware device.

2. The method of claim 1 , wherein the predictions of one or more future events include a next subsequent event.

3. The method as recited in claim 1 , wherein the observations further comprise one or more video images, and wherein the RNN utilizes image recognition to interpret any of the observations that include the one or more video images.

4. The method as recited in claim 1 , wherein the embedding vectors include vectors selected from the group consisting of positive embedding vectors and negative embedding vectors.

5. The method as recited in claim 1 , wherein the predictions of one or more future events are calculated based on a positive embedding vector and a negative embedding vector.

6. The method as recited in claim 1 , wherein the observations include at least one previously unobserved observation.

7. The method as recited in claim 1 , wherein the observations include audio, wherein the RNN utilizes voice recognition to interpret the observations that include the audio.

8. The method as recited in claim 1 , wherein the initiating an action includes the RNN initiating the action utilizing an actuator mechanism.

9. The method as recited in claim 1 , wherein the initiating an action is automatically performed by the RNN.

10. The method as recited in claim 1 , wherein the observations further include a subset from a plurality of more than one action.

11. A system for training a recursive neural network (RNN), the system comprising:

a processor, configured to:

generate, using the RNN, a plurality of embedding vectors based on a plurality of observations, wherein the observations include (i) a subject, (ii) an action taken by the subject, and (iii) an object on which the subject is taking the action on, wherein the subject and object are constant;

generate based on a prediction score, predictions of one or more future events based on one or more comparisons of at least some of the plurality of embedding vectors, the RNN being trained to distinguish between likely and unlikely future events and to rank and score future events determined to be the likely events, wherein the prediction score represents a probability that an event Y will occur if an event X occurs first, a rank loss is calculated based on a determined accuracy of the predictions, and margin loss is minimized during training of the RNN; and

initiate based on the predictions, an action to a hardware device to mitigate expected harm to at least one item selected from the group consisting of the hardware device, another hardware device related to the hardware device, and a person related to the hardware device.

12. The system as recited in claim 11 , wherein the observations further comprise one or more video images, and wherein the RNN utilizes image recognition to interpret any of the observations that include the one or more video images.

13. The system as recited in claim 11 , wherein the embedding vectors include vectors selected from the group consisting of positive embedding vectors and negative embedding vectors.

14. The system as recited in claim 11 , wherein the observations include audio, wherein the RNN utilizes voice recognition to interpret the observations that include the audio.

15. The system as recited in claim 11 , wherein the predictions of one or more future events are calculated based on a positive embedding vector and a negative embedding vector.

16. The system as recited in claim 11 , wherein the observations include at least one previously unobserved observation.

17. A non-transitory computer-readable storage medium comprising a computer-readable program for training a recursive neural network (RNN), wherein the computer-readable program when executed on a computer causes the computer to perform the steps of:

generating, by a processor using the RNN, a plurality of embedding vectors based on a plurality of observations, wherein the observations include (i) a subject, (ii) an action taken by the subject, and (iii) an object on which the subject is taking the action on, wherein the subject and object are constant;

generating, by the processor, based on a prediction score, predictions of one or more future events based on one or more comparisons of at least some of the plurality of embedding vectors, the RNN being trained to distinguish between likely and unlikely future events and to rank and score future events determined to be the likely events, wherein the prediction score represents a probability that an event Y will occur if an event X occurs first, a rank loss is calculated based on a determined accuracy of the predictions, and margin loss is minimized during training of the RNN; and

initiating, by the processor, based on the predictions, an action to a hardware device to mitigate expected harm to at least one item selected from the group consisting of the hardware device, another hardware device related to the hardware device, and a person related to the hardware device.

18. The non-transitory computer-readable storage medium of claim 17 , wherein the predictions of one or more future events include a next subsequent event.

19. The non-transitory computer-readable storage medium of claim 17 , wherein the observations further comprise one or more video images, and wherein the RNN utilizes image recognition to interpret any of the observations that include the one or more video images.

20. The non-transitory computer-readable storage medium of claim 17 , wherein the observations include audio, wherein the RNN utilizes voice recognition to interpret the observations that include the audio.