Mapper component for a neuro-linguistic behavior recognition system
Techniques are disclosed for generating a sequence of symbols based on input data for a neuro-linguistic model. The model may be used by a behavior recognition system to analyze the input data. A mapper component of a neuro-linguistic module in the behavior recognition system receives one or more normalized vectors generated from the input data. The mapper component generates one or more clusters based on a statistical distribution of the normalized vectors. The mapper component evaluates statistics and identifies statistically relevant clusters. The mapper component assigns a distinct symbol to each of the identified clusters.
1. A method comprising:
receiving a normalized vector of feature values generated from input data, each feature value from the normalized vector of feature values being calculated based on a feature from a plurality of features;
for each feature value in the normalized vector of feature values:
evaluating a distribution of a plurality of clusters in a cluster space corresponding to the feature from the plurality of features that is associated with the feature value,
mapping the feature value to a single cluster from the plurality of clusters based on the distribution,
updating the distribution of the plurality of clusters based on the mapping to produce an updated distribution, and
determining, based on the updated distribution, whether or not to merge the plurality of clusters with a further cluster from the cluster space;
determining a symbol for a statistically significant cluster from the plurality of clusters; and
transmitting the symbol to a behavior recognition system.
2. The method of claim 1 , wherein the normalized vector of feature values is a first normalized vector of feature values, the input data is first input data, and the updated distribution is a first updated distribution, the method further comprising:
receiving a second normalized vector of feature values generated from second input data; and
for each feature value in the second normalized vector of feature values:
mapping the feature value to a single cluster from the plurality of clusters,
updating the distribution of the plurality of clusters based on the mapping, to produce a second updated distribution, and
upon determining, based on the second updated distribution, that the single cluster has statistical significance, outputting a symbol from a plurality of symbols associated with the single cluster.
3. The method of claim 1 , further comprising:
determining, based on the distribution of the plurality of clusters, that at least two clusters from the plurality of clusters overlap in the cluster space, and
merging the at least two clusters in response to determining that the at least two clusters overlap in the cluster space.
4. The method of claim 1 , wherein a statistical significance of each cluster from the plurality of clusters is determined based on a statistical significance score that indicates a number of the feature values that map to the cluster over time.
5. The method of claim 4 , further comprising, upon determining that feature values mapping to the cluster have not been received after a period based on a function of time, decaying the statistical significance of the cluster.
6. The method of claim 1 , wherein statistics associated with each cluster from the plurality of clusters include a mean, a variance, and a standard deviation.
7. The method of claim 1 , wherein each feature value in the normalized vector of feature values is within a range of 0 and 1, inclusive.
8. A computer-readable storage medium storing instructions, which, when executed on a processor, performs an operation comprising:
receiving a normalized vector of feature values generated from input data, each feature value from the normalized vector of feature values being calculated based on a feature from a plurality of features;
for each feature value in the normalized vector of feature values:
evaluating a distribution of a plurality of clusters in a cluster space corresponding to the feature from the plurality of features that is associated with the feature value,
mapping the feature value to a single cluster from the plurality of clusters based on the distribution,
updating the distribution of the plurality of clusters based on the mapping to produce an updated distribution, and
determining, based on the updated distribution, whether or not to merge the plurality of clusters with a further cluster from the cluster space;
determining a symbol for a statistically significant cluster from the plurality of clusters; and
transmitting the symbol to a behavior recognition system.
9. The computer-readable storage medium of claim 8 , wherein the normalized vector of feature values is a first normalized vector of feature values, the input data is first input data, and the updated distribution is a first updated distribution, the operation further comprising:
receiving a second normalized vector of feature values generated from second input data; and
for each feature value in the second normalized vector of feature values:
mapping the feature value to a single cluster from the plurality of clusters,
updating the distribution of the plurality of clusters based on the mapping, to produce a second updated distribution, and
upon determining, based on the second updated distribution, that the single cluster has statistical significance, outputting a symbol from a plurality of symbols associated with the single cluster.
10. The computer-readable storage medium of claim 8 , wherein the operation further comprises:
determining, based on the evaluated distribution of the plurality of clusters, that at least two clusters from the plurality of clusters overlap in the cluster space, and
merging the at least two clusters in response to determining that the at least two clusters overlap in the cluster space.
11. The computer-readable storage medium of claim 8 , wherein a statistical significance of each cluster from the plurality of clusters is determined based on a statistical significance score that indicates a number of the feature values that map to the cluster over time.
12. The computer-readable storage medium of claim 11 , wherein the operation further comprises, upon determining that feature values mapping to the cluster have not been received after a period based on a function of time, decaying the statistical significance of the cluster.
13. The computer-readable storage medium of claim 8 , wherein statistics associated with each cluster from the plurality of clusters include a mean, variance, and standard deviation.
14. The computer-readable storage medium of claim 8 , wherein each feature value in the normalized vector of feature values is within a range of 0 and 1, inclusive.
15. A system, comprising:
a processor; and
a memory storing one or more application programs configured to perform an operation comprising:
receiving a normalized vector of feature values generated from input data, each feature value from the normalized vector of feature values being calculated based on a feature from a plurality of features;
for each feature value in the normalized vector of feature values:
evaluating a distribution of a plurality of clusters in a cluster space corresponding to the feature from the plurality of features that is associated with the feature value,
mapping the feature value to a single cluster from the plurality of clusters based on the distribution,
updating the distribution of the plurality of clusters based on the mapping to produce an updated distribution, and
determining, based on the updated distribution, whether or not to merge the plurality of clusters with a further cluster from the cluster space;
determining a symbol for a statistically significant cluster from the plurality of clusters; and
transmitting the symbol to a behavior recognition system.
16. The system of claim 15 , wherein the normalized vector of feature values is a first normalized vector of feature values, the input data is first input data, and the updated distribution is a first updated distribution, the operation further comprising:
receiving a second normalized vector of feature values generated from second input data; and
for each feature value in the second normalized vector of feature values:
mapping the feature value to a single cluster from the plurality of clusters,
updating the distribution of the plurality of clusters based on the mapping, to produce a second updated distribution, and
upon determining, based on the second updated distribution, that the single cluster has statistical significance, outputting a symbol from a plurality of symbols associated with the single cluster.
17. The system of claim 15 , wherein the operation further comprises:
determining, based on the evaluated distribution of the plurality of clusters, that at least two clusters from the plurality of clusters overlap in the cluster space, and
merging the at least two clusters in response to determining that the at least two clusters overlap in the cluster space.
18. The system of claim 15 , wherein a statistical significance of each cluster from the plurality of clusters is determined based on a statistical significance score that indicates a number of the feature values that map to the cluster over time.
19. The system of claim 15 , wherein statistics associated with each cluster from the plurality of clusters include a mean, variance, and standard deviation.
20. The system of claim 15 , wherein each feature value in the normalized vector of feature values is within a range of 0 and 1, inclusive.