Refined learning data representation for classifiers
In one embodiment, a learning machine device initializes thresholds of a data representation of one or more data features, the thresholds specifying a first number of pre-defined bins (e.g., uniform and equidistant bins). Next, adjacent bins of the pre-defined bins having substantially similar weights may be reciprocally merged, the merging resulting in a second number of refined bins that is less than the first number. Notably, while merging, the device also learns weights of a linear decision rule associated with the one or more data features. Accordingly, a data-driven representation for a data-driven classifier may be established based on the refined bins and learned weights.
1. A method, comprising:
initializing, at a learning machine device, thresholds of a data representation of one or more data features, the thresholds specifying a first number of pre-defined bins;
reciprocally merging adjacent bins of the pre-defined bins having substantially similar weights, the merging resulting in a second number of refined bins that is less than the first number;
simultaneously learning weights of a linear decision rule associated with the one or more data features while merging; and
establishing a data-driven representation for a data-driven classifier based on the refined bins and learned weights.
2. The method as in claim 1 , wherein the pre-defined bins are uniform and equidistant.
3. The method as in claim 1 , further comprising:
using the data-driven classifier on traffic in a computer network.
4. The method as in claim 1 , further comprising:
sharing the data-driven classifier with one or more other devices.
5. The method as in claim 1 , wherein the data-driven classifier is a linear support vector machine (SVM).
6. The method as in claim 1 , wherein substantially similar weights comprise one or more of equal weights, similar weights, and weights having a same sign.
7. The method as in claim 1 , wherein the one or more features comprise a feature pair, and wherein correlation between feature values of the feature pair is used for the data representation.
8. The method as in claim 1 , wherein the one or more features are binarized.
9. The method as in claim 1 , wherein the one or more features are real-valued.
10. An apparatus, comprising:
a processor configured to execute one or more processes; and
a memory configured to store a process executable by the processor, the process when executed operable to:
initialize thresholds of a data representation of one or more data features, the thresholds specifying a first number of pre-defined bins;
reciprocally merge adjacent bins of the pre-defined bins having substantially similar weights, the merging resulting in a second number of refined bins that is less than the first number;
simultaneously learn weights of a linear decision rule associated with the one or more data features while merging; and
establish a data-driven representation for a data-driven classifier based on the refined bins and learned weights.
11. The apparatus as in claim 10 , wherein the pre-defined bins are uniform and equidistant.
12. The apparatus as in claim 10 , wherein the process when executed is further operable to:
use the data-driven classifier on traffic in a computer network.
13. The apparatus as in claim 10 , wherein the process when executed is further operable to:
share the data-driven classifier with one or more other devices.
14. The apparatus as in claim 10 , wherein the data-driven classifier is a linear support vector machine (SVM).
15. The apparatus as in claim 10 , wherein substantially similar weights comprise one or more of equal weights, similar weights, and weights having a same sign.
16. The apparatus as in claim 10 , wherein the one or more features comprise a feature pair, and wherein correlation between feature values of the feature pair is used for the data representation.
17. The apparatus as in claim 10 , wherein the one or more features are binarized.
18. The apparatus as in claim 10 , wherein the one or more features are real-valued.
19. A tangible, non-transitory, computer-readable media having software encoded thereon, the software when executed by a processor operable to:
initialize thresholds of a data representation of one or more data features, the thresholds specifying a first number of pre-defined bins;
reciprocally merge adjacent bins of the pre-defined bins having substantially similar weights, the merging resulting in a second number of refined bins that is less than the first number;
simultaneously learn weights of a linear decision rule associated with the one or more data features while merging; and
establish a data-driven representation for a data-driven classifier based on the refined bins and learned weights.
20. The computer-readable media as in claim 19 , wherein the pre-defined bins are uniform and equidistant.