Fast approximate wavelet tracking on streams
View Patent ↗The first fast solution to the problem of tracking wavelet representations of one-dimensional and multi-dimensional data streams based on a stream synopsis, the Group-Count Sketch (GCS) is provided. By imposing a hierarchical structure of groups over the data and applying the GCS, our algorithms can quickly recover the most important wavelet coefficients with guaranteed accuracy. A tradeoff between query time and update time is established, by varying the hierarchical structure of groups, allowing the right balance to be found for specific data streams. Experimental analysis confirmed this tradeoff, and showed that all the methods significantly outperformed previously known methods in terms of both update time and query time, while maintaining a high level of accuracy.
1. A computer-implemented method for approximate wavelet tracking, comprising:
sketching a wavelet coefficient vector as updates arrive in a streaming model;
performing updates by using a hash-based grouping and search technique on a group-count sketch (GCS) synopsis comprising a three-dimensional array indexed by a repetition index, a bucket index, and a sub-bucket index; and
providing an estimate of an energy of at least one group of elements from the wavelet coefficient vector.
2. The method of claim 1 , further comprising:
using the GCS synopsis to track a set of large wavelet coefficients.
3. The method of claim 2 , wherein the set of large wavelet coefficients are identified in polylogarithmic time.
4. The method of claim 1 , wherein the hash-based group and search technique comprises:
for each repetition, computing the bucket index using a first hash function and computing the sub-bucket index using a second hash function; and
locating an element to update in the GCS synopsis with the bucket index and the sub-bucket index.
5. The method of claim 1 , wherein a synopsis space is sublinear, a per-item update time is sublinear, and a query time is sublinear.
6. A system for approximate wavelet tracking, comprising:
a processor;
a sketching component causing the processor to sketch a wavelet coefficient vector as updates arrive in a streaming model;
an update component causing the processor to perform updates by using a hash-based grouping and search technique on a group-count sketch (GCS) synopsis comprising a three-dimensional array indexed by a repetition index, a bucket index, and a sub-bucket index; and
an estimator causing the processor to provide an estimate of an energy of at least one group of elements from the wavelet coefficient vector.
7. The system of claim 6 , further comprising:
a tracker for using the GCS synopsis to track a set of large wavelet coefficients.
8. The system of claim 7 , wherein the set of large wavelet coefficients are identified in polylogarithmic time.
9. The system of claim 6 , wherein the hash-based group and search technique comprises:
for each repetition, computing the bucket index using a first hash function and computing the sub-bucket index using a second hash function; and
locating an element to update in the GCS synopsis with the bucket index and the sub-bucket index.
10. The system of claim 6 , wherein a synopsis space is sublinear, a per-item update time is sublinear, and a query time is sublinear.
11. A computer readable medium storing instructions for performing a method for approximate wavelet tracking, the method comprising:
sketching a wavelet coefficient vector as updates arrive in a streaming model;
performing updates by using a hash-based grouping and search technique on a group-count sketch (GCS) synopsis comprising a three-dimensional array indexed by a repetition index, a bucket index, and a sub-bucket index; and
providing an estimate of an energy of at least one group of elements from the wavelet coefficient vector.
12. The computer readable medium of claim 11 , further comprising:
using the GCS synopsis to track a set of large wavelet coefficients.
13. The computer readable medium of claim 12 , wherein the set of large wavelet coefficients are identified in polylogarithmic time.
14. The computer readable medium of claim 11 , wherein the hash-based group and search technique comprises:
for each repetition, computing the bucket index using a first hash function and computing the sub-bucket index using a second hash function; and
locating an element to update in the GCS synopsis with the bucket index and the sub-bucket index.
15. The computer readable medium of claim 11 , wherein a synopsis space is sublinear, a per-item update time is sublinear, and a query time is sublinear.