Vector database sharding
Multimodal search
JaguarDB quantization
JaguarDB Vector API
Best Vector databases
JaguarDB in Docker
Setup JaguarDB with tar package
Setup JaguarDB on multiple nodes
Vector index sharing
How zeromove works
Video introduction
|
|
Example: Anomaly Detection
An anomaly occurs when feature values deviate from the norm considerably enough to be regarded as a rare exception. Jaguar detects anomalies based on the assumption that such outliers, or exceptions, should stand apart from the significant portion of data in the dataset.Anomaly detection can be adopted in applications such as fraud detection in industries of insurance and banking, intrusion detection in computer networks and national surveillance, medical informatics in diagnosis and disorder detection, and fault/damage detection in commerce and industry.
The following anomaly detection statement is used to check if a query vector is anomalous:
select anomalous(v, '0.1, 0.2, 0.3, 0.4, 0.5, 0.3, 0.1',
'type=euclidean_fraction_float, sigmas=3,activation=0.3')
from vectab
jag.query(qs)
while jag.fetch():
print(jag.jsonString())
In this context, the field "v" may represent a vector from textual data or imagery data. The search process entails a comparison between the query vector and the other vectors within the vector database, assessing whether the query vector exhibits a significant deviation from the rest. This deviation is measured using standard deviation. The optional parameter "sigmas" defines how many times the query vector must deviate from the mean of the collected vectors in the table to be considered significant. The optional parameter "activation" is a value between 0 and 1, specifying the percentage of vector components that must surpass the number of standard deviations to be deemed significant. The default values are "sigmas" set to 2 and "activation" set to 0.3. The results are returned in JSON format:
{"anomalous":"YES","prate":"0.388671875", "activation":"0.3","sigmas":"2"}
In the JSON result, string "YES" reports that the query vector is deemed anomalous. The value of "prate" indicates a rating value 0.388671875 which exceeds the activation value of 0.3 in this case. If the value prate is more than the activation level, the query vector will be considered anomalous. User can fine-tune the paramters to suite their own data patterns to achieve optimized detection accuracy.
|
