Back

Speaker "Ayush Sarkar" Details Back

 

Topic

Probabilistic Sequential Pattern Mining for Event-Transaction Queries

Abstract

In this talk, a three-tier pipeline architecture will be presented that leverages the advantages of sequential pattern mining for log processing under a probabilistic paradigm. The transactions occurring in a database management system are modeled as Bayesian networks adhering to the Markov assumption. The system employs probabilistic graph processing techniques for extracting "actionable knowledge" by executing sequential pattern mining algorithms on normalized logs. The implementation provides an online query-based system where a user inquires about a particular event in an event log and the system outputs the log lines estimated to have occurred in the same transaction. In the experiment, a 43.17% recall has been achieved for the relevant log lines, demonstrating a trade-off between the recall and computation cost.
Who is this presentation for?
Data scientists, AI/ML practitioners, Database designers, architects and developers.
Prerequisite knowledge:
Database and Machine Learning
 

Profile

Ayush Sarkar has multiple years of experience in solving real-world business problems in various industry verticals involving AI/ML from edge to cloud. Ayush is currently pursuing a Master’s degree in Computer Science at the University of Illinois, Urbana-Champaign, with a focus on artificial intelligence, machine learning, software architecture, mathematical models, smart contracts in blockchain, and integration of AI and blockchain. He developed a novel three-tier architecture to reduce bandwidth for streaming 360 videos, employing super resolution techniques at the edge for enhanced video quality of experience (QoE). At Intel Corporation, Ayush worked on enhancing Batch GEMM (General Matrix-Matrix Multiplication) operations for 1D dilated Convolution Neural Networks (1D_CNNs) for applications like NLP, speech recognition, etc. Ayush has multiple patents and publications.