Algorithms are the beating heart of AIOps. Innovative AI research keeps the blood flowing.
As Chief Science Officer, I head the AI research and development team here at Moogsoft. Our R&D team is proud of the deep bank of intellectual property we’ve developed and commercialized. As of last count, Moogsoft’s AI research team has authored 50 patents for original and unique algorithms such as Vertex Entropy, Tempus, Cookbook and Probable Root Cause. The suite of algorithms applied in Moogsoft AIOps Platform spans both kinds of machine learning: unsupervised, which identifies patterns in large data sets; and supervised, which learns by example over time.
This month, our CEO Phil Tee, the R&D team, and myself all got to participate at what we scientists consider the fun part of our jobs – presenting new AI research to our peers at IEEE. The 16th International Symposium on Integrated Network Management (IM19) was held from April 8-12 in Washington, DC. The overarching theme of this year’s conference was the “intelligent management of the next wave of cyber and social networks”.
Advances in data analytics and machine learning for network and service management were featured in a special workshop called Analytics for Network and Service Management, or AnNet for short.
Moogsoft’s R&D team sponsors doctoral student projects in #datascience from prominent UK universities. One PhD candidate recently presented original AI research at #IEEE #IM19 conference’s #AnNet workshop.
Moogsoft CEO Presents Keynote Address at AnNet 2019 Workshop
Phil Tee delivered the opening keynote presentation on “The Unreasonable Effectiveness of Graph Theory”. His talk detailed the use of AI and algorithmic techniques in fault management and network analysis.
As anyone in IT Operations challenged with managing today’s complex environment knows, correlating system and network events to determine root cause isn’t a walk in the park. Techniques applied over the last few decades have ranged from rules based expert systems to the more modern approach of mathematical algorithms. A curious feature of the space is the repeated motif of graph theory as a common thread in all approaches. (Still with me?) Phil’s talk surveyed some of the manifestations of graph theory across different applications, and identified that the convenient equivalence between graphs and matrices as potentially being a driving factor in the “unreasonable effectiveness of graphs” for fault localization.
Did I lose you? Trust me, he was brilliant.
Moogsoft R&D Team Presents 2 Original AI Research Papers at IEEE IM19
In addition, we presented the following original AI research papers at AnNet and IM19 respectively:
- A Method for Temporal Event Correlation; co-authored by Phil and myself
- Functional Topology Inference from Network Events; co-authored by the University of Sussex and the Moogsoft R&D team; including Antoine Messager, George Parisis, Istvan Kiss, Luc Berthouze, Phil and me
(Unfortunately, these papers are available only from IEEE, so online distribution and access is restricted to current members.)
Finally, it’s important to note that ongoing academic outreach is a key part of our data science success. Moogsoft R&D provides a real-world AI research laboratory for doctoral candidates to pursue their theses. They benefit, but also we benefit from continually tapping the best and brightest minds to expand our thinking in new directions. In the U.K., we are currently hosting PhD students from Oxford University and University of Sussex. This process of identifying graduate projects for future sponsorship continues, with Moogsoft customers being the primary beneficiaries of our work.
