Multi-Domain Analytics

ENSIGN solves a critical problem in the application of machine learning in data science – how to gain deep insight from the entirety of massive-scale multidimensional unlabeled data while supporting, but not requiring, heroic up-front feature engineering. The approach builds on advanced, well-founded techniques from the field of spectral hypergraph analytics.

These techniques have been extended and made computationally tractable using Reservoir’s patented data structures and proprietary algorithmic advances. Combined with supporting tools, these advances enable a broad range of potential use cases, scalable and streaming operation, and the ability to leverage a variety of computing configurations.

Learn more about ENSIGN

Core Capabilities

ENSIGN is written in ANSI C and can run on any size platform from desktop PCs to massive shared and distributed-memory supercomputers with or without GPUs. Using Dask for parallelism and an in-memory Python workflow, ENSIGN supports massive scale data from end to end.

Hypergraphs are more than just expanded graphs. The time of being limited to statistical measures and basic graph metrics like centrality is over. ENSIGN provides a way to make smarter use of large-volume linked data that exploits rather than flattens intrinsic multidimensional correlations.

ENSIGN includes support for generating reports and visualizing results. With ENSIGN, it is possible to see the patterns in data and understand the influence of specific actors.

No need for labeling or heroic feature engineering. Just ingest and go. ENSIGN supports a variety of multi-domain use cases in security, forensic analysis, and scientific discovery. Exploratory tools provide novel ways to examine and visualize interesting correlations.

Joint analysis allows two or more datasets with overlapping dimensions to be analyzed together without the combinatorial explosion of a relational join. Streaming analysis allows new data to be added to an existing analysis. Both operations provide massive, practical computational efficiency for modern use cases.

ENSIGN is callable from a Python interface and includes supporting analytic tools written in Python. This means ENSIGN can seamlessly interoperate with leading edge data science tools and packages such as Anaconda.

See it in Action

ENSIGN engineers have developed Jupyter notebooks to showcase a variety of problems that ENSIGN can address. We’ve highlighted a few of our projects below through the link below.

The notebooks showcase code, narrative text, and equations, to offer a glimpse into ENSIGN as a data science tool and to highlight informative applications.

Meet Some of Our Team

James Ezick

VP Engineering
Bio

Muthu Baskaran

Fellow & Managing Engineer
Bio

Dimitri Leggas

Research Engineer
Bio

Brendan von Hofe

Research Engineer
Bio

Get in touch with one of our experts today

The Latest

Recent Publications

An All–at–Once CP Decomposition Method for Count Tensors

CANDECOMP/PARAFAC (CP) tensor decomposition is a popular method for detecting latent behaviors in real– world data sets. As data sets grow larger and more elaborate, more sophisticated CP decomposition algorithms are required to enable these discoveries. Data sets from many applications can be represented as count tensors. To decompose count

Read More »

Filtered Tensor Construction and Decomposition for Drug Repositioning

Drug repositioning (also called “drug repurposing”) is a drug development strategy that saves time and money by finding new uses for existing drugs. While a variety of computational approaches to drug repositioning exist, recent work has shown that tensor decomposition, an unsupervised learning technique for finding latent structure in multidimensional

Read More »

Knowledge-guided Tensor Decomposition for Baselining and Anomaly Detection

We introduce a flexible knowledge-guided penalty for incorporating known or expected patterns of activity into tensor decomposition. Our modified tensor decomposition both enables efficient identification of semantically-related patterns across data sets and provides a means for identifying anomalous patterns. Specifically, we modify the loss function for a CP tensor decomposition

Read More »