Get Outta Dodge Big Data, Sparse Modelling’s The New Sheriff in Town

Sparse Modelling

With so many industry experts and academics shouting from the rooftops that Big Data is the future of business, there is little opposition coming out that, in fact, there are other approaches that can be just as commercially successful in the long run.

One of them, little known to those outside the world of data science, is in the field of small data, or Sparse Modelling, an approach to data processing that one Japanese AI startup believes will reap success for industrial and medical applications.

HACARUS, founded in 2014 by Kenshin Fujiwara, is a Kyoto-based company whose proprietary AI engine is built using Sparse Modelling.

But what exactly is this Sparse Modelling and how will it help customers?

HACARUS

HARARUS’ Sparse Modelling-based AI can make highly accurate predictions by taking advantage of its three main strengths:

The technology does not require large amounts of data for training; it can provide explainable solutions; its lightweight design offers high speeds and low power consumption in a variety of environments.

And this is where HACARUS’ LASSO platform comes in — a core enabler for Sparse Modelling

“LASSO, a key algorithm behind Sparse modeling was first developed around 25 years ago. It is uncertain about how it was created, but when Professor Robert Tibshirani of Stanford University proposed LASSO, it became widely recognized in the field of data science.”

 — HACARUS

HACARUS even presents examples of Sparse Modelling outclassing deep learning (DL) by comparing its “Sparse Modelling based approach with Classifier (SVM) and Deep Learning (CNN) techniques for detection of defects on Solar Cells, concluding that Sparse Modelling far outperforms the competition. Not only is accuracy higher, but it also creates AI models faster — even when using a far smaller dataset.”

With 100+ AI projects across the medical and manufacturing fields undertaken since its founding, HACARUS’ IP is proven to work.

Along with its proprietary technology, HACARUS also offers several services. These include:

• HACARUS Inspect

applying Sparse Modeling based AI to provide accurate and reliable inspections, that allows humans to make faster and better decisions

• HACARUS Lens

providing cost-effective and efficient solutions through our high-precision optical reading AI

• HACARUS Dojo

ensuring successful AI introduction and method selection

• HACARUS Edge

Through COLIGO, its Edge AI platform, HACARUS provides customers with a wide range of applications and services for tailor-made development, including implementations for IoT, FPGA and other edge use cases

The main man behind HACARUS’ rise is Founder and CEO Kenshin Fujiwara. A founder/co-founder of several Japanese startups, Fujiwara has experience and expertise in business development in AI, IoT, web, mobile, B2B domains, hiring, fundraising, handling communication with VCs, strategy planning, alliance with partner companies, setting up regional branches/offices, and selling company asset or whole companies via M&A.

In October 2019, Fujiwara was named the Challenging Spirt 2019 for the Kansai Region, in the yearly EoY competition hosted by Ernst & Young.

He obtained a BS in Computer Science from California State University-Northridge.

Backed by the likes of Osaka Gas and Miyako Capital (Kyoto University), in 2018 HACARUS raised approximately $2.5 million in a Series A round — this was after seed funding amounting to some $930,000 back in 2016 and 2017. In 2020, it managed to raise an undisclosed amount during a Series B round, which inside sources close to the deal estimated to be in the “millions of US dollars.”

Realizing DL methods require a vast amount of training data, can only do inference on the edge while requiring internet connectivity to re-train the model and — in Fujiwara’s own words — DL algorithms are “a black box” quite unfathomable to humans, it seems certain that HACARUS’ AI technology is finding answers to some big problems of DL-based AI and proves Sparse Modelling has a commercial future.