Stochastic gradient-based optimization is crucial to optimize neural networks. While popular approaches heuristically adapt the step size and direction by rescaling gradients, a more principled approach to improve optimizers requires second-order information. Such methods precondition the gradient using the objective's Hessian. Yet, computing the Hessian is usually expensive and effectively using second-order information in the stochastic gradient setting is non-trivial. We propose using Information-Theoretic Trust Region Optimization (arTuRO) for improved updates with uncertain second-order information. By modeling the network parameters as a Gaussian distribution and using a Kullback-Leibler divergence-based trust region, our approach takes bounded steps accounting for the objective's curvature and uncertainty in the parameters. Before each update, it solves the trust region problem for an optimal step size, resulting in a more stable and faster optimization process. We approximate the diagonal elements of the Hessian from stochastic gradients using a simple recursive least squares approach, constructing a model of the expected Hessian over time using only first-order information. We show that arTuRO combines the fast convergence of adaptive moment-based optimization with the generalization capabilities of SGD.

Arturo, an AI-powered platform that derives property insights and predictive analytics from aerial and satellite imagery, has tapped Brett Antonides as vice president of apps, analytics and visualization and Dustin Montoya as vice president of platform infrastructure and engineering. With extensive industry knowledge and firsthand experience in geospatial technology, data analytics and artificial intelligence, these new executives will help refine Arturo's industry-leading technology and drive strategic initiatives and partnerships to support the company's continued expansion. "We recognize that a major factor in successfully scaling Arturo is our ability to add the best, brightest and most passionate people to our team," said Ben Tuttle, chief technology officer at Arturo. "We are thrilled to be able to make such strategic hires on the heels of our Series B fundraise that will help us continue to innovate our technology platform and grow our business. With the extensive technical and product backgrounds Brett and Dustin bring to the table, they will each play a major role in bolstering Arturo's capabilities and functionality, and will help drive the company's success in the years ahead."

A startup is employing machine learning to process aerial imagery and remotely analyze insurance risks to properties around the country. Why it matters: The combination of AI and aerial imagery from satellites and even stratospheric balloons can help insurers quickly judge property risks without an in-person visit, saving money and time. How it works: Arturo's AI model can identify potentially risky characteristics of a property -- like roof tiles in need of repair or a pool that lacks a fence -- and estimate the likelihood of an insurable accident in the future. Background: Arturo's business model is a combination of two major technological trends: the ever-increasing growth of aerial imagery that can capture detailed pictures of the ground and the power of machine learning. The big picture: Insurance might seem like the blandest of businesses, but since its origins hundreds of years ago, the field has focused on using available data to try to predict the future -- which happens to be precisely what machine learning is good at.

Arturo, Inc. –Arturo is a deep learning spin-out from American Family Insurance focused on delivering highly accurate measurement and predictive data for the Property & Casualty (P&C) Insurance, Reinsurance, REIT, and PERE markets. The company is headquartered in Chicago, IL, and has representation across North America to support both the Insurance and Re-Insurance Industry. Arturo has recently added to its locations a presence in Ottawa, Canada. Arturo, Inc. has raised a total of $8M in funding over two rounds. Their latest funding was raised on Apr 7, 2020, from a Series A round.

AI.Reverie - AI.Reverie is a simulation platform that trains AI to understand the world. They offer a suite of synthetic data and vision APIs to help businesses across different industries train their machine learning algorithms and improve their AI accuracy and repeatability. Key industries AI.Reverie has solutions for including Agriculture, Industrial, including managing construction sites, Smart Cities, and Smart Homes. AI.Reverie has raised a total of $5.6M in funding over three rounds. Their latest funding was raised on Apr 14, 2020.

Specialty re/insurer Canopius has partnered with technology start-up Arturo on artificial intelligence (AI) and deep-learning property analytics. The integration of Arturo's AI-powered technology will enable Canopius to gain access to the physical property characteristic and predictive analytics using the latest satellite, aerial, and ground-level data. As a result, Canopius will be able to make more informed and differentiated pricing decisions at the point of underwriting. Canopius chief digital officer Marek Shafer said: "Arturo's, AI-powered image analytics capability is hugely impressive. Canopius is excited to be harnessing this pioneering technology, which will help to fine-tune our risk selection process and improve point-of-sale underwriting."

CHICAGO & LONDON--(BUSINESS WIRE)--Arturo, the artificial intelligence and deep-learning property analytics start-up, today announces it has entered into a partnership with UK-based global specialty (re)insurer Canopius AG ("Canopius"). Arturo's deep-learning property analysis capability will be integrated into Canopius, providing physical property characteristic and predictive analytics, enabling more informed and differentiated pricing decisions at the point of underwriting Marek Shafer, chief digital officer, Canopius, said: "Arturo's, AI-powered image analytics capability is hugely impressive. Canopius is excited to be harnessing this pioneering technology, which will help to fine-tune our risk selection process and improve point-of-sale underwriting." John-Isaac Clark, CEO at Arturo said: "Arturo was built from the ground up within American Family Insurance and we deeply understand the specific challenges our insurance customers are trying to solve. We are incredibly excited to announce our relationship with Canopius and looking forward to working together to further enhance our product to meet their current and future needs."

What began as an assignment to the research and development team at American Family Insurance Co. now stands alone as an independent company and today announced its first major customer. Arturu, a Chicago-based insurtech startup launched by American Family three years ago, said in a press release that it has penned a deal with Hippo Insurance to provide "real-time" analytics on property characteristics. "Hippo's partnership with Arturo has allowed us to pre-fill very valuable property data and information directly into the customer's application including roof material, pool presence, and more," stated Michael Gulla, senior director of underwriting at Hippo. "These insights streamline the client's onboarding process, while helping us define a highly accurate initial rating and top-quality underwriting requirements." Hippo, based in Mountain View, California, began working with Arturo from its very start as an insurance provider, and even before Arturo became a company.

CHICAGO & MADISON, Wis.--(BUSINESS WIRE)--Arturo, an artificial intelligence (AI) and deep-learning property analytics start-up, announced today the close of Seed Series Transaction, with lead investor and Arturo co-developer American Family Insurance. Following the close of the transaction, Madison-based American Family Insurance will remain a customer as well as an investor in Arturo. Arturo grew out of investment and more than three years of research by American Family Insurance on the application of artificial intelligence (AI) and deep learning to satellite, aerial, drone, and ground-level imagery to accurately assess physical property characteristics for residential and commercial properties. Arturo's AI-powered analytics generate detailed property information often in under five seconds. It enables a variety of businesses that insure, lend, invest, or manage residential or commercial properties to make more informed decisions and better manage risk with the most up to date information available.