top of page
Writer's pictureCANA

December 2019 Newsletter

Updated: Sep 13, 2022


CANA Connection Newsletter

December 2019


On behalf of CANA Advisors, we wish all a safe, relaxing holiday season and a productive, prosperous New Year full of curiosity and opportunity!


We hope you enjoy reading about our team’s participation in the recent NeurIPS conference in western Canada as well as our profile of Walt DeGrange, CANA’s very own DAC! Finally, we spend some time reflecting on our CANA Foundation efforts this past year, and how we are leaning forward not only to provide solutions for our clients but also to contribute our time and focus to our communities we work and live in!


Cheers to 2020!

~ Team CANA

 


Highlights from NeurIPS 2019 in Vancouver BC

NeurIPS December 2019

By Lucia Darrow

location photo
NeurIPS 2019 in Vancouver BC

This December 13,000 AI researchers descended on Vancouver, British Columbia for the 33rd Conference on Neural Information Processing Systems (NeurIPS). This conference is the buzziest event in AI, with impressive sponsors and vendors and keynotes from the top researchers in the field. The work presented at NeurIPS represents the current state of AI research, and gives a glimpse of what to expect in the near future.


Lucia Darrow and Jerome Dixon of CANA Advisors attended the conference, seeking to discover the latest trends in different areas of AI research. Approaching the conference with an operations research perspective, Lucia focused on talks in the optimization track, from convex optimization to non-convex optimization as commonly found in neural network problems. From a life sciences perspective, Jerome engaged with the wealth of talks, posters and researchers in one of the most rapidly advancing and impactful areas of AI research: computational biology and health. Here are a couple of their takeaways:


Artificial Intelligence is successful in a narrow area of intelligence. Two keynote speakers Blaise Aguera y Arcas and Yoshua Bengio discussed the topic of intelligence and what changes in approach are needed to reach new levels in AI. Current methods are optimization-driven and lack flexibility. Models are trained to pass a test or meet a certain score, but don’t have the inherent attributes that make up human or animal intelligence. Aguera y Arcas argued for a shift to meta learning and evolution strategies while Bengio spoke about System 2 Intelligence, characterized by conscious reasoning and representation learning.

Monitor showing artificial ears hearing music
Demo: How does music sound to artificial ears?

AI researchers need to raise the bar on reproducibility of research. Human decision making is found at every step in the data science cycle, from data selection and cleaning to model and algorithm selection. In a talk titled, Veridical Data Science, Bin Yu of UC Berkeley called on researchers to focus on creating more reproducible and well-articulated research.


Reinforcement Learning and Federated Learning two trending areas. A large amount of talks and papers on these two topics. Reinforcement learning, is a type of dynamic programming that trains algorithms using a system of reward and punishment. A reinforcement learning algorithm, or agent, learns by interacting with its environment. The agent receives rewards by performing correctly and penalties for performing incorrectly. The agent learns without intervention from a human by maximizing its reward and minimizing its penalty function.

Federated Learning is a machine learning setting where the goal is to train a high-quality centralized model with training data distributed over a large number of clients each with unreliable and relatively slow network connections. A big use case for this is sensitive data (i.e. healthcare or HIPAA compliance). The algorithm learns on user data and then passes up the model but not the sensitive client information.


photo
Attendees at NeurIPS 2019

Energy and natural resource use as key considerations for the future machine learning. In several talks, the growing carbon footprint of deep neural nets was stressed. Training large models requires massive amounts of energy, much of which is spent in data transfer and redundant training. The normalization of emissions reporting on publications in this area is something to look out for in the next few years, especially with tools like ML CO2 Impact developed by Mila.


Bottom layer to the Big Data Stack. NeurIPS is heavy on the lower level AI algorithms from design to development. The developers of XGBoost, scikit-learn, tensorflow, and other well known algorithms in attendance and easy to meet and connect with. NeurIPS, Apache Spark AI Conference, and the Strata Big Data Conference are all highly recommended for any company or agency starting or planning to optimize their AI journey.

Interested in learning more? Check out recordings of NeurIPS sessions here: https://slideslive.com/neurips.


 


TEAM CANA MEMBER SHOWCASE

Walt DeGrange

Director of Analytics Capabilities


"Let not your mind run on what you lack as much as on what you have already."

Marcus Aurelius

Photograph of Walt DeGrange
Walt DeGrange | Director of Analytics Capabilities

CANA Team member Walt DeGrange has been with CANA Advisors for 5 and a half years and has over 17 years of experience in scoping, performing and implementing analytic solutions. He is an Operations Research Professional with implemented analytics solutions in Government, Industry, and Academia. He is currently the Course Director for the Military Operations Research Society (MORS) Critical Skills for Analytics Professionals and a past Chairperson of the INFORMS Sports Analytics Section.


Walt is also an educator and teaches online for the University of Arkansas and serves as an Executive Practicum Advisor at the NC State Poole School of Management. As a Team CANA speaker and course instructor, he contributes his experience and skills to the CANA Advisors’ Learning and Development seminars and workshops.


In his role as the Director of Analytics Capabilities (DAC) at CANA Advisors, he is responsible for recruiting, training, and equipping all of CANA's analytics professionals. Walt also leads CANA's sports analytics research area and coordinates the monthly CANA Advisors Roundtable or CAR as we like to call it. The CAR provides team members with a virtual online opportunity to share tips, analytical techniques, and discuss recent project successes. If you would like to learn more about Walt and how he and CANA Advisors can take your analytics to the next level, please contact him at wdegrange@canallc.com


 


CANA Foundation logo

CANA FOUNDATION

Three Years and Counting!

By Kenny McRostie


An important part of the success and culture at CANA Advisors is the CANA Foundation.


Founded in 2017, CANA Foundation is wrapping up its third full year. Based on a commitment to create positive change in the lives of people, organizations and the environment, CANA Foundation provides an avenue for the company and its employees to give back and share with others in both corporate and individual successes. Within the Foundation’s five focus areas – advancing veterans support & activities, health & wellness, civic duty & education, community planning & strategy, and environmental sustainment – CANA employees are able to give purpose to the work they do each day through giving back in areas they feel passionate and care about, with the company alongside and supporting all the way.


The year 2019 saw some very impactful initiatives. From development of an app to help the Colorado Mountain Club (CMC) monitor recreational impact in the Rocky Mountains to supporting one of CANA’s team members in their involvement with a local organization that provides comprehensive services for victims of sexual assault and violent crimes, CANA Foundation provided resources that enabled CANA’s employees to give back in their communities in some pretty incredible ways.



One initiative that stood out was undertaken by Kurt Eades, in an area with tremendous significance to him, Veteran Suicide Awareness. Kurt volunteered to help set up, run and clean up for the off-road fundraising event put on by Jon’s Mission for 22 to raise awareness and support for veteran suicide. A close friend of the founders of Jon’s Mission for 22, who lost their son to suicide after a long battle with Post Traumatic Stress Disorder (PTSD), Kurt served as a trail guide, videographer using CANA CDS01 Drone and his personal Jeep with a GoPro, safety, and generally anything that was asked of him. It was an incredible day of remembering, celebrating and grieving those who lost their battle with PTSD as over 250 off-roaders showed up to support the cause of veteran suicide awareness. CANA Foundation was honored to come alongside Kurt in this cause and provided paid time for a portion of the numerous hours Kurt gave to the event over a three-day period. The money raised from this event will go directly to Jon’s mission and several other organizations fighting hard to educate and help veterans, their families and those directly affected by veteran suicide. This is what CANA Foundation is all about – giving back and positively impacting people’s lives.


The 2019 initiative season caps off with the Foundation’s volunteer paid time off (VPTO) “Give Back” Campaign. In its third year, this is an initiative put forth by the company to allow each CANA employee to take up to eight hours of paid time off to volunteer in their local community supporting one or more of the Foundation's five focus areas. As we transition into 2020, the Foundation continues to evolve and grow, and our hope is that the Foundation’s impact will become even more tangible and genuine to the recipients with each volunteer minute spent, each dollar donated, and each act of kindness carried out.



 

Connect with CANA Advisors at one of these upcoming events in 2020!

 


CANA Advisors logo

CANA Advisors

7371 Atlas Walk Way Gainesville, Virginia 20155 Telephone (703) 317-7378 Facsimile (571) 248-2563


The CANA Connection Newsletter Copyright ©2019 CANA LLC.

89 views1 comment

Recent Posts

See All
bottom of page