Today, we are interviewing Renee Yao, Founder and CEO of Neo Ivy Capital. We started our conversation by comparing notes on the very limited number of women we encountered at the Neural Information Processing Systems (NIPS) conference in Barcelona, in 2016 (both of us attended unbeknownst to one another). We found that, in addition to our interest in machine learning and other neural information processing techniques, we also share with Renee a mission to encourage women, especially younger generations, to embrace a career in quantitative finance and investing. Like her, we are committed to increasing the level of female representation in this fascinating field. This article is another step towards our common goal!
What Contributed to Renee’s Success?We asked Renee to tell us the things that contributed most to her success. She conveyed three basic themes that characterized her professional career:
- Theme #1: DEDICATION
- Theme #2: INNOVATION
- Theme #3: INSPIRATION
Renee’s next frontier: Reinforcement Learning and Deep Convolutional NetworksThese days, it is common in asset management circles, to toss around words like Machine Learning (ML), Deep Learning (DL) and so on (here is a good primer on the subject by Druce Vertes). These collective terms form the foundation for the concept of Artificial Intelligence, or AI. But what is AI? Renee shares that it took her two years to come up with a straightforward explanation to share with her potential investors.
AI is a process designed to train a machine so that the machine will think and analyze data similar to a human being including having the ability to improve itself spontaneously. This process is quite different from traditional data mining methods that typically uses statistical techniques to fit long time series of historical data and ultimately is capable of building only static models.Renee and her current team of five researchers began studying and focusing on AI back in 2014 when, thanks to their strong ties to academic circles, they ran into the working version of the Alpha Go paper, which was later published in 2016 in Nature. It is advantageous to participate in academic research groups, where papers are discussed long prior to their official publication and often provides a running head start on seminal research. The Alpha Go paper is one example, where prior exposure was key to Renee’s ability to apply the techniques of deep learning neural networks to her investment process. Alpha Go has its roots in the ancient Chinese game of Go. Alpha Go functions by selecting game’s moves via supervised learning from human experts and reinforcement learning from self-play. If this sounds like a foreign language to you, Renee has you covered. Investor education is as important (if not more) as the complexities of the algorithms. She built a nice demo that can be found on the Neo Ivy Capital website. Check it out here. It shows, in a very impactful way, how reinforcement learning works. There has already been an evolution of Alpha Go, namely Alpha Go Zero, based on a new paper published last year in Nature. This time the algorithm functions only with reinforcement learning, without human data, guidance or domain knowledge beyond game rules. It is like Alpha Go is teaching itself and the results have been very good considering Alpha Go Zero defeated 100-0 Alpha Go, which in turn defeated the human world champion of the game. To quote Professor David Silver:
…with fewer data and less computational intensity, Alpha Go Zero is performing much better than its predecessor, which was programmed by using millions of moves.Renee believes this is exciting progress. In traditional data mining, static models are used—the researcher discovers a pattern and applies it the same way into the future with no opportunity to learn from its mistakes. Artificial Intelligence is dynamic, constantly improving and learning. So what do her clients think? Renee always gets a couple of questions from clients and potential clients: what is your edge and how long will you maintain this edge? She uses visuals (here) to show how different her approach is from other forms of statistical modeling. She believes that the dynamic nature of her models, which are based on continuous learning will help keep her edge for longer. Critics of AI are worried about the poor interpretability of these models. We asked Renee how she deals with explaining the performances of her strategies to her investors. First of all, she points out that due to the complexity and idiosyncrasies of her investment process, it is critical to surround yourself with clients who share your vision and views.