Our approach when designing and implementing our technology is inspired by, and in many ways, builds upon foundational concepts outlined in the following books:
- “Superforecasting, The Art and Science of Prediction,” by Philip Tetlock and Dan Garder
- “The Beginning to Infinity, Explanations that Transform the World,” by David Deutsch
- “The Wisdom of Crowds,” by James Surowiecki
Concepts and Ideas
These books offer a wealth of information and wisdom when it comes to forecasting, but there are three main concepts that inform our solution design that we want to highlight as an introduction to our conceptual methodology:
- In Superforecasting, Tetlock summarizes his findings by defining two broad groups of experts: hedgehogs and foxes. Those who displayed poorer superforecasting skills (hedgehogs) tended to organize their thinking around Big Ideas. They sought to squeeze complex problems into preferred cause-effect templates. The other group consisted of more pragmatic experts; these foxes, “gathered as much information from as many sources as they could. They talked about possibilities and probabilities, not certainties, and they readily admitted when they were wrong and changed their minds” Foxes are better at forecasting than hedgehogs.
- In the Beginning of Infinity, by David Deutsch, he argues the need to constantly adapt to changing inputs to make precise predictions. This is reinforced by Tetlock in Superforecasting, where he introduces the concept of “perpetual beta”, which acknowledges and reinforces the benefit of constant reevaluation, updating and self-improvement.
- In Wisdom of Crowds, Surowiecki shows that with the aggregation of information in groups, it results in decisions that are often better than could have been made by any single member of the group.
Solution Design and Technology
These three ideas go a long way to introducing the conceptual methodology behind our platform. They are therefore reflected in our solution design and technology, firstly in the form of our rapid-prototyping solution architecture; secondly, through our approach to data, algorithm and parameter exploration via the Auto-series feature; and thirdly through our UI, in the form of the Group Dashboard, which ensembles and stacks the models into an aggregate output visualization.
Build models in minutes, no data science expertise required. Facilitated through Data Lake and streamlined automated data ingestion and model building workflow.
A comprehensive framework and process for building accurate and robust machine learning models that constantly learn and adapt to changing conditions.
Combines model outputs via user-defined or automated success criteria (filters) that leverages the wisdom of multiple models produced via rapid prototyping and auto-series.
These three aspects of our technology in many ways underpin the success of our platform insofar as they facilitate a dynamic process of continuous improvement via exploration, assessment and adaptation. A process that isn’t focused on one “Big Idea” (hedgehogs), but explores thousands of permutations and combinations of data, algorithm selection and parameter possibilities (foxes), and then aggregates the individual model outputs into a challenger/champion framework (wisdom of crowds). We then revisit each hypothesis periodically to adapt to changing conditions and inputs, where winners are reselected based on the new results, and the process repeats ad infinitum (perpetual beta). The more compute resources the user can allocate to this process, the better the results.
Where do we go from here?
This post is intended as an introduction to the conceptual methodology and these components by no means define the entirety of what we have designed into Quantum ML. In the following series of posts we will walk you through different aspects of our solution so as to better outline, define and deep dive into what’s behind our technology and solution. Our primary goal is to bring clarity and confidence to our existing (and potential) customers, so we encourage you to reach out to us with any questions, suggestions and even criticisms. Our philosophy with regards to constant adaptation and improvement isn’t just limited to our technology, but our business practices too.
Quantum Data Technologies is a data science services and solutions company based out of Vancouver, BC, and New York. We specialize in Machine Learning and AI, and build robust, cross-industry data science solutions for major global enterprises. Our core product, Quantum ML, is a cloud-based data platform that uses Automated Machine Learning to help users predict markets, understand market drivers, and manage risk. With Quantum ML, users can easily deploy highly accurate ML models to predict any financial instrument – no data science or coding expertise required. We also offer consultancy and data services in addition to, or in augmentation of, our ML solutions, including data sales, data services, and custom software development.