Quantum computing pioneer D-Wave Systems today announced a new business unit – Quadrant – to provide machine learning services based on lessons from its quantum computing research. Quadrant will specialize in the use of generative learning models which require smaller sets of labelled data to generate models than typical discriminative methods. As a proof point of the approach’s power, D-Wave is calling attention to its winning effort in a recent Siemens medical imaging grand challenge – CATARACT – to automate identification of surgical instruments used in cataract surgery.
“D-Wave is committed to tackling real-world problems, today. Quadrant is a natural extension of the scientific and technological advances from D-Wave as we continue to explore new applications for our quantum systems,” said Vern Brownell, CEO at D-Wave in today’s announcement.
D-Wave was an early entrant into quantum computing field and its adiabatic annealing approach to quantum computing has drawn fans and critics. Nevertheless, it is the only vendor offering near-production ready machines currently. D-Wave’s approach is best used for problems that can be described as energy landscapes whose solution is finding the lowest energy state. This typically translates into optimization and minimization kinds of problems. For example, D-Wave and VW ran a test exercise to analyze taxi traffic in Beijing and select the shortest route for cabs based existing traffic conditions.
Application of AI techniques is a natural area of pursuit. Quadrant leverages generative machine learning, which requires less labeled data than common discriminative models. This can be very powerful: In the CATARACT challenge, for example, a relatively small dataset of 50 videos of cataract surgery was available and Quadrant was able to identify surgical tools used in cataract surgery with 99.71% accuracy.
Cataract surgery is the most common surgical procedure in the world with roughly 19 million cataract surgeries are performed annually. “The main motivation for annotating tool usage is to design efficient solutions for surgical workflow analysis. Analyzing the surgical workflow has potential applications in report generation, surgical training and even real-time decision support,” according to challenge organizers.
Generative learning algorithms such as those put forward by Quadrant are expected to have potential for use in AI across many disciplines and industries. Here’s a brief excerpt from Quadrant literature describing its approach.
“Deep neural networks are more than up to this task, and often have hundreds of thousands or millions of parameters. However, to train models of this richness without overfitting requires large amounts of training data. Before the extraordinary potential of machine learning can be realized across a range of domains, this situation must change, we must be able to learn with less data. Quadrant has developed a number of learning algorithms to do just that.
…As an example, imagine building a model that predicts a person’s weight (y) based on observations of their height (x). A generative model may cluster humans into gender, and then based on the gender, generate both height and weight. By inferring the gender, the weight prediction is based upon both height and gender. In contrast, a discriminative model which does not learn how to generate height and weight is comparatively impoverished; it predicts weight only as a function of height…”
Right now, Quadrant is running on GPUs because D-Wave wanted more companies to have access to the algorithms, but the models are “quantum ready” meaning they are equipped to run on D-Wave’s forthcoming next-gen quantum processor, according to a company spokesman.
Link to CATARACT study: https://cataracts.grand-challenge.org/home/