Research News

Cleveland Clinic researchers are unlocking quantum computing’s full potential through the creation of a new computing paradigm inspired by the human brain. Fabio Cumbo, PhD, Research Associate in the lab of Daniel Blankenberg, PhD, Associate Staff, Computational Life Sciences, is developing the model, called quantum hyperdimensional computing (QHDC).  

Dr. Cumbo published the first-ever implementation of QHDC on two distinct experiments in Nature’s npj Unconventional Computing.  

What is hyperdimensional computing?  

Hyperdimensional computing (HDC) is a type of computing based in neuroscience. It follows the idea that a concept in the brain is not stored on one single neuron. For example, when you think of a cat, there is no single neuron in your brain solely responsible for knowing what a cat is. That information is spread across thousands or millions of neurons, so if one neuron fails, you still remember what a cat is.  

In computing, if data is spread across vectors, the system can produce an accurate calculation even if there is an error with one vector. To enable this, HDC mimics the brain by using long vectors, a type of data structure that contains thousands of dimensions.

QHDC is the application of this framework on quantum hardware. While HDC relies on massive high-dimensional representations of information, mapping these directly onto current quantum computers is a significant bottleneck.  

QHDC bridges this gap by leveraging quantum properties, like quantum superposition, to efficiently encode and process these complex spaces. HDC uses hyperdimensional vectors, while quantum computing gets its power from qubits that use the quantum principle of superposition to exist in multiple states at once. This makes QHDC ideal for biomedical research where data is complex and often has an unknowable number of possible outcomes. 

How does QHDC advance quantum computing frameworks?  

Dr. Blankenberg, senior author of the paper, says that even though the use of quantum computers is rapidly growing, researchers are still learning how to create frameworks and algorithms that can take full advantage of quantum computers’ potential. Current "quantum versions" of artificial intelligence (AI) and neural networks involve complicated workflows that can take a long time to develop and run.  

“Most quantum computing software is still built by borrowing ideas from classical computing,” Dr. Cumbo, lead author of the paper, says. “I had the idea to explore a type of computation that works naturally on a quantum computer, rather than forcing it to fit a classical framework.”  

Dr. Cumbo previously ran several studies on hyperdimensional computing, which helped him realize its compatibility with quantum computers.  

How did QHDC perform?  

The team tested the framework on a classical computer, an idealized quantum simulator and a quantum computer. These tests allowed the researchers to see not only how the framework performed, but also how it compared to current computing methods.  

The first test was a symbolic reasoning model to demonstrate the framework’s reasoning abilities. The second test was a machine learning test, which measured the framework’s ability to classify images and learn from the task.  

The results showed that QHDC was able to perform 500 times faster than other methods. 

“This work lays the foundation for a new class of quantum computing algorithms that can advance the speed and efficiency of biomedical research,” Dr. Cumbo says. “We will continue to explore the possibilities of QHDC by applying it to larger models and seeing if the speed and accuracy can maintain.”