According to LivesCience, Australian researchers have developed new quantum machine learning technologies to drive and generate new semiconductor designs that will help improve the chip design process. This paper, published in Advanced Science, demonstrates how to generate new models that can improve chip design efficiency by encoding data in quantum states to search for patterns before analyzing results using machine learning.
As the paper suggests, modern design processes for complex processors and the semiconductors within them are complex and require absolute accuracy. There are many steps in the process of laying the latest wafers and silicon layers that will ultimately create chips. This new technique may be most useful in the final part of this process.
Once chips are wrapped in packages and can be integrated into real-world devices, it is important to have a deep understanding of how the manufacturing process allows the semiconductor and metal packaging layers of the chip to be known as the electrical flow between them, otherwise ohm contact resistance. Although it is particularly challenging modeling, researchers believe their new techniques can make it much easier, allowing for potential advancements in modern chip designs.
The report used 159 samples of Gallium Nitride High Electron Mobility Transistors (GAN HEMTS) commonly used in high-end electronics. They first developed a technique called Quantum Kernel-Aligned Regressor (QKAR), in which variables in the manufacturing process have the greatest effect on the contact resistance of ohms, and then convert classical data into quantum states. Quantum computing systems can analyze that data and search for patterns. The results of that analysis are then fed into machine learning algorithms, where the data can be analyzed and applied to the chip design process to find greater efficiencies that may be found in manufacturing.
This model, which combines quantum and machine learning elements, is said to outweigh more traditional machine learning and deep learning algorithms. This study suggests that QKAR is an effective method between 8.8% and 20.1% over the other models.
This could allow for much more subtle chip design processes in the future, but it may require the production of new, more sophisticated quantum computing hardware before exploiting it for full effectiveness.
“These findings illustrate the potential of QML to effectively handle high-dimensional small regression tasks in the semiconductor domain, and point to promising tools for future deployment in real-world applications as quantum hardware continues to mature,” the study states.
While the technologies outlined in this study may not be ready to revolutionize chipmaking, the combination of machine learning and quantum computing technology highlights the impact of quantum computing across a wide range of industries, even if large quantum computing hardware is still viable. Both traditional and quantum computing have their own advantages, so combining techniques allows best-world scenarios to be delivered in a wide range of potential applications.
to follow Google News Tom Hardware Get the latest news, analysis and reviews in your feed. surely[フォロー]Please click the button.
