Researchers have developed an extension of parameterized Instantaneous Quantum Polynomial (IQP) circuits, adapting them to work with integer data using the qudit formalism. Traditionally, quantum generative learning models based on IQP circuits have been effective for binary distributions. However, their application to non-binary datasets presented significant limitations, as converting integer values into qubit-compatible binary representations often distorted the original metric structure of the data.

The new methodology addresses this limitation by encoding each integer-valued pixel into a fixed-length bit-string. Quantum gates have been transformed to operate under the qudit formalism, allowing for a more natural and efficient representation of non-binary data. As part of this generative machine learning approach, a suitable loss function for circuit training has been designed, and a method for calculating the covariance matrix among features has been developed.

The validity of this method has been demonstrated using energy deposits from single-particle electron showers in the electromagnetic calorimeter of the CLIC detector. This advance is crucial because it enables parameterized IQP circuits, a promising tool in quantum machine learning, to effectively handle data that is not intrinsically binary. The ability to process integer data directly, without the information loss inherent in binary conversions, opens new avenues for quantum generative learning.

Beyond its application in particle physics, the proposed method is extensible to other areas that utilize quantum generative machine learning with non-binary data. This potentially includes fields such as image processing, signal analysis, or material simulation, where data often comes in integer or multi-valued formats. The development of this qudit extension for IQP circuits represents a significant step towards broadening the applicability of quantum machine learning models to more complex and diverse real-world problems.