Abstract

The research goals of the project “Photonic Integrated Quantum machine Learning” (PIQLearn) lie within the quantum machine learning framework, namely the interplay between quantum mechanics and machine learning (ML).As such, one guideline of the project is to design the photonic version of widely studied standard ML algorithms, with a strong focus on integrated circuits. Another intriguing aspect that will be investigated is whether state-of-art quantum devices can grant a performance enhancement with respect to standard algorithms for real-world tasks. This question has become particularly significant, given that quantum advantage has only been demonstrated on tasks with no known applications. Given the relevance of the topic, several studies have been already conducted within this research area, but mostly relying on superconducting qubits, where the quantum circuit formalism has a straightforward translation. However, photonic platforms could constitute a game changer in this context because they are naturally fitting to adaptive protocols, e.g. ML training, where optimal parameters need to be iteratively searched. This suitability stems from their versatility, ease of reconfigurability and high fidelity in the performed operations. Another crucial aspect that encourages the use of photonics is the lower energy consumptions of optical artificial neural networks, which has been recently demonstrated. PIQLearn aims to take full advantage of these features, to boost the research on QML, in the European and Austrian research landscapes. This is why this project strives also for an increase in the know-how of the manufacturing of photonic integrated platforms in Austria. To achieve this result, our project will benefit from the synergy between the University of Vienna (UW) and Joanneum Research (JR), located in Graz (Austria). In this collaboration, UW provides the experimental expertise on the generation of single photon states (through spontaneous parametric down-conversion), on the exploitation of photonic integrated circuits (PICs), from a user’s point of view, and on the theory related to the design and simulation of ML algorithms. JR will complement this knowledge, by leveraging their experience on femtosecond laser micromachining, which is currently the most versatile technique to implement PICs. Hence, within PIQLearn, UW will devote its efforts to the design and implementation of novel QML algorithms oriented to real-world applications, while JR will work towards the realization of low-loss, low-power and reliable universal photonic processors, compatible with a broad wavelength spectrum. This interplay will allow to take a significant step forward in QML, exploring novel algorithms and exploiting innovative photonic platforms. The results of this project will pave the way to further investigations on photonic QML algorithms, which might prove a keystone within quantum computing, applied to real-world tasks. More in general, PIQLearn will bring remarkable insights on the potentialities of computational models with optical implementations, towards the design of future, more sustainable, computation algorithms.