Extreme learning machines explore nonlinear random projections to perform computing tasks on high-dimensional output spaces. Since training only occurs at the output layer, the approach has the potential to speed up the training process and the capacity to turn any physical system into a computing platform. Yet, requiring strong nonlinear dynamics, optical solutions operating at fast processing rates and low power can be hard to achieve with conventional nonlinear optical materials. In this context, this manuscript explores the possibility of using atomic gases in near-resonant conditions to implement an optical extreme learning machine leveraging their enhanced nonlinear optical properties. Our results suggest that these systems have the potential not only to work as an optical extreme learning machine but also to perform these computations at the few-photon level, paving opportunities for energy-efficient computing solutions.
Extreme learning machines explore nonlinear random projections to perform computing tasks on high-dimensional output spaces. Since training only occurs at the output layer, the approach has the potential to speed up the training process and the capacity to turn any physical system into a computing platform. Yet, requiring strong nonlinear dynamics, optical solutions operating at fast processing rates and low power can be hard to achieve with conventional nonlinear optical materials. In this context, this manuscript explores the possibility of using atomic gases in near-resonant conditions to implement an optical extreme learning machine leveraging their enhanced nonlinear optical properties. Our results suggest that these systems have the potential not only to work as an optical extreme learning machine but also to perform these computations at the few-photon level, paving opportunities for energy-efficient computing solutions.
Recent advancements in machine learning have led to an exponential increase in computational demands, driving the need for innovative computing platforms. Quantum computing, with its Hilbert space scaling exponentially with the number of particles, emerges as a promising solution. In this work, we implement a quantum extreme machine learning (QELM) protocol leveraging indistinguishable photon pairs and multimode fiber as a random densly connected layer. We experimentally study QELM performance based on photon coincidences -- for distinguishable and indistinguishable photons -- on an image classification task. Simulations further show that increasing the number of photons reveals a clear quantum advantage. We relate this improved performance to the enhanced dimensionality and expressivity of the feature space, as indicated by the increased rank of the feature matrix in both experiment and simulation.
Abstract Quantum reservoir computing and quantum extreme learning machines are two emerging approaches that have demonstrated their potential both in classical and quantum machine learning tasks. They exploit the quantumness of physical systems combined with an easy training strategy, achieving an excellent performance. The increasing interest in these unconventional computing approaches is fueled by the availability of diverse quantum platforms suitable for implementation and the theoretical progresses in the study of complex quantum systems. In this review article, recent proposals and first experiments displaying a broad range of possibilities are reviewed when quantum inputs, quantum physical substrates and quantum tasks are considered. The main focus is the performance of these approaches, on the advantages with respect to classical counterparts and opportunities.
We train a model atom to recognize hand-written digits between 0 and 9, employing intense light--matter interaction as a computational resource. For training, individual images of hand-written digits in the range 0-9 are converted into shaped laser pulses (data input pulses). Simultaneously with an input pulse, another shaped pulse (program pulse), polarized in the orthogonal direction, is applied to the atom and the system evolves quantum mechanically according to the time-dependent Schrödinger equation. The purpose of the optimal program pulse is to direct the system into specific atomic final states that correspond to the input digits. A success rate of about 40\% is demonstrated here for a basic optimization scheme, so far limited by the computational power to find the optimal program pulse in a high-dimensional search space. This atomic-intelligence image-recognition scheme is scalable towards larger (e.g. molecular) systems, is readily reprogrammable towards other learning/classification tasks and operates on time scales down to tens of femtoseconds. It has the potential to outpace other currently implemented machine-learning approaches, including the fastest optical on-chip neuromorphic systems and optical accelerators, by orders of magnitude.
[eng] Quantum photonic technologies have achieved considerable success across a range of disciplines, including quantum computing, quantum communications, and quantum metrology. Conversely, in the current era of information, the ability to discern patterns within voluminous data sets is highly sought after, with Machine Learning (ML) having emerged as a powerful solution to address this challenge. The processing speed of photonic hardware, in conjunction with the potential for computation enhancement resulting from quantum properties, positions quantum photonics as a promising platform for the implementation of ML algorithms. In this thesis, we focus on reservoir computing (RC) as an ML framework that is suitable for time series processing. It provides fast and energy-efficient training in comparison to more conventional methods and is also appropriate for physical implementations. In this thesis, we explore the possibilities brought by quantum photonic setups in the field of quantum RC. Specifically, we propose and study different platform designs that are able to perform different temporal benchmark tasks by providing tunable memory and nonlinearity. Furthermore, we leverage the capabilities of state-of-the-art quantum photonic hardware and identify novel strategies to address key challenges in RC in quantum substrates, as well as expand the potential of quantum machine learning. Firstly, an approach is provided to overcome some of the limitations of quantum projective measurements, thereby enabling information processing in real time. This is achieved by generating a physical ensemble of quantum reservoirs that recirculate through an optical fiber inside a feedback loop. We investigate the detrimental effects of sampling noise due to finite ensembles and provide a strategy to mitigate them. Secondly, the consequences of increasing quantum squeezing in the dynamics of such setups were studied, and it was found that it monotonically improves the noise robustness of the platforms in realistic scenarios. Finally, we propose a novel design in which photonic cluster states, a specific kind of highly entangled quantum network, are used as resources to implement quantum RC. In this configuration, external input information is encoded into the quantum network via quantum teleportation by engineering the measurement basis of the protocol. Furthermore, the platform has the capacity to implement global quantum gates through local operations (single-mode measurements), thus establishing the foundation for distributed quantum ML.
Quantum reservoir computing has emerged as a promising paradigm within the field of quantum machine learning, harnessing the inherent properties of quantum systems to optimise and enhance information processing capabilities. Here, we explore the potential of quantum-inspired machine learning methodologies by leveraging the complex dynamics of quantum reservoirs to address computationally challenging tasks with enhanced efficiency and accuracy. To this end, we employ an open quantum system model comprising two-level atomic ensembles coupled to Lorentzian photonic cavities to construct a quantum physical reservoir computer layer for a recurrent neural network. We evaluate the effectiveness of this approach by applying it to a standard machine learning image-recognition problem and benchmarking its performance against a conventional neural network of similar architecture, but lacking the quantum physical reservoir computer layer. Remarkably, as the dataset size increases, the quantum physical reservoir computer outperforms the conventional neural network, requiring fewer training epochs and a smaller dataset to achieve comparable accuracy. Furthermore, we employ the quantum physical reservoir computing approach to model the dynamics of open quantum systems, focusing on atomic system ensembles interacting with a structured photonic reservoir associated with a photonic band-gap material. Our results reveal that the quantum reservoir computer provides equally powerful representations for quantum dynamical problems, maintaining effectiveness even under constraints of limited training data.
We train a model atom to recognize pixel-drawn digits based on hand-written numbers in the range 0-9, employing intense light-matter interaction as a computational resource. For training, the images of the digits are converted into shaped laser pulses (data input pulses). Simultaneously with an input pulse, another shaped pulse (program pulse), polarized in the orthogonal direction, is applied to the atom and the system evolves quantum mechanically according to the time-dependent Schrödinger equation. The purpose of the optimal program pulse is to direct the system into specific atomic final states (classification states) that correspond to the input digits. A success rate of about 40% is achieved when using a basic optimization scheme that might be limited by the computational resources for finding the optimal program pulse in a high-dimensional search space. Our key result is the demonstration that the laser-programmed atom is able to generalize, i.e. successful classification is not limited to the training examples, but also the classification of previously unseen images is improved by training. This atom-sized machine-learning image-recognition scheme operates on time scales down to tens of femtoseconds, is scalable towards larger (e.g. molecular) systems, and is readily reprogrammable towards other learning/classification tasks. An experimental implementation of the scheme using ultrafast polarization pulse shaping and differential photoelectron detection is within reach.