Search

Modern machine learning relies nearly exclusively on dedicated electronic hardware accelerators. Photonic approaches, with low consumption and high operation speed, are increasingly considered for inference but, to date, remain mostly limited to relatively basic tasks. Simultaneously, the problem of training deep and complex neural networks, overwhelmingly performed through backpropagation, remains a significant limitation to the size and, consequently, the performance of current architectures and a major compute and energy bottleneck. Here, we experimentally implement a versatile and scalable training algorithm, called direct feedback alignment, on a hybrid electronic-photonic platform. An optical processing unit performs large-scale random matrix multiplications, which is the central operation of this algorithm, at speeds up to 1500 TeraOps. We perform optical training of one of the most recent deep learning architectures, including Transformers, with more than 1B parameters, and obtain good performances on both language and vision tasks. We study the compute scaling of our hybrid optical approach, and demonstrate a potential advantage for ultra-deep and wide neural networks, thus opening a promising route to sustain the exponential growth of modern artificial intelligence beyond traditional von Neumann approaches., Comment: 19 pages, 4 figures

Utilizing a quantum system for reservoir computing has recently received a lot of attention. Key challenges are related to how on can optimally en- and decode classical information, as well as what constitutes a good reservoir. Our main contribution is a generalization of the standard way to robustly en- and decode time series into subspaces defined by the cosets of a given stabilizer. A key observation is the necessity to perform the decoding step, which in turn ensures a consistent way of encoding. This provides a systematic way to encode classical information in a robust way. We provide a numerical analysis on a discrete time series given by two standard maps, namely the logistic and the H\'enon map. Our numerical findings indicate that the system's performance is increasing with the length of the training data.

Risk averse decision making under uncertainty in partially observable domains is a fundamental problem in AI and essential for reliable autonomous agents. In our case, the problem is modeled using partially observable Markov decision processes (POMDPs), when the value function is the conditional value at risk (CVaR) of the return. Calculating an optimal solution for POMDPs is computationally intractable in general. In this work we develop a simplification framework to speedup the evaluation of the value function, while providing performance guarantees. We consider as simplification a computationally cheaper belief-MDP transition model, that can correspond, e.g., to cheaper observation or transition models. Our contributions include general bounds for CVaR that allow bounding the CVaR of a random variable X, using a random variable Y, by assuming bounds between their cumulative distributions. We then derive bounds for the CVaR value function in a POMDP setting, and show how to bound the value function using the computationally cheaper belief-MDP transition model and without accessing the computationally expensive model in real-time. Then, we provide theoretical performance guarantees for the estimated bounds. Our results apply for a general simplification of a belief-MDP transition model and support simplification of both the observation and state transition models simultaneously.

Supervised models for speech enhancement are trained using artificially generated mixtures of clean speech and noise signals. However, the synthetic training conditions may not accurately reflect real-world conditions encountered during testing. This discrepancy can result in poor performance when the test domain significantly differs from the synthetic training domain. To tackle this issue, the UDASE task of the 7th CHiME challenge aimed to leverage real-world noisy speech recordings from the test domain for unsupervised domain adaptation of speech enhancement models. Specifically, this test domain corresponds to the CHiME-5 dataset, characterized by real multi-speaker and conversational speech recordings made in noisy and reverberant domestic environments, for which ground-truth clean speech signals are not available. In this paper, we present the objective and subjective evaluations of the systems that were submitted to the CHiME-7 UDASE task, and we provide an analysis of the results. This analysis reveals a limited correlation between subjective ratings and several supervised nonintrusive performance metrics recently proposed for speech enhancement. Conversely, the results suggest that more traditional intrusive objective metrics can be used for in-domain performance evaluation using the reverberant LibriCHiME-5 dataset developed for the challenge. The subjective evaluation indicates that all systems successfully reduced the background noise, but always at the expense of increased distortion. Out of the four speech enhancement methods evaluated subjectively, only one demonstrated an improvement in overall quality compared to the unprocessed noisy speech, highlighting the difficulty of the task. The tools and audio material created for the CHiME-7 UDASE task are shared with the community.

Supervised speech enhancement models are trained using artificially generated mixtures of clean speech and noise signals, which may not match real-world recording conditions at test time. This mismatch can lead to poor performance if the test domain significantly differs from the synthetic training domain. This paper introduces the unsupervised domain adaptation for conversational speech enhancement (UDASE) task of the 7th CHiME challenge. This task aims to leverage real-world noisy speech recordings from the target domain for unsupervised domain adaptation of speech enhancement models. The target domain corresponds to the multi-speaker reverberant conversational speech recordings of the CHiME-5 dataset, for which the ground-truth clean speech reference is unavailable. Given a CHiME-5 recording, the task is to estimate the clean, potentially multi-speaker, reverberant speech, removing the additive background noise. We discuss the motivation for the CHiME-7 UDASE task and describe the data, the task, and the baseline system.

We present a novel formalism to both understand and construct mixers that preserve a given subspace. The method connects and utilizes the stabilizer formalism that is used in error correcting codes. This can be useful in the setting when the quantum approximate optimization algorithm (QAOA), a popular meta-heuristic for solving combinatorial optimization problems, is applied in the setting where the constraints of the problem lead to a feasible subspace that is large but easy to specify. The proposed method gives a systematic way to construct mixers that are resource efficient in the number of controlled not gates and can be understood as a generalization of the well-known X and XY mixers and a relaxation of the Grover mixer: Given a basis of any subspace, a resource efficient mixer can be constructed that preserves the subspace. The numerical examples provided show a dramatic reduction of CX gates when compared to previous results. We call our approach logical X-Mixer or logical X QAOA ($\textbf{LX-QAOA}$), since it can be understood as dividing the subspace into code spaces of stabilizers S and consecutively applying logical rotational X gates associated with these code spaces. Overall, we hope that this new perspective can lead to further insight into the development of quantum algorithms.

Accurate segmentation of thalamic nuclei, crucial for understanding their role in healthy cognition and in pathologies, is challenging to achieve on standard T1-weighted (T1w) magnetic resonance imaging (MRI) due to poor image contrast. White-matter-nulled (WMn) MRI sequences improve intrathalamic contrast but are not part of clinical protocols or extant databases. Here, we introduce Histogram-based polynomial synthesis (HIPS), a fast preprocessing step that synthesizes WMn-like image contrast from standard T1w MRI using a polynomial approximation. HIPS was incorporated into our Thalamus Optimized Multi-Atlas Segmentation (THOMAS) pipeline, developed and optimized for WMn MRI. HIPS-THOMAS was compared to a convolutional neural network (CNN)-based segmentation method and THOMAS modified for T1w images (T1w-THOMAS). The robustness and accuracy of the three methods were tested across different image contrasts, scanner manufacturers, and field strength. HIPS-synthesized images improved intra-thalamic contrast and thalamic boundaries, and their segmentations yielded significantly better mean Dice, lower percentage of volume error, and lower standard deviations compared to both the CNN method and T1w-THOMAS. Finally, using THOMAS, HIPS-synthesized images were as effective as WMn images for identifying thalamic nuclei atrophy in alcohol use disorders subjects relative to healthy controls, with a higher area under the ROC curve compared to T1w-THOMAS (0.79 vs 0.73)., Comment: 5 Figures, 3 Tables, 1 Supplemental Figure, Submitted to Scientific Reports Apr 12 2023

This paper describes our submission to the Second Clarity Enhancement Challenge (CEC2), which consists of target speech enhancement for hearing-aid (HA) devices in noisy-reverberant environments with multiple interferers such as music and competing speakers. Our approach builds upon the powerful iterative neural/beamforming enhancement (iNeuBe) framework introduced in our recent work, and this paper extends it for target speaker extraction. We therefore name the proposed approach as iNeuBe-X, where the X stands for extraction. To address the challenges encountered in the CEC2 setting, we introduce four major novelties: (1) we extend the state-of-the-art TF-GridNet model, originally designed for monaural speaker separation, for multi-channel, causal speech enhancement, and large improvements are observed by replacing the TCNDenseNet used in iNeuBe with this new architecture; (2) we leverage a recent dual window size approach with future-frame prediction to ensure that iNueBe-X satisfies the 5 ms constraint on algorithmic latency required by CEC2; (3) we introduce a novel speaker-conditioning branch for TF-GridNet to achieve target speaker extraction; (4) we propose a fine-tuning step, where we compute an additional loss with respect to the target speaker signal compensated with the listener audiogram. Without using external data, on the official development set our best model reaches a hearing-aid speech perception index (HASPI) score of 0.942 and a scale-invariant signal-to-distortion ratio improvement (SI-SDRi) of 18.8 dB. These results are promising given the fact that the CEC2 data is extremely challenging (e.g., on the development set the mixture SI-SDR is -12.3 dB). A demo of our submitted system is available at WAVLab CEC2 demo.

Smart city management is going through a remarkable transition, in terms of quality and diversity of services provided to the end-users. The stakeholders that deliver pervasive applications are now able to address fundamental challenges in the big data value chain, from data acquisition, data analysis and processing, data storage and curation, and data visualisation in real scenarios. Industry 4.0 is pushing this trend forward, demanding for servitization of products and data, also for the smart cities sector where humans, sensors and devices are operating in strict collaboration. The data produced by the ubiquitous devices must be processed quickly to allow the implementation of reactive services such as situational awareness, video surveillance and geo-localization, while always ensuring the safety and privacy of involved citizens. This paper proposes a modular architecture to (i) leverage innovative technologies for data acquisition, management and distribution (such as Apache Kafka and Apache NiFi), (ii) develop a multi-layer engineering solution for revealing valuable and hidden societal knowledge in smart cities environment, and (iii) tackle the main issues in tasks involving complex data flows and provide general guidelines to solve them. We derived some guidelines from an experimental setting performed together with leading industrial technical departments to accomplish an efficient system for monitoring and servitization of smart city assets, with a scalable platform that confirms its usefulness in numerous smart city use cases with different needs., Comment: This work was funded by the European Union's Horizon 2020 research and innovation programme MARVEL under grant agreement No 957337. This publication reflects the authors views only. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible

Recent work on monaural source separation has shown that performance can be increased by using fully learned filterbanks with short windows. On the other hand it is widely known that, for conventional beamforming techniques, performance increases with long analysis windows. This applies also to most hybrid neural beamforming methods which rely on a deep neural network (DNN) to estimate the spatial covariance matrices. In this work we try to bridge the gap between these two worlds and explore fully end-to-end hybrid neural beamforming in which, instead of using the Short-Time-Fourier Transform, also the analysis and synthesis filterbanks are learnt jointly with the DNN. In detail, we explore two different types of learned filterbanks: fully learned and analytic. We perform a detailed analysis using the recent Clarity Challenge data and show that by using learnt filterbanks it is possible to surpass oracle-mask based beamforming for short windows., Comment: accepted at ICASSP 2022

The investigation of spatio-temporal couplings (STCs) of broadband light beams is becoming a key topic for the optimization as well as applications of ultrashort laser systems. This calls for accurate measurements of STCs. Yet, it is only recently that such complete spatio-temporal or spatio-spectral characterization has become possible, and it has so far mostly been implemented at the output of the laser systems, where experiments take place. In this survey, we present for the first time STC measurements at different stages of a collection of high-power ultrashort laser systems, all based on the chirped-pulse amplification (CPA) technique, but with very different output characteristics. This measurement campaign reveals spatio-temporal effects with various sources, and motivates the expanded use of STC characterization throughout CPA laser chains, as well as in a wider range of types of ultrafast laser systems. In this way knowledge will be gained not only about potential defects, but also about the fundamental dynamics and operating regimes of advanced ultrashort laser systems.