Search

Recommender systems have become integral to digital experiences, shaping user interactions and preferences across various platforms. Despite their widespread use, these systems often suffer from algorithmic biases that can lead to unfair and unsatisfactory user experiences. This study introduces an interactive tool designed to help users comprehend and explore the impacts of algorithmic harms in recommender systems. By leveraging visualizations, counterfactual explanations, and interactive modules, the tool allows users to investigate how biases such as miscalibration, stereotypes, and filter bubbles affect their recommendations. Informed by in-depth user interviews, this tool benefits both general users and researchers by increasing transparency and offering personalized impact assessments, ultimately fostering a better understanding of algorithmic biases and contributing to more equitable recommendation outcomes. This work provides valuable insights for future research and practical applications in mitigating bias and enhancing fairness in machine learning algorithms.

We study the integration problem over the $s$-dimensional unit cube on four types of Banach spaces of integrands. First we consider Haar wavelet spaces, consisting of functions whose Haar wavelet coefficients exhibit a certain decay behavior measured by a parameter $\alpha >0$. We study the worst case error of integration over the norm unit ball and provide upper error bounds for quasi-Monte Carlo (QMC) cubature rules based on arbitrary $(t,m,s)$-nets as well as matching lower error bounds for arbitrary cubature rules. These results show that using arbitrary $(t,m,s)$-nets as sample points yields the best possible rate of convergence. Afterwards we study spaces of integrands of fractional smoothness $\alpha \in (0,1)$ and state a sharp Koksma-Hlawka-type inequality. More precisely, we show that on those spaces the worst case error of integration is equal to the corresponding fractional discrepancy. Those spaces can be continuously embedded into tensor product Bessel potential spaces, also known as Sobolev spaces of dominated mixed smoothness, with the same set of parameters. The latter spaces can be embedded into suitable Besov spaces of dominating mixed smoothness $\alpha$, which in turn can be embedded into the Haar wavelet spaces with the same set of parameters. Therefore our upper error bounds on Haar wavelet spaces for QMC cubatures based on $(t,m,s)$-nets transfer (with possibly different constants) to the corresponding spaces of integrands of fractional smoothness and to Sobolev and Besov spaces of dominating mixed smoothness. Moreover, known lower error bounds for periodic Sobolev and Besov spaces of dominating mixed smoothness show that QMC integration based on arbitrary $(t,m,s)$-nets yields the best possible convergence rate on periodic as well as on non-periodic Sobolev and Besov spaces of dominating smoothness., Comment: 56 pages

Video generation using diffusion-based models is constrained by high computational costs due to the frame-wise iterative diffusion process. This work presents a Diffusion Reuse MOtion (Dr. Mo) network to accelerate latent video generation. Our key discovery is that coarse-grained noises in earlier denoising steps have demonstrated high motion consistency across consecutive video frames. Following this observation, Dr. Mo propagates those coarse-grained noises onto the next frame by incorporating carefully designed, lightweight inter-frame motions, eliminating massive computational redundancy in frame-wise diffusion models. The more sensitive and fine-grained noises are still acquired via later denoising steps, which can be essential to retain visual qualities. As such, deciding which intermediate steps should switch from motion-based propagations to denoising can be a crucial problem and a key tradeoff between efficiency and quality. Dr. Mo employs a meta-network named Denoising Step Selector (DSS) to dynamically determine desirable intermediate steps across video frames. Extensive evaluations on video generation and editing tasks have shown that Dr. Mo can substantially accelerate diffusion models in video tasks with improved visual qualities.

Audio Event Recognition (AER) traditionally focuses on detecting and identifying audio events. Most existing AER models tend to detect all potential events without considering their varying significance across different contexts. This makes the AER results detected by existing models often have a large discrepancy with human auditory perception. Although this is a critical and significant issue, it has not been extensively studied by the Detection and Classification of Sound Scenes and Events (DCASE) community because solving it is time-consuming and labour-intensive. To address this issue, this paper introduces the concept of semantic importance in AER, focusing on exploring the differences between human perception and model inference. This paper constructs a Multi-Annotated Foreground Audio Event Recognition (MAFAR) dataset, which comprises audio recordings labelled by 10 professional annotators. Through labelling frequency and variance, the MAFAR dataset facilitates the quantification of semantic importance and analysis of human perception. By comparing human annotations with the predictions of ensemble pre-trained models, this paper uncovers a significant gap between human perception and model inference in both semantic identification and existence detection of audio events. Experimental results reveal that human perception tends to ignore subtle or trivial events in the event semantic identification, while model inference is easily affected by events with noises. Meanwhile, in event existence detection, models are usually more sensitive than humans., Comment: Dataset homepage: https://github.com/Voltmeter00/MAFAR

Both freeze-in of very weakly coupled dark matter and freeze-out of initially thermalized dark matter from the primordial heat bath provide interesting possibilities for dark matter creation in the early universe. Both scenarios allow for a calculation of baryon-dark matter coupling constants as a function of dark matter mass due to the constraint that freeze-in or freeze-out produce the observed dark matter abundance. Here we compare the resulting coupling constants in the two scenarios if dark matter couples to baryons through an antisymmetric tensor portal. The freeze-in scenario predicts much smaller coupling in agreement with the nonthermalization postulate. We find that the couplings as a function of mass behave very differently in the two scenarios., Comment: 6 pages, 6 figures

In this work, we explore a time-fractional diffusion equation of order $\alpha \in (0,1)$ with a stochastic diffusivity parameter. We focus on efficient estimation of the expected values (considered as an infinite dimensional integral on the parametric space corresponding to the random coefficients) of linear functionals acting on the solution of our model problem. To estimate the expected value computationally, the infinite expansions of the random parameter need to be truncated. Then we approximate the high-dimensional integral over the random field using a high-order quasi-Monte Carlo method. This follows by approximating the deterministic solution over the space-time domain via a second-order accurate time-stepping scheme in combination with a spatial discretization by Galerkin finite elements. Under reasonable regularity assumptions on the given data, we show some regularity properties of the continuous solution and investigate the errors from estimating the expected value. We report on numerical experiments that complement the theoretical results.

We derive new small deviations bounds for a class of Gaussian multiplicative chaos measures obtained from Gaussian fields with zero spatial average. The upper bound holds for a class of fields closely related to the $\star$-scale invariant Gaussian fields in arbitrary number of dimensions, and the lower bound holds for the Gaussian free field with zero spatial average on the torus. The upper bound is obtained by a modification of the method that was used in \cite{LRV}, and the lower bound is obtained by applying the Donsker--Varadhan variational formula. We also give the probabilistic path integral formulation of the massless Sinh--Gordon model on a torus of side length $R$, and study its partition function $R$ tends to infinity. We apply the small deviation bounds for Gaussian multiplicative chaos to obtain lower and upper bounds for the logarithm of the partition function, leading to the existence of a non-zero and finite subsequential infinite volume limit for the free energy., Comment: Preliminary version

Increase in data, size, or compute can lead to sudden learning of specific capabilities by a neural network -- a phenomenon often called "emergence''. Beyond scientific understanding, establishing the causal factors underlying such emergent capabilities is crucial to enable risk regulation frameworks for AI. In this work, we seek inspiration from study of emergent properties in other fields and propose a phenomenological definition for the concept in the context of neural networks. Our definition implicates the acquisition of general structures underlying the data-generating process as a cause of sudden performance growth for specific, narrower tasks. We empirically investigate this definition by proposing an experimental system grounded in a context-sensitive formal language and find that Transformers trained to perform tasks on top of strings from this language indeed exhibit emergent capabilities. Specifically, we show that once the language's underlying grammar and context-sensitivity inducing structures are learned by the model, performance on narrower tasks suddenly begins to improve. We then analogize our network's learning dynamics with the process of percolation on a bipartite graph, establishing a formal phase transition model that predicts the shift in the point of emergence observed in our experiments when changing the data structure. Overall, our experimental and theoretical frameworks yield a step towards better defining, characterizing, and predicting emergence in neural networks., Comment: Preprint

The rise of generative artificial intelligence (GenAI) is transforming the telecom industry. GenAI models, particularly large language models (LLMs), have emerged as powerful tools capable of driving innovation, improving efficiency, and delivering superior customer services in telecom. This paper provides an overview of GenAI for telecom from theory to practice. We review GenAI models and discuss their practical applications in telecom. Furthermore, we describe the key technology enablers and best practices for applying GenAI to telecom effectively. We highlight the importance of retrieval augmented generation (RAG) in connecting LLMs to telecom domain specific data sources to enhance the accuracy of the LLMs' responses. We present a real-world use case on RAG-based chatbot that can answer open radio access network (O-RAN) specific questions. The demonstration of the chatbot to the O-RAN Alliance has triggered immense interest in the industry. We have made the O-RAN RAG chatbot publicly accessible on GitHub., Comment: 7 pages, 6 figures, submitted for possible publication

Due to the divergence-instability, the accuracy of low-order conforming finite element methods (FEMs) for nearly incompressible elasticity equations deteriorates as the Lam\'e parameter $\lambda\to\infty$, or equivalently as the Poisson ratio $\nu\to1/2$. This effect is known as \itshape{locking} or \itshape{non-robustness}. For the piecewise linear case, the error in the ${\bf L}^2$-norm of the standard Galerkin conforming FEM is bounded by $C\lambda h^2$, resulting in poor accuracy for practical values of $h$ if $\lambda$ is sufficiently large. In this paper, we show that for 2D problems the locking phenomenon can be controlled by replacing $\lambda$ with $\lambda_h=\lambda/(1+\lambda h/L)$ in the stiffness matrix, where $L$ is the diameter of the body $\Omega$. We prove that with this modification, the error in the ${\bf L}^2$-norm is bounded by $Ch$ for a constant $C$ that does not depend on $\lambda$. Numerical experiments confirm this convergence behaviour and show that, for practical meshes, our method is more accurate than the standard method if the material is nearly incompressible. Our analysis also shows that the error in the ${\bf H}^1$-norm is bounded by $Ch^{1/2}$, but our numerical experiments suggest that this bound is not sharp.

In exploring the simulation of human rhythmic perception and synchronization capabilities, this study introduces a computational model inspired by the physical and biological processes underlying rhythm processing. Utilizing a reservoir computing framework that simulates the function of cerebellum, the model features a dual-neuron classification and incorporates parameters to modulate information transfer, reflecting biological neural network characteristics. Our findings demonstrate the model's ability to accurately perceive and adapt to rhythmic patterns within the human perceptible range, exhibiting behavior closely aligned with human rhythm interaction. By incorporating fine-tuning mechanisms and delay-feedback, the model enables continuous learning and precise rhythm prediction. The introduction of customized settings further enhances its capacity to stimulate diverse human rhythmic behaviors, underscoring the potential of this architecture in temporal cognitive task modeling and the study of rhythm synchronization and prediction in artificial and biological systems. Therefore, our model is capable of transparently modelling cognitive theories that elucidate the dynamic processes by which the brain generates rhythm-related behavior.

As Fifth generation (5G) cellular systems transition to softwarized, programmable, and intelligent networks, it becomes fundamental to enable public and private 5G deployments that are (i) primarily based on software components while (ii) maintaining or exceeding the performance of traditional monolithic systems and (iii) enabling programmability through bespoke configurations and optimized deployments. This requires hardware acceleration to scale the Physical (PHY) layer performance, programmable elements in the Radio Access Network (RAN) and intelligent controllers at the edge, careful planning of the Radio Frequency (RF) environment, as well as end-to-end integration and testing. In this paper, we describe how we developed the programmable X5G testbed, addressing these challenges through the deployment of the first 8-node network based on the integration of NVIDIA Aerial RAN CoLab (ARC), OpenAirInterface (OAI), and a near-real-time RAN Intelligent Controller (RIC). The Aerial Software Development Kit (SDK) provides the PHY layer, accelerated on Graphics Processing Unit (GPU), with the higher layers from the OAI open-source project interfaced with the PHY through the Small Cell Forum (SCF) Functional Application Platform Interface (FAPI). An E2 agent provides connectivity to the O-RAN Software Community (OSC) near-real-time RIC. We discuss software integration, the network infrastructure, and a digital twin framework for RF planning. We then profile the performance with up to 4 Commercial Off-the-Shelf (COTS) smartphones for each base station with iPerf and video streaming applications, measuring a cell rate higher than 500 Mbps in downlink and 45 Mbps in uplink., Comment: 15 pages, 15 figures, 3 tables. arXiv admin note: text overlap with arXiv:2310.17062

We investigate experimentally the quantum coherence of an electronic two-level system in a double quantum dot under continuous charge detection. The charge-state of the two-level system is monitored by a capacitively coupled single quantum dot detector that imposes a back-action effect to the system. The measured back-action is well described by an additional decoherence rate, approximately linearly proportional to the detector electron tunneling rate. We provide a model for the decoherence rate arising due to level detuning fluctuations induced by detector charge fluctuations. The theory predicts a factor of two lower decoherence rate than observed in the experiment, suggesting the need for a more elaborate theory accounting for additional sources of decoherence., Comment: 5 pages, 3 figures

The measurement problem of quantum mechanics concerns the question under which circumstances coherent wave evolution becomes disrupted to produce eigenstates of observables, instead of evolving superpositions of eigenstates. The problem needs to be addressed already within wave mechanics, before second quantization, because low-energy interactions can be dominated by particle-preserving potential interactions. We discuss a scattering array of harmonic oscillators which can detect particles penetrating the array through interaction with a short-range potential. Evolution of the wave function of scattered particles, combined with Heisenberg's assertion that quantum jumps persist in wave mechanics, indicates that the wave function will collapse around single oscillator sites if the scattering is inelastic, while it will not collapse around single sites for elastic scattering. The Born rule for position observation is then equivalent to the statement that the wave function for inelastic scattering amounts to an epistemic superposition of possible scattering states, in the sense that it describes a sum of probability amplitudes for inelastic scattering off different scattering centers, whereas at most one inelastic scattering event can happen at any moment in time. Within this epistemic interpretation of the wave function, the actual underlying inelastic scattering event corresponds to a quantum jump, whereas the continuously evolving wave function only describes the continuous evolution of probability amplitudes for scattering off different sites. Quantum jumps then yield definite position observations as defined by the spatial resolution of the oscillator array., Comment: 5 pages

We live in a rich and varied acoustic world, which is experienced by individuals or communities as a soundscape. Computational auditory scene analysis, disentangling acoustic scenes by detecting and classifying events, focuses on objective attributes of sounds, such as their category and temporal characteristics, ignoring the effect of sounds on people and failing to explore the relationship between sounds and the emotions they evoke within a context. To fill this gap and to automate soundscape analysis, which traditionally relies on labour-intensive subjective ratings and surveys, we propose the soundscape captioning (SoundSCap) task. SoundSCap generates context-aware soundscape descriptions by capturing the acoustic scene, event information, and the corresponding human affective qualities. To this end, we propose an automatic soundscape captioner (SoundSCaper) composed of an acoustic model, SoundAQnet, and a general large language model (LLM). SoundAQnet simultaneously models multi-scale information about acoustic scenes, events, and perceived affective qualities, while LLM generates soundscape captions by parsing the information captured by SoundAQnet to a common language. The soundscape caption's quality is assessed by a jury of 16 audio/soundscape experts. The average score (out of 5) of SoundSCaper-generated captions is lower than the score of captions generated by two soundscape experts by 0.21 and 0.25, respectively, on the evaluation set and the model-unknown mixed external dataset with varying lengths and acoustic properties, but the differences are not statistically significant. Overall, SoundSCaper-generated captions show promising performance compared to captions annotated by soundscape experts. The models' code, LLM scripts, human assessment data and instructions, and expert evaluation statistics are all publicly available., Comment: Code: https://github.com/Yuanbo2020/SoundSCaper

We present a superconducting cavity-coupled double quantum dot (DQD) photodiode that achieves a maximum photon-to-electron conversion efficiency of 25% in the microwave domain. With a higher-quality-factor cavity and improved device design to prevent photon leakages through unwanted pathways, our device measures microwave signals down to 100 aW power level and achieves sensitivity to probe microwave signals with one photon at a time in the cavity. We analyze the photodiode operation using Jaynes-Cummings input-output theory, identifying the key improvements of stronger cavity-DQD coupling needed to achieve near-unity photodetection efficiency. The results presented in this work represent a crucial advancement toward near unity microwave photodetection efficiency with single cavity-photon sensitivity for studies of photon statistics in the microwave range and applications related to quantum information processing., Comment: 8 pages, 4 figures

We propose reconstruction advantage measures to audit label privatization mechanisms. A reconstruction advantage measure quantifies the increase in an attacker's ability to infer the true label of an unlabeled example when provided with a private version of the labels in a dataset (e.g., aggregate of labels from different users or noisy labels output by randomized response), compared to an attacker that only observes the feature vectors, but may have prior knowledge of the correlation between features and labels. We consider two such auditing measures: one additive, and one multiplicative. These incorporate previous approaches taken in the literature on empirical auditing and differential privacy. The measures allow us to place a variety of proposed privatization schemes -- some differentially private, some not -- on the same footing. We analyze these measures theoretically under a distributional model which encapsulates reasonable adversarial settings. We also quantify their behavior empirically on real and simulated prediction tasks. Across a range of experimental settings, we find that differentially private schemes dominate or match the privacy-utility tradeoff of more heuristic approaches.

Let $\alpha$ be a Steinhaus or a Rademacher random multiplicative function. For a wide class of multiplicative functions $f$ we show that the sum $\sum_{n \le x}\alpha(n) f(n)$, normalised to have mean square $1$, has a non-Gaussian limiting distribution. More precisely, we establish a generalised central limit theorem with random variance determined by the total mass of a random measure associated with $\alpha f$. Our result applies to $d_z$, the $z$-th divisor function, as long as $z$ is strictly between $0$ and $\tfrac{1}{\sqrt{2}}$. Other examples of admissible $f$-s include any multiplicative indicator function with the property that $f(p)=1$ holds for a set of primes of density strictly between $0$ and $\tfrac{1}{2}$., Comment: 42 pages, 2 figures. Typos fixed, abstract and discussion of previous works updated. Codes for simulation experiment available on authors' personal page; comments are still welcome

Let $\alpha \colon \mathbb{N} \to S^1$ be the Steinhaus multiplicative function: a completely multiplicative function such that $(\alpha(p))_{p\text{ prime}}$ are i.i.d.~random variables uniformly distributed on $S^1$. Helson conjectured that $\mathbb{E}|\sum_{n\le x}\alpha(n)|=o(\sqrt{x})$ as $x \to \infty$, and this was solved in strong form by Harper. We give a short proof of the conjecture using a result of Saksman and Webb on a random model for the zeta function., Comment: 7 pages. Typos fixed and discussion of previous works added; comments are welcome

Context detection involves labeling segments of an online stream of data as belonging to different tasks. Task labels are used in lifelong learning algorithms to perform consolidation or other procedures that prevent catastrophic forgetting. Inferring task labels from online experiences remains a challenging problem. Most approaches assume finite and low-dimension observation spaces or a preliminary training phase during which task labels are learned. Moreover, changes in the transition or reward functions can be detected only in combination with a policy, and therefore are more difficult to detect than changes in the input distribution. This paper presents an approach to learning both policies and labels in an online deep reinforcement learning setting. The key idea is to use distance metrics, obtained via optimal transport methods, i.e., Wasserstein distance, on suitable latent action-reward spaces to measure distances between sets of data points from past and current streams. Such distances can then be used for statistical tests based on an adapted Kolmogorov-Smirnov calculation to assign labels to sequences of experiences. A rollback procedure is introduced to learn multiple policies by ensuring that only the appropriate data is used to train the corresponding policy. The combination of task detection and policy deployment allows for the optimization of lifelong reinforcement learning agents without an oracle that provides task labels. The approach is tested using two benchmarks and the results show promising performance when compared with related context detection algorithms. The results suggest that optimal transport statistical methods provide an explainable and justifiable procedure for online context detection and reward optimization in lifelong reinforcement learning., Comment: 10 pages excluding references and bibliography. Accepted at CoLLAs 2024

To efficiently utilize the scarce wireless resource, the random access scheme has been attaining renewed interest primarily in supporting the sporadic traffic of a large number of devices encountered in the Internet of Things (IoT). In this paper we investigate the performance of slotted ALOHA -- a simple and practical random access scheme -- in connection with the grant-free random access protocol applied for user-centric cell-free massive MIMO. More specifically, we provide the expression of the sum-throughput under the assumptions of the capture capability owned by the centralized detector in the uplink. Further, a comparative study of user-centric cell-free massive MIMO with other types of networks is provided, which allows us to identify its potential and possible limitation. Our numerical simulations show that the user-centric cell-free massive MIMO has a good trade-off between performance and fronthaul load, especially at low activation probability regime., Comment: Accepted for presentation at IEEE PIMRC 2024, Valencia, Spain. Copyright may be transferred without notice, after which this version may no longer be accessible

The concept of counterfactual explanations (CE) has emerged as one of the important concepts to understand the inner workings of complex AI systems. In this paper, we translate the idea of CEs to linear optimization and propose, motivate, and analyze three different types of CEs: strong, weak, and relative. While deriving strong and weak CEs appears to be computationally intractable, we show that calculating relative CEs can be done efficiently. By detecting and exploiting the hidden convex structure of the optimization problem that arises in the latter case, we show that obtaining relative CEs can be done in the same magnitude of time as solving the original linear optimization problem. This is confirmed by an extensive numerical experiment study on the NETLIB library.

While diffusion models effectively generate remarkable synthetic images, a key limitation is the inference inefficiency, requiring numerous sampling steps. To accelerate inference and maintain high-quality synthesis, teacher-student distillation is applied to compress the diffusion models in a progressive and binary manner by retraining, e.g., reducing the 1024-step model to a 128-step model in 3 folds. In this paper, we propose a single-fold distillation algorithm, SFDDM, which can flexibly compress the teacher diffusion model into a student model of any desired step, based on reparameterization of the intermediate inputs from the teacher model. To train the student diffusion, we minimize not only the output distance but also the distribution of the hidden variables between the teacher and student model. Extensive experiments on four datasets demonstrate that our student model trained by the proposed SFDDM is able to sample high-quality data with steps reduced to as little as approximately 1%, thus, trading off inference time. Our remarkable performance highlights that SFDDM effectively transfers knowledge in single-fold distillation, achieving semantic consistency and meaningful image interpolation.

This paper introduces a novel approach to predicting periodic time series using reservoir computing. The model is tailored to deliver precise forecasts of rhythms, a crucial aspect for tasks such as generating musical rhythm. Leveraging reservoir computing, our proposed method is ultimately oriented towards predicting human perception of rhythm. Our network accurately predicts rhythmic signals within the human frequency perception range. The model architecture incorporates primary and intermediate neurons tasked with capturing and transmitting rhythmic information. Two parameter matrices, denoted as c and k, regulate the reservoir's overall dynamics. We propose a loss function to adapt c post-training and introduce a dynamic selection (DS) mechanism that adjusts $k$ to focus on areas with outstanding contributions. Experimental results on a diverse test set showcase accurate predictions, further improved through real-time tuning of the reservoir via c and k. Comparative assessments highlight its superior performance compared to conventional models.

Coherent states are a well-established tool of quantum optics to describe electromagnetic waves in terms of photons. However, they do not describe the near-field regime of radiation sources. Instead, we generically use classical solutions of Maxwell's equations to describe radiation in the near-field regime. The classical solutions provide linear relations between currents and emitted electromagnetic fields, whereas evolution of states at the quantum level proceeds through unitary time evolution operators involving photon operators. This begs questions how the classical radiation equations relate to unitary quantum evolution, and how we can describe macroscopic fields from antennas or magnetic coils in terms of elementary photons. The present paper answers both questions through the construction of generalized Glauber states for radiation emitters., Comment: 9 pages. Version 2 introduces a phase factor for the photon states in Sec. 2, Eqs. (30,31), such that the photon states do not only yield the correct expectation values for the electromagnetic fields, but also satisfy the interaction picture evolution equation for general time-dependent source currents

Spoken language interaction is at the heart of interpersonal communication, and people flexibly adapt their speech to different individuals and environments. It is surprising that robots, and by extension other digital devices, are not equipped to adapt their speech and instead rely on fixed speech parameters, which often hinder comprehension by the user. We conducted a speech comprehension study involving 39 participants who were exposed to different environmental and contextual conditions. During the experiment, the robot articulated words using different vocal parameters, and the participants were tasked with both recognising the spoken words and rating their subjective impression of the robot's speech. The experiment's primary outcome shows that spaces with good acoustic quality positively correlate with intelligibility and user experience. However, increasing the distance between the user and the robot exacerbated the user experience, while distracting background sounds significantly reduced speech recognition accuracy and user satisfaction. We next built an adaptive voice for the robot. For this, the robot needs to know how difficult it is for a user to understand spoken language in a particular setting. We present a prediction model that rates how annoying the ambient acoustic environment is and, consequentially, how hard it is to understand someone in this setting. Then, we develop a convolutional neural network model to adapt the robot's speech parameters to different users and spaces, while taking into account the influence of ambient acoustics on intelligibility. Finally, we present an evaluation with 27 users, demonstrating superior intelligibility and user experience with adaptive voice parameters compared to fixed voice., Comment: IEEE Robotics and Automation Letters (IEEE RAL)

Despite their prevalence in deep-learning communities, over-parameterized models convey high demands of computational costs for proper training. This work studies the fine-grained, modular-level learning dynamics of over-parameterized models to attain a more efficient and fruitful training strategy. Empirical evidence reveals that when scaling down into network modules, such as heads in self-attention models, we can observe varying learning patterns implicitly associated with each module's trainability. To describe such modular-level learning capabilities, we introduce a novel concept dubbed modular neural tangent kernel (mNTK), and we demonstrate that the quality of a module's learning is tightly associated with its mNTK's principal eigenvalue $\lambda_{\max}$. A large $\lambda_{\max}$ indicates that the module learns features with better convergence, while those miniature ones may impact generalization negatively. Inspired by the discovery, we propose a novel training strategy termed Modular Adaptive Training (MAT) to update those modules with their $\lambda_{\max}$ exceeding a dynamic threshold selectively, concentrating the model on learning common features and ignoring those inconsistent ones. Unlike most existing training schemes with a complete BP cycle across all network modules, MAT can significantly save computations by its partially-updating strategy and can further improve performance. Experiments show that MAT nearly halves the computational cost of model training and outperforms the accuracy of baselines., Comment: Accepted at NeurIPS 2023

The Cherenkov Telescope Array (CTA) is a future ground-based observatory for gamma-ray astronomy providing unparalleled sensitivity in the energy range from 20 GeV up to 300 TeV. CTA will consist of telescopes with three different sizes. The Medium-Sized Telescopes (MSTs) will have 12 m reflectors with a tessellated mirror design of 86 mirror facets each. Each mirror facet is mounted on the mirror support structure with two actuators that are adjustable in length to align the mirrors, and a freely rotating fixpoint. Image resolution and pointing accuracy constraints impose limits on the backlash and deformation of the actuators and the fixpoint under various weight and wind loads. In this contribution, the test stand to measure the backlash and deformation behaviour of actuators and fixpoints is described and the measurement procedure is explained., Comment: 7 pages, 5 figures. Accepted as Proceeding of the 7th Heidelberg International Symposium on High-Energy Gamma-Ray Astronomy (Gamma2022)

Decoding EEG signals is crucial for unraveling human brain and advancing brain-computer interfaces. Traditional machine learning algorithms have been hindered by the high noise levels and inherent inter-person variations in EEG signals. Recent advances in deep neural networks (DNNs) have shown promise, owing to their advanced nonlinear modeling capabilities. However, DNN still faces challenge in decoding EEG samples of unseen individuals. To address this, this paper introduces a novel approach by incorporating the conditional identification information of each individual into the neural network, thereby enhancing model representation through the synergistic interaction of EEG and personal traits. We test our model on the WithMe dataset and demonstrated that the inclusion of these identifiers substantially boosts accuracy for both subjects in the training set and unseen subjects. This enhancement suggests promising potential for improving for EEG interpretability and understanding of relevant identification features., Comment: Accepted by 6th International Conference on Advances in Signal Processing and Artificial Intelligence (ASPAI' 2024)

Industrial anomaly detection is an important task within computer vision with a wide range of practical use cases. The small size of anomalous regions in many real-world datasets necessitates processing the images at a high resolution. This frequently poses significant challenges concerning memory consumption during the model training and inference stages, leaving some existing methods impractical for widespread adoption. To overcome this challenge, we present the tiled ensemble approach, which reduces memory consumption by dividing the input images into a grid of tiles and training a dedicated model for each tile location. The tiled ensemble is compatible with any existing anomaly detection model without the need for any modification of the underlying architecture. By introducing overlapping tiles, we utilize the benefits of traditional stacking ensembles, leading to further improvements in anomaly detection capabilities beyond high resolution alone. We perform a comprehensive analysis using diverse underlying architectures, including Padim, PatchCore, FastFlow, and Reverse Distillation, on two standard anomaly detection datasets: MVTec and VisA. Our method demonstrates a notable improvement across setups while remaining within GPU memory constraints, consuming only as much GPU memory as a single model needs to process a single tile., Comment: To appear at CVPR 24 Visual Anomaly Detection Workshop. Research conducted during Google Summer of Code 2023 at OpenVINO (Intel). GSoC 2023 page: https://summerofcode.withgoogle.com/archive/2023/projects/WUSjdxGl

Significantly simplifying the creation of optimization models for real-world business problems has long been a major goal in applying mathematical optimization more widely to important business and societal decisions. The recent capabilities of Large Language Models (LLMs) present a timely opportunity to achieve this goal. Therefore, we propose research at the intersection of LLMs and optimization to create a Decision Optimization CoPilot (DOCP) - an AI tool designed to assist any decision maker, interacting in natural language to grasp the business problem, subsequently formulating and solving the corresponding optimization model. This paper outlines our DOCP vision and identifies several fundamental requirements for its implementation. We describe the state of the art through a literature survey and experiments using ChatGPT. We show that a) LLMs already provide substantial novel capabilities relevant to a DOCP, and b) major research challenges remain to be addressed. We also propose possible research directions to overcome these gaps. We also see this work as a call to action to bring together the LLM and optimization communities to pursue our vision, thereby enabling much more widespread improved decision-making.

In a recent work (Dick et al, arXiv:2310.06187), we considered a linear stochastic elasticity equation with random Lam\'e parameters which are parameterized by a countably infinite number of terms in separate expansions. We estimated the expected values over the infinite dimensional parametric space of linear functionals ${\mathcal L}$ acting on the continuous solution $\vu$ of the elasticity equation. This was achieved by truncating the expansions of the random parameters, then using a high-order quasi-Monte Carlo (QMC) method to approximate the high dimensional integral combined with the conforming Galerkin finite element method (FEM) to approximate the displacement over the physical domain $\Omega.$ In this work, as a further development of aforementioned article, we focus on the case of a nearly incompressible linear stochastic elasticity equation. To serve this purpose, in the presence of stochastic inhomogeneous (variable Lam\'e parameters) nearly compressible material, we develop a new locking-free symmetric nonconforming Galerkin FEM that handles the inhomogeneity. In the case of nearly incompressible material, one known important advantage of nonconforming approximations is that they yield optimal order convergence rates that are uniform in the Poisson coefficient. Proving the convergence of the nonconforming FEM leads to another challenge that is summed up in showing the needed regularity properties of $\vu$. For the error estimates from the high-order QMC method, which is needed to estimate the expected value over the infinite dimensional parametric space of ${\mathcal L}\vu,$ we %rely on (Dick et al. 2022). We are required here to show certain regularity properties of $\vu$ with respect to the random coefficients. Some numerical results are delivered at the end.

Stepwise inference protocols, such as scratchpads and chain-of-thought, help language models solve complex problems by decomposing them into a sequence of simpler subproblems. Despite the significant gain in performance achieved via these protocols, the underlying mechanisms of stepwise inference have remained elusive. To address this, we propose to study autoregressive Transformer models on a synthetic task that embodies the multi-step nature of problems where stepwise inference is generally most useful. Specifically, we define a graph navigation problem wherein a model is tasked with traversing a path from a start to a goal node on the graph. Despite is simplicity, we find we can empirically reproduce and analyze several phenomena observed at scale: (i) the stepwise inference reasoning gap, the cause of which we find in the structure of the training data; (ii) a diversity-accuracy tradeoff in model generations as sampling temperature varies; (iii) a simplicity bias in the model's output; and (iv) compositional generalization and a primacy bias with in-context exemplars. Overall, our work introduces a grounded, synthetic framework for studying stepwise inference and offers mechanistic hypotheses that can lay the foundation for a deeper understanding of this phenomenon.