Search

The superconformal index of $\frac12$-BPS states of $N=4$ U(N) super Yang-Mills theory has a known infinite $q$-series expression with successive terms suppressed by $q^N$. We derive a holographic bulk interpretation of this series by evaluating the corresponding functional integral in the dual $AdS_5 \times S^5$. The integral localizes to a product of small fluctuations of the vacuum and of the collective modes of an arbitrary number of giant-gravitons wrapping an $S^3$ of maximal size inside the $S^5$. The quantum mechanics of the small fluctuations of one maximal giant is described by a supersymmetric version of the Landau problem. The quadratic fluctuation determinant reduces to a sum over the supersymmetric ground states, and precisely reproduces the first non-trivial term in the infinite series. Further, we show that the terms corresponding to multiple giants are obtained precisely by the matrix versions of the above super-quantum-mechanics.

Biomarker analysis of athletes' urinary steroid profiles is crucial for the success of anti-doping efforts. Current statistical analysis methods generate personalised limits for each athlete based on univariate modelling of longitudinal biomarker values from the urinary steroid profile. However, simultaneous modelling of multiple biomarkers has the potential to further enhance abnormality detection. In this study, we propose a multivariate Bayesian adaptive model for longitudinal data analysis, which extends the established single-biomarker model in forensic toxicology. The proposed approach employs Markov chain Monte Carlo sampling methods and addresses the scarcity of confirmed abnormal values through a one-class classification algorithm. By adapting decision boundaries as new measurements are obtained, the model provides robust and personalised detection thresholds for each athlete. We tested the proposed approach on a database of 229 athletes which includes longitudinal steroid profiles classified as normal, atypical, or confirmed abnormal. Our results demonstrate improved detection performance, highlighting the potential value of a multivariate approach in doping detection., Comment: 25 pages, main manuscript pgs. 1-19, appendix A pgs. 19-22, appendix B pgs. 23-25

Deep learning algorithms are increasingly employed at the edge. However, edge devices are resource constrained and thus require efficient deployment of deep neural networks. Pruning methods are a key tool for edge deployment as they can improve storage, compute, memory bandwidth, and energy usage. In this paper we propose a novel accurate pruning technique that allows precise control over the output network size. Our method uses an efficient optimal transportation scheme which we make end-to-end differentiable and which automatically tunes the exploration-exploitation behavior of the algorithm to find accurate sparse sub-networks. We show that our method achieves state-of-the-art performance compared to previous pruning methods on 3 different datasets, using 5 different models, across a wide range of pruning ratios, and with two types of sparsity budgets and pruning granularities., Comment: ICCV 2023

The desired outcome for urban digital twins is an automatically generated detailed 3D model of a building from aerial imagery, footprints, LiDAR, or a fusion of these. Such 3D models have applications in architecture, civil engineering, urban planning, construction, real estate, GIS, and many others. Further, the visualization of large-scale data in conjunction with the generated 3D models is often a recurring and resource-intensive task. However, a completely automated end-to-end workflow is complex, requiring many steps to achieve a high-quality visualization. Methods for building reconstruction approaches have come a long way from previously manual approaches to semi-automatic or automatic approaches. The next step after reconstructing buildings is visualizing the buildings and their context. Advances in real-time rendering using game engines have enabled the extension of building reconstruction methods to procedurally generated context generation. This paper aims to complement existing methods of 3D building generation. First, we present a literature review covering different options for procedurally generated context generation and visualization methods in-depth, focusing on workflows and data pipelines. Next, we present a semi-automated workflow that extends the building reconstruction pipeline to include procedural context generation (terrain and vegetation) using Unreal Engine and, finally, the integration of various types of large-scale urban analysis data for visualization. We conclude with a series of challenges faced in achieving such pipelines and the limitations of the current approach. The steps for a complete, end-to-end solution involve developing robust systems for building detection, rooftop recognition, and geometry generation and importing and visualizing data in the same 3D environment.

The need to repeatedly shuttle around synaptic weight values from memory to processing units has been a key source of energy inefficiency associated with hardware implementation of artificial neural networks. Analog in-memory computing (AIMC) with spatially instantiated synaptic weights holds high promise to overcome this challenge, by performing matrix-vector multiplications (MVMs) directly within the network weights stored on a chip to execute an inference workload. However, to achieve end-to-end improvements in latency and energy consumption, AIMC must be combined with on-chip digital operations and communication to move towards configurations in which a full inference workload is realized entirely on-chip. Moreover, it is highly desirable to achieve high MVM and inference accuracy without application-wise re-tuning of the chip. Here, we present a multi-core AIMC chip designed and fabricated in 14-nm complementary metal-oxide-semiconductor (CMOS) technology with backend-integrated phase-change memory (PCM). The fully-integrated chip features 64 256x256 AIMC cores interconnected via an on-chip communication network. It also implements the digital activation functions and processing involved in ResNet convolutional neural networks and long short-term memory (LSTM) networks. We demonstrate near software-equivalent inference accuracy with ResNet and LSTM networks while implementing all the computations associated with the weight layers and the activation functions on-chip. The chip can achieve a maximal throughput of 63.1 TOPS at an energy efficiency of 9.76 TOPS/W for 8-bit input/output matrix-vector multiplications.

Artificial intelligence in the form of deep learning is now very popular and directly implemented in many areas of science and technology. In the present work we study time evolution of Discrete Breathers in one-dimensional nonlinear chains using the framework of Convolutional Neural Networks. We focus on differentiating discrete breathers which are localized nonlinear modes from linearized phonon modes. The breathers are localized in space and time-periodic solutions of non-linear discrete lattices while phonons are the linear collective oscillations of interacting atoms and molecules. We show that deep learning neural networks are indeed able not only to distinguish breather from phonon modes but also determine with high accuracy the underlying nonlinear on-site potentials that generate breathers. This work can have extensions to more complex natural systems.

Visual oddity task was conceived as a universal ethnic-independent analytic intelligence test for humans. Advancements in artificial intelligence led to important breakthroughs, yet competing with humans on such analytic intelligence tasks remains challenging and typically resorts to non-biologically-plausible architectures. We present a biologically realistic system that receives inputs from synthetic eye movements - saccades, and processes them with neurons incorporating dynamics of neocortical neurons. We introduce a procedurally generated visual oddity dataset to train an architecture extending conventional relational networks and our proposed system. Both approaches surpass the human accuracy, and we uncover that both share the same essential underlying mechanism of reasoning. Finally, we show that the biologically inspired network achieves superior accuracy, learns faster and requires fewer parameters than the conventional network.

The Schur index of a $4$ dimensional $\mathcal{N}=2$ superconformal field theory counts (with sign) bosonic and fermionic states that preserve $4$ supercharges. We consider the Schur indices of $4$d $\mathcal{N}=4$ super Yang-Mills and $\mathcal{N}=2$ circular quiver gauge theories with gauge groups $U(N)$ or $SU(N)$. We calculate the exponentially dominant part of their asymptotic expansions as the index parameter $q$ approaches any root of unity. We find that some of the indices exhibit ``small" ($\mathcal{O}(N^0)$ as $N \rightarrow \infty$) exponential growth, which is much smaller than an $\mathcal{O}(N^2)$ exponential growth of states that is indicative of a black hole. This implies that the indices do not capture a growth of states that would correspond to a supersymmetric black hole that preserves 4 supercharges in the holographic dual AdS theory. Interestingly, the exponentially dominant part in the Schur asymptotics we consider, depends on the parity of the rank $N$.

Machine learning methodologies can be adopted in cultural applications and propose new ways to distribute or even present the cultural content to the public. For instance, speech analytics can be adopted to automatically generate subtitles in theatrical plays, in order to (among other purposes) help people with hearing loss. Apart from a typical speech-to-text transcription with Automatic Speech Recognition (ASR), Speech Emotion Recognition (SER) can be used to automatically predict the underlying emotional content of speech dialogues in theatrical plays, and thus to provide a deeper understanding how the actors utter their lines. However, real-world datasets from theatrical plays are not available in the literature. In this work we present GreThE, the Greek Theatrical Emotion dataset, a new publicly available data collection for speech emotion recognition in Greek theatrical plays. The dataset contains utterances from various actors and plays, along with respective valence and arousal annotations. Towards this end, multiple annotators have been asked to provide their input for each speech recording and inter-annotator agreement is taken into account in the final ground truth generation. In addition, we discuss the results of some indicative experiments that have been conducted with machine and deep learning frameworks, using the dataset, along with some widely used databases in the field of speech emotion recognition.

This study outlines the synthesis and physicochemical characteristics of a solution-processable iron manganite (FeMnO3) nanoparticles via a chemical combustion method using tartartic acid as a fuel and demonstrates the performance of this material as a n-type photoactive layer in all-oxide solar cells. It is shown that the solution combustion synthesis (SCS) method enables the formation of pure crystal phase FeMnO3 with controllable particle size. XRD pattern and morphology images from TEM confirm the purity of FeMnO3 phase and the relative small crystallite size (~13 nm), firstly reported in the literature. Moreover, to assemble a network of connected FeMnO3 nanoparticles, \b{eta}-alanine was used as a capping agent and dimethylformamide (DMF) as a polar aprotic solvent for the colloidal dispersion of FeMnO3 NPs. This procedure yields a ~500 nm thick photoactive layer. The proposed method is crucial to obtain functional solution processed NiO/FeMnO3 heterojunction inorganic photovoltaics. The optoelectronic properties of the heterojunction were established. These solar cells demonstrate a high open circuit voltage of 1.31 V with sufficient fill factor of 54.3% and low short circuit current of 0.07 mA cm-2 delivering a power conversion efficiency of 0.05% under 100 mW cm-2 illumination. This work expands on the burgeoning of environmentally friendly, low-cost, sustainable solar cell material that derive from metal oxides., Comment: 19 pages, 8 figures

For many diseases it is reasonable to assume that the hazard rate is not constant across time, but also that it changes in different time intervals. To capture this, we work here with a piecewise survival model. One of the major problems in such piecewise models is to determine the time points of change of the hazard rate. From the practical point of view this can provide very important information as it may reflect changes in the progress of a disease. We present piecewise Weibull regression models with covariates. The time points where change occurs are assumed unknown and need to be estimated. The equality of hazard rates across the distinct phases is also examined to verify the exact number of phases. An example based on herpes zoster data has been used to demonstrate the usefulness of the developed methodology.

Automatic speech recognition (ASR) is a capability which enables a program to process human speech into a written form. Recent developments in artificial intelligence (AI) have led to high-accuracy ASR systems based on deep neural networks, such as the recurrent neural network transducer (RNN-T). However, the core components and the performed operations of these approaches depart from the powerful biological counterpart, i.e., the human brain. On the other hand, the current developments in biologically-inspired ASR models, based on spiking neural networks (SNNs), lag behind in terms of accuracy and focus primarily on small scale applications. In this work, we revisit the incorporation of biologically-plausible models into deep learning and we substantially enhance their capabilities, by taking inspiration from the diverse neural and synaptic dynamics found in the brain. In particular, we introduce neural connectivity concepts emulating the axo-somatic and the axo-axonic synapses. Based on this, we propose novel deep learning units with enriched neuro-synaptic dynamics and integrate them into the RNN-T architecture. We demonstrate for the first time, that a biologically realistic implementation of a large-scale ASR model can yield competitive performance levels compared to the existing deep learning models. Specifically, we show that such an implementation bears several advantages, such as a reduced computational cost and a lower latency, which are critical for speech recognition applications., Comment: Accepted at the Efficient Natural Language and Speech Processing workshop at NeurIPS 2021