Searcharxiv⌕ Search

arXiv subjects

Robert Wille

Publications and source records attributed to Robert Wille.

At least 127 records · Page 7Linked to original sources
arXiv2024

Efficient Post-Training Augmentation for Adaptive Inference in Heterogeneous and Distributed IoT Environments

Early Exit Neural Networks (EENNs) present a solution to enhance the efficiency of neural network deployments. However, creating EENNs is challenging and requires specialized domain knowledge, due to the large amount of additional design choices. To address this issue, we propose an automated augmentation flow that focuses on converting an existing model into an EENN. It performs all required design decisions for the deployment to heterogeneous or distributed hardware targets: Our framework constructs the EENN architecture, maps its subgraphs to the hardware targets, and configures its decision mechanism. To the best of our knowledge, it is the first framework that is able to perform all of these steps. We evaluated our approach on a collection of Internet-of-Things and standard image classification use cases. For a speech command detection task, our solution was able to reduce the mean operations per inference by 59.67%. For an ECG classification task, it was able to terminate all samples early, reducing the mean inference energy by 74.9% and computations by 78.3%. On CIFAR-10, our solution was able to achieve up to a 58.75% reduction in computations. The search on a ResNet-152 base model for CIFAR-10 took less than nine hours on a laptop CPU. Our proposed approach enables the creation of EENN optimized for IoT environments and can reduce the inference cost of Deep Learning applications on embedded and fog platforms, while also significantly reducing the search cost - making it more accessible for scientists and engineers in industry and research. The low search cost improves the accessibility of EENNs, with the potential to improve the efficiency of neural networks in a wide range of practical applications.

cs.LG↗
arXiv2024

Temporal Decisions: Leveraging Temporal Correlation for Efficient Decisions in Early Exit Neural Networks

Deep Learning is becoming increasingly relevant in Embedded and Internet-of-things applications. However, deploying models on embedded devices poses a challenge due to their resource limitations. This can impact the model's inference accuracy and latency. One potential solution are Early Exit Neural Networks, which adjust model depth dynamically through additional classifiers attached between their hidden layers. However, the real-time termination decision mechanism is critical for the system's efficiency, latency, and sustained accuracy. This paper introduces Difference Detection and Temporal Patience as decision mechanisms for Early Exit Neural Networks. They leverage the temporal correlation present in sensor data streams to efficiently terminate the inference. We evaluate their effectiveness in health monitoring, image classification, and wake-word detection tasks. Our novel contributions were able to reduce the computational footprint compared to established decision mechanisms significantly while maintaining higher accuracy scores. We achieved a reduction of mean operations per inference by up to 80% while maintaining accuracy levels within 5% of the original model. These findings highlight the importance of considering temporal correlation in sensor data to improve the termination decision.

cs.LG↗
arXiv2024

Towards Hamiltonian Simulation with Decision Diagrams

This paper proposes a novel approach to Hamiltonian simulation using Decision Diagrams (DDs), which are an exact representation based on exploiting redundancies in representations of quantum states and operations. While the simulation of Hamiltonians has been studied extensively, scaling these simulations to larger or more complex systems is often challenging and may require approximations or new simulation methods altogether. DDs offer such an alternative that has not yet been applied to Hamiltonian simulation. In this work, we investigate the behavior of DDs for this task. To this end, we review the basics of DDs such as their construction and present how the relevant operations for Hamiltonian simulation are implemented in this data structure -- leading to the first DD-based Hamiltonian simulation approach. Based on several series of evaluations and comparisons, we then discuss insights about the performance of this complementary approach. Overall, these studies show that DDs indeed may offer a promising new data structure which, for certain examples, can provide orders of magnitudes of improvement compared to the state-of-the-art, yet also comes with its own, fundamentally different, limitations.

quant-ph↗
arXiv2024

Accelerated Computational Fluid Dynamics Simulations of Microfluidic Devices by Exploiting Higher Levels of Abstraction

The design of microfluidic devices is a cumbersome and tedious process that can be significantly improved by simulation. Methods based on Computational Fluid Dynamics (CFD) are considered state-of-the-art, but require extensive compute time - oftentimes limiting the size of microfluidic devices that can be simulated. Simulation methods that abstract the underlying physics on a higher level generally provide results instantly, but the fidelity of these methods is usually worse. In this work, a simulation method that accelerates CFD simulations by exploiting simulation methods on higher levels of abstraction is proposed. Case studies confirm that the proposed method accelerates CFD simulations by multiple factors (often several orders of magnitude) while maintaining the fidelity of CFD simulations.

cs.CE↗
arXiv2023

Variational Quantum Circuit Design for Quantum Reinforcement Learning on Continuous Environments

Quantum Reinforcement Learning (QRL) emerged as a branch of reinforcement learning (RL) that uses quantum submodules in the architecture of the algorithm. One branch of QRL focuses on the replacement of neural networks (NN) by variational quantum circuits (VQC) as function approximators. Initial works have shown promising results on classical environments with discrete action spaces, but many of the proposed architectural design choices of the VQC lack a detailed investigation. Hence, in this work we investigate the impact of VQC design choices such as angle embedding, encoding block architecture and postprocessesing on the training capabilities of QRL agents. We show that VQC design greatly influences training performance and heuristically derive enhancements for the analyzed components. Additionally, we show how to design a QRL agent in order to solve classical environments with continuous action spaces and benchmark our agents against classical feed-forward NNs.

quant-ph↗
arXiv2023

Hybrid Circuit Mapping: Leveraging the Full Spectrum of Computational Capabilities of Neutral Atom Quantum Computers

Quantum computing based on Neutral Atoms (NAs) provides a wide range of computational capabilities, encompassing high-fidelity long-range interactions with native multi-qubit gates, and the ability to shuttle arrays of qubits. While previously these capabilities have been studied individually, we propose the first approach of a fast hybrid compiler to perform circuit mapping and routing based on both high-fidelity gate interactions and qubit shuttling. We delve into the intricacies of the compilation process when combining multiple capabilities and present effective solutions to address resulting challenges. The final compilation strategy is then showcased across various hardware settings, revealing its versatility, and highlighting potential fidelity enhancements achieved through the strategic utilization of combined gate- and shuttling-based routing. With the additional multi-qubit gate support for both routing capabilities, the proposed approach is able to take advantage of the full spectrum of computational capabilities offered by NAs.

quant-ph↗
arXiv2023

Atomic Defect-Aware Physical Design of Silicon Dangling Bond Logic on the H-Si(100)2x1 Surface

Although fabrication capabilities of Silicon Dangling Bonds have rapidly advanced from manual labor-driven laboratory work to automated manufacturing in just recent years, sub-nanometer substrate defects still pose a hindrance to production due to the need for atomic precision. In essence, unpassivated or missing surface atoms, contaminants, and structural deformations disturb the fabricated logic or prevent its realization altogether. Moreover, design automation techniques in this domain have not yet adopted any defect-aware behavior to circumvent the present obstacles. In this paper, we derive a surface defect model for design automation from experimentally verified defect types that we apply to identify sensitivities in an established gate library in an effort to generate more robust designs. Furthermore, we present an automatic placement and routing algorithm that considers scanning tunneling microscope data obtained from physical experiments to lay out dot-accurate circuitry that is resilient against the presence of atomic surface defects. This culminates in a holistic evaluation on surface data of varying defect rates that enables us to quantify the severity of such defects. We project that fabrication capabilities must achieve defect rates of around 0.1 %, if charged defects can be completely eliminated, or < 0.1 %, otherwise. This realization sets the pace for future efforts to scale up this promising circuit technology.

physics.app-ph↗
arXiv2023

Using Boolean Satisfiability for Exact Shuttling in Trapped-Ion Quantum Computers

Trapped ions are a promising technology for building scalable quantum computers. Not only can they provide a high qubit quality, but they also enable modular architectures, referred to as Quantum Charge Coupled Device (QCCD) architecture. Within these devices, ions can be shuttled (moved) throughout the trap and through different dedicated zones, e.g., a memory zone for storage and a processing zone for the actual computation. However, this movement incurs a cost in terms of required time steps, which increases the probability of decoherence, and, thus, should be minimized. In this paper, we propose a formalization of the possible movements in ion traps via Boolean satisfiability. This formalization allows for determining the minimal number of time steps needed for a given quantum algorithm and device architecture, hence reducing the decoherence probability. An empirical evaluation confirms that -- using the proposed approach -- minimal results (i.e., the lower bound) can be determined for the first time. An open-source implementation of the proposed approach is publicly available at https://github.com/cda-tum/mqt-ion-shuttler.

quant-ph↗
arXiv2023

Recommending Solution Paths for Solving Optimization Problems with Quantum Computing

Solving real-world optimization problems with quantum computing requires choosing between a large number of options concerning formulation, encoding, algorithm and hardware. Finding good solution paths is challenging for end users and researchers alike. We propose a framework designed to identify and recommend the best-suited solution paths. This introduces a novel abstraction layer that is required to make quantum-computing-assisted solution techniques accessible to end users without requiring a deeper knowledge of quantum technologies. State-of-the-art hybrid algorithms, encoding and decomposition techniques can be integrated in a modular manner and evaluated using problem-specific performance metrics. Equally, tools for the graphical analysis of variational quantum algorithms are developed. Classical, fault tolerant quantum and quantum-inspired methods can be included as well to ensure a fair comparison resulting in useful solution paths. We demonstrate and validate our approach on a selected set of options and illustrate its application on the capacitated vehicle routing problem (CVRP). We also identify crucial requirements and the major design challenges for the proposed automation layer within a quantum-assisted solution workflow for optimization problems.

quant-ph↗
arXiv2023

Temporal Patience: Efficient Adaptive Deep Learning for Embedded Radar Data Processing

Radar sensors offer power-efficient solutions for always-on smart devices, but processing the data streams on resource-constrained embedded platforms remains challenging. This paper presents novel techniques that leverage the temporal correlation present in streaming radar data to enhance the efficiency of Early Exit Neural Networks for Deep Learning inference on embedded devices. These networks add additional classifier branches between the architecture's hidden layers that allow for an early termination of the inference if their result is deemed sufficient enough by an at-runtime decision mechanism. Our methods enable more informed decisions on when to terminate the inference, reducing computational costs while maintaining a minimal loss of accuracy. Our results demonstrate that our techniques save up to 26% of operations per inference over a Single Exit Network and 12% over a confidence-based Early Exit version. Our proposed techniques work on commodity hardware and can be combined with traditional optimizations, making them accessible for resource-constrained embedded platforms commonly used in smart devices. Such efficiency gains enable real-time radar data processing on resource-constrained platforms, allowing for new applications in the context of smart homes, Internet-of-Things, and human-computer interaction.

cs.LG↗
arXiv2023

Mixed-Dimensional Quantum Circuit Simulation with Decision Diagrams

Quantum computers promise to solve several categories of problems faster than classical computers ever could. Current research mostly focuses on qubits, i.e., systems where the unit of information can assume only two levels. However, the underlying physics of most (if not all) of the technological platforms supports more than two levels, commonly referred to as qudits. Performing computations with qudits increases the overall complexity while, at the same time, reducing the number of operations and providing a lower error rate. Furthermore, qudits with different number of levels can be mixed in one system to ease the experimental control and keep representations as compact as possible. Exploiting these capabilities requires dedicated software support to tackle the increased complexity in an automated and efficient fashion. In this paper, we present a qudit simulator that handles mixed-dimensional systems based on Decision Diagrams (DDs). More precisely, we discuss the type of decision diagram introduced as underlying data structure as well as the resulting implementation. Experimental evaluations demonstrate that the proposed solution is capable of efficiently simulating mixed-dimensional quantum circuits, with specific use cases including more than 100 qudits in one circuit. The source code of the simulator is available via github.com/cda-tum/MiSiM under the MIT~license.

quant-ph↗
arXiv2023

Simulating Charged Defects in Silicon Dangling Bond Logic Systems to Evaluate Logic Robustness

Recent research interest in emerging logic systems based on quantum dots has been sparked by the experimental demonstration of nanometer-scale logic devices composed of atomically sized quantum dots made of silicon dangling bonds (SiDBs), along with the availability of SiQAD, a computer-aided design tool designed for this technology. Latest design automation frameworks have enabled the synthesis of SiDB circuits that reach the size of $32\times10^3\,\text{nm}^{2}$ -- orders of magnitude more complex than their hand-designed counterparts. However, current SiDB simulation engines do not take defects into account, which is important to consider for these sizable systems. This work proposes a formulation for incorporating fixed-charge simulation into established ground state models to cover an important class of defects that has a non-negligible effect on nearby SiDBs at the $10\,\text{nm}$ scale and beyond. The formulation is validated by implementing it into SiQAD's simulation engine and computationally reproducing experiments on multiple defect types, revealing a high level of accuracy. The new capability is applied towards studying the tolerance of several established logic gates against the introduction of a single nearby defect to establish the corresponding minimum required clearance. These findings are compared against existing metrics to form a foundation for logic robustness studies.

cond-mat.mes-hall↗
arXiv2023

The Need for Speed: Efficient Exact Simulation of Silicon Dangling Bond Logic

The Silicon Dangling Bond (SiDB) logic platform, an emerging computational beyond-CMOS nanotechnology, is a promising competitor due to its ability to achieve integration density and clock speed values that are several orders of magnitude higher compared to current CMOS fabrication nodes. However, the exact physical simulation of SiDB layouts, which is an essential component of any design validation workflow, is computationally expensive. In this paper, we propose a novel algorithm called QuickExact, which aims to be both, efficient and exact. To this end, we are introducing three techniques, namely 1) Physically-informed Search Space Pruning, 2) Partial Solution Caching, and 3) Effective State Enumeration. Extensive experimental evaluations confirm that, compared to the state-of-the-art algorithm, the resulting approach leads to a paramount runtime advantage of more than a factor of 5000 on randomly generated layouts and more than a factor of 2000 on an established gate library.

physics.app-ph↗
arXiv2023

A Hybrid Classical Quantum Computing Approach to the Satellite Mission Planning Problem

Hundreds of satellites equipped with cameras orbit the Earth to capture images from locations for various purposes. Since the field of view of the cameras is usually very narrow, the optics have to be adjusted and rotated between single shots of different locations. This is even further complicated by the fixed speed -- determined by the satellite's altitude -- such that the decision what locations to select for imaging becomes even more complex. Therefore, classical algorithms for this Satellite Mission Planning Problem (SMPP) have already been proposed decades ago. However, corresponding classical solutions have only seen evolutionary enhancements since then. Quantum computing and its promises, on the other hand, provide the potential for revolutionary improvement. Therefore, in this work, we propose a hybrid classical quantum computing approach to solve the SMPP combining the advantages of quantum hardware with decades of classical optimizer development. Using the Variational Quantum Eigensolver (VQE), Quantum Approximate Optimization Algorithm (QAOA), and its warm-start variant (W-QAOA), we demonstrate the applicability of solving the SMPP for up to 21 locations to choose from. This proof-of-concept -- which is available on GitHub (https://github.com/cda-tum/mqt-problemsolver) as part of the Munich Quantum Toolkit (MQT) -- showcases the potential of quantum computing in this application domain and represents a first step toward competing with classical algorithms in the future.

quant-ph↗
arXiv2023

Reducing the Compilation Time of Quantum Circuits Using Pre-Compilation on the Gate Level

In order to implement a quantum computing application, problem instances must be encoded into a quantum circuit and then compiled for a specific platform. The lengthy compilation process is a key bottleneck in this workflow, especially for problems that arise repeatedly with a similar yet distinct structure (each of which requires a new compilation run thus far). In this paper, we aim to overcome this bottleneck by proposing a comprehensive pre-compilation technique that tries to minimize the time spent for compiling recurring problems while preserving the solution quality as much as possible. The following concepts underpin the proposed approach: Beginning with a problem class and a corresponding quantum algorithm, a predictive encoding scheme is applied to encode a representative problem instance into a general-purpose quantum circuit for that problem class. Once the real problem instance is known, the previously constructed circuit only needs to be adjusted -- with (nearly) no compilation necessary. Experimental evaluations on QAOA for the MaxCut problem as well as a case study involving a satellite mission planning problem show that the proposed approach significantly reduces the compilation time by several orders of magnitude compared to Qiskit's compilation schemes while maintaining comparable compiled circuit quality. All implementations are available on GitHub (https://github.com/cda-tum/mqt-problemsolver) as part of the Munich Quantum Toolkit (MQT).

quant-ph↗
arXiv2023

Multi-Task Cross-Modality Attention-Fusion for 2D Object Detection

Accurate and robust object detection is critical for autonomous driving. Image-based detectors face difficulties caused by low visibility in adverse weather conditions. Thus, radar-camera fusion is of particular interest but presents challenges in optimally fusing heterogeneous data sources. To approach this issue, we propose two new radar preprocessing techniques to better align radar and camera data. In addition, we introduce a Multi-Task Cross-Modality Attention-Fusion Network (MCAF-Net) for object detection, which includes two new fusion blocks. These allow for exploiting information from the feature maps more comprehensively. The proposed algorithm jointly detects objects and segments free space, which guides the model to focus on the more relevant part of the scene, namely, the occupied space. Our approach outperforms current state-of-the-art radar-camera fusion-based object detectors in the nuScenes dataset and achieves more robust results in adverse weather conditions and nighttime scenarios.

cs.CV↗
arXiv2023

MQT Bench: Benchmarking Software and Design Automation Tools for Quantum Computing

Quantum software tools for a wide variety of design tasks on and across different levels of abstraction are crucial in order to eventually realize useful quantum applications. This requires practical and relevant benchmarks for new software tools to be empirically evaluated and compared to the current state of the art. Although benchmarks for specific design tasks are commonly available, the demand for an overarching cross-level benchmark suite has not yet been fully met and there is no mutual consolidation in how quantum software tools are evaluated thus far. In this work, we propose the MQT Bench benchmark suite (as part of the Munich Quantum Toolkit, MQT) based on four core traits: (1) cross-level support for different abstraction levels, (2) accessibility via an easy-to-use web interface (https://www.cda.cit.tum.de/mqtbench) and a Python package, (3) provision of a broad selection of benchmarks to facilitate generalizability, as well as (4) extendability to future algorithms, gate-sets, and hardware architectures. By comprising more than 70,000 benchmark circuits ranging from 2 to 130 qubits on four abstraction levels, MQT Bench presents a first step towards benchmarking different abstraction levels with a single benchmark suite to increase comparability, reproducibility, and transparency.

quant-ph↗
arXiv2023

Detection of Sensor-To-Sensor Variations using Explainable AI

With the growing concern for air quality and its impact on human health, interest in environmental gas monitoring has increased. However, chemi-resistive gas sensing devices are plagued by issues of sensor reproducibility during manufacturing. This study proposes a novel approach for detecting sensor-to-sensor variations in sensing devices using the explainable AI (XAI) method of SHapley Additive exPlanations (SHAP). This is achieved by identifying sensors that contribute the most to environmental gas concentration estimation via machine learning, and measuring the similarity of feature rankings between sensors to flag deviations or outliers. The methodology is tested using artificial and realistic Ozone concentration profiles to train a Gated Recurrent Unit (GRU) model. Two applications were explored in the study: the detection of wrong behaviors of sensors in the train dataset, and the detection of deviations in the test dataset. By training the GRU with the pruned train dataset, we could reduce computational costs while improving the model performance. Overall, the results show that our approach improves the understanding of sensor behavior, successfully detects sensor deviations down to 5-10% from the normal behavior, and leads to more efficient model preparation and calibration. Our method provides a novel solution for identifying deviating sensors, linking inconsistencies in hardware to sensor-to-sensor variations in the manufacturing process on an AI model-level.

cs.LG↗