Search

Financial market prediction and optimal trading strategy development remain challenging due to market complexity and volatility. Our research in quantum finance and reinforcement learning for decision-making demonstrates the approach of quantum-classical hybrid algorithms to tackling real-world financial challenges. In this respect, we corroborate the concept with rigorous backtesting and validate the framework's performance under realistic market conditions, by including fixed transaction cost per trade. This paper introduces a Quantum Attention Deep Q-Network (QADQN) approach to address these challenges through quantum-enhanced reinforcement learning. Our QADQN architecture uses a variational quantum circuit inside a traditional deep Q-learning framework to take advantage of possible quantum advantages in decision-making. We gauge the QADQN agent's performance on historical data from major market indices, including the S&P 500. We evaluate the agent's learning process by examining its reward accumulation and the effectiveness of its experience replay mechanism. Our empirical results demonstrate the QADQN's superior performance, achieving better risk-adjusted returns with Sortino ratios of 1.28 and 1.19 for non-overlapping and overlapping test periods respectively, indicating effective downside risk management., Comment: Accepted at the 2024 IEEE International Conference on Quantum Computing and Engineering (QCE24), QCRL, September 2024

In this study, we develop a novel quantum machine learning (QML) framework to analyze cybersecurity vulnerabilities using data from the 2022 CISA Known Exploited Vulnerabilities catalog, which includes detailed information on vulnerability types, severity levels, common vulnerability scoring system (CVSS) scores, and product specifics. Our framework preprocesses this data into a quantum-compatible format, enabling clustering analysis through our advanced quantum techniques, QCSWAPK-means and QkernelK-means. These quantum algorithms demonstrate superior performance compared to state-of-the-art classical clustering techniques like k-means and spectral clustering, achieving Silhouette scores of 0.491, Davies-Bouldin indices below 0.745, and Calinski-Harabasz scores exceeding 884, indicating more distinct and well-separated clusters. Our framework categorizes vulnerabilities into distinct groups, reflecting varying levels of risk severity: Cluster 0, primarily consisting of critical Microsoft-related vulnerabilities; Cluster 1, featuring medium severity vulnerabilities from various enterprise software vendors and network solutions; Cluster 2, with high severity vulnerabilities from Adobe, Cisco, and Google; and Cluster 3, encompassing vulnerabilities from Microsoft and Oracle with high to medium severity. These findings highlight the potential of QML to enhance the precision of vulnerability assessments and prioritization, advancing cybersecurity practices by enabling more strategic and proactive defense mechanisms., Comment: Accepted at the 2024 IEEE International Conference on Quantum Computing and Engineering (QCE24), QML@QCE, September 2024

Portfolio Optimization (PO) is a financial problem aiming to maximize the net gains while minimizing the risks in a given investment portfolio. The novelty of Quantum algorithms lies in their acclaimed potential and capability to solve complex problems given the underlying Quantum Computing (QC) infrastructure. Utilizing QC's applicable strengths to the finance industry's problems, such as PO, allows us to solve these problems using quantum-based algorithms such as Variational Quantum Eigensolver (VQE) and Quantum Approximate Optimization Algorithm (QAOA). While the Quantum potential for finance is highly impactful, the architecture and composition of the quantum circuits have not yet been properly defined as robust financial frameworks/algorithms as state of the art in present literature for research and design development purposes. In this work, we propose a novel scalable framework, denoted PO-QA, to systematically investigate the variation of quantum parameters (such as rotation blocks, repetitions, and entanglement types) to observe their subtle effect on the overall performance. In our paper, the performance is measured and dictated by convergence to similar ground-state energy values for resultant optimal solutions by each algorithm variation set for QAOA and VQE to the exact eigensolver (classical solution). Our results provide effective insights into comprehending PO from the lens of Quantum Machine Learning in terms of convergence to the classical solution, which is used as a benchmark. This study paves the way for identifying efficient configurations of quantum circuits for solving PO and unveiling their inherent inter-relationships., Comment: Accepted at the 2024 IEEE International Conference on Quantum Computing and Engineering (QCE24), September 2024

Autonomous embedded systems (e.g., robots) typically necessitate intelligent computation with low power/energy processing for completing their tasks. Such requirements can be fulfilled by embodied neuromorphic intelligence with spiking neural networks (SNNs) because of their high learning quality (e.g., accuracy) and sparse computation. Here, the employment of event-based data is preferred to ensure seamless connectivity between input and processing parts. However, state-of-the-art SNNs still face a long training time to achieve high accuracy, thereby incurring high energy consumption and producing a high rate of carbon emission. Toward this, we propose FastSpiker, a novel methodology that enables fast SNN training on event-based data through learning rate enhancements targeting autonomous embedded systems. In FastSpiker, we first investigate the impact of different learning rate policies and their values, then select the ones that quickly offer high accuracy. Afterward, we explore different settings for the selected learning rate policies to find the appropriate policies through a statistical-based decision. Experimental results show that our FastSpiker offers up to 10.5x faster training time and up to 88.39% lower carbon emission to achieve higher or comparable accuracy to the state-of-the-art on the event-based automotive dataset (i.e., NCARS). In this manner, our FastSpiker methodology paves the way for green and sustainable computing in realizing embodied neuromorphic intelligence for autonomous embedded systems., Comment: To appear at the 18th International Conference on Control, Automation, Robotics and Vision (ICARCV), December 2024, Dubai, UAE

Recent advancements in quantum computing have led to the emergence of hybrid quantum neural networks, such as Quanvolutional Neural Networks (QuNNs), which integrate quantum and classical layers. While the susceptibility of classical neural networks to adversarial attacks is well-documented, the impact on QuNNs remains less understood. This study introduces RobQuNN, a new methodology to enhance the robustness of QuNNs against adversarial attacks, utilizing quantum circuit expressibility and entanglement capability alongside different adversarial strategies. Additionally, the study investigates the transferability of adversarial examples between classical and quantum models using RobQuNN, enhancing our understanding of cross-model vulnerabilities and pointing to new directions in quantum cybersecurity. The findings reveal that QuNNs exhibit up to 60\% higher robustness compared to classical networks for the MNIST dataset, particularly at low levels of perturbation. This underscores the potential of quantum approaches in improving security defenses. In addition, RobQuNN revealed that QuNN does not exhibit enhanced resistance or susceptibility to cross-model adversarial examples regardless of the quantum circuit architecture., Comment: 6 pages, 3 figures

Quantization techniques are widely used to improve inference speed and deployment of large language models. While a wide body of work examines the impact of quantization on LLMs in English, none have evaluated across languages. We conduct a thorough analysis of quantized multilingual LLMs, focusing on performance across languages and at varying scales. We use automatic benchmarks, LLM-as-a-Judge, and human evaluation, finding that (1) harmful effects of quantization are apparent in human evaluation, which automatic metrics severely underestimate: a 1.7% average drop in Japanese across automatic tasks corresponds to a 16.0% drop reported by human evaluators on realistic prompts; (2) languages are disparately affected by quantization, with non-Latin script languages impacted worst; and (3) challenging tasks like mathematical reasoning degrade fastest. As the ability to serve low-compute models is critical for wide global adoption of NLP technologies, our results urge consideration of multilingual performance as a key evaluation criterion for efficient models., Comment: Findings of EMNLP 2024 Camera-Ready

Preference optimization techniques have become a standard final stage for training state-of-art large language models (LLMs). However, despite widespread adoption, the vast majority of work to-date has focused on first-class citizen languages like English and Chinese. This captures a small fraction of the languages in the world, but also makes it unclear which aspects of current state-of-the-art research transfer to a multilingual setting. In this work, we perform an exhaustive study to achieve a new state-of-the-art in aligning multilingual LLMs. We introduce a novel, scalable method for generating high-quality multilingual feedback data to balance data coverage. We establish the benefits of cross-lingual transfer and increased dataset size in preference training. Our preference-trained model achieves a 54.4% win-rate against Aya 23 8B, the current state-of-the-art multilingual LLM in its parameter class, and a 69.5% win-rate or higher against widely used models like Gemma-1.1-7B-it, Llama-3-8B-Instruct, Mistral-7B-Instruct-v0.3. As a result of our study, we expand the frontier of alignment techniques to 23 languages covering half of the world's population.

We investigate a surprising limitation of LLMs: their inability to consistently generate text in a user's desired language. We create the Language Confusion Benchmark (LCB) to evaluate such failures, covering 15 typologically diverse languages with existing and newly-created English and multilingual prompts. We evaluate a range of LLMs on monolingual and cross-lingual generation reflecting practical use cases, finding that Llama Instruct and Mistral models exhibit high degrees of language confusion and even the strongest models fail to consistently respond in the correct language. We observe that base and English-centric instruct models are more prone to language confusion, which is aggravated by complex prompts and high sampling temperatures. We find that language confusion can be partially mitigated via few-shot prompting, multilingual SFT and preference tuning. We release our language confusion benchmark, which serves as a first layer of efficient, scalable multilingual evaluation at https://github.com/for-ai/language-confusion., Comment: EMNLP 2024 Main Conference Camera-ready

This technical report introduces Aya 23, a family of multilingual language models. Aya 23 builds on the recent release of the Aya model (\"Ust\"un et al., 2024), focusing on pairing a highly performant pre-trained model with the recently released Aya collection (Singh et al., 2024). The result is a powerful multilingual large language model serving 23 languages, expanding state-of-art language modeling capabilities to approximately half of the world's population. The Aya model covered 101 languages whereas Aya 23 is an experiment in depth vs breadth, exploring the impact of allocating more capacity to fewer languages that are included during pre-training. Aya 23 outperforms both previous massively multilingual models like Aya 101 for the languages it covers, as well as widely used models like Gemma, Mistral and Mixtral on an extensive range of discriminative and generative tasks. We release the open weights for both the 8B and 35B models as part of our continued commitment for expanding access to multilingual progress.

Recent trends have shown that autonomous agents, such as Autonomous Ground Vehicles (AGVs), Unmanned Aerial Vehicles (UAVs), and mobile robots, effectively improve human productivity in solving diverse tasks. However, since these agents are typically powered by portable batteries, they require extremely low power/energy consumption to operate in a long lifespan. To solve this challenge, neuromorphic computing has emerged as a promising solution, where bio-inspired Spiking Neural Networks (SNNs) use spikes from event-based cameras or data conversion pre-processing to perform sparse computations efficiently. However, the studies of SNN deployments for autonomous agents are still at an early stage. Hence, the optimization stages for enabling efficient embodied SNN deployments for autonomous agents have not been defined systematically. Toward this, we propose a novel framework called SNN4Agents that consists of a set of optimization techniques for designing energy-efficient embodied SNNs targeting autonomous agent applications. Our SNN4Agents employs weight quantization, timestep reduction, and attention window reduction to jointly improve the energy efficiency, reduce the memory footprint, optimize the processing latency, while maintaining high accuracy. In the evaluation, we investigate use cases of event-based car recognition, and explore the trade-offs among accuracy, latency, memory, and energy consumption. The experimental results show that our proposed framework can maintain high accuracy (i.e., 84.12% accuracy) with 68.75% memory saving, 3.58x speed-up, and 4.03x energy efficiency improvement as compared to the state-of-the-art work for NCARS dataset. In this manner, our SNN4Agents framework paves the way toward enabling energy-efficient embodied SNN deployments for autonomous agents., Comment: Accepted for publication at Frontiers in Robotics and AI (FROBT) - Section Robot Vision and Artificial Perception

Autonomous Driving (AD) systems are considered as the future of human mobility and transportation. Solving computer vision tasks such as image classification and object detection/segmentation, with high accuracy and low power/energy consumption, is highly needed to realize AD systems in real life. These requirements can potentially be satisfied by Spiking Neural Networks (SNNs). However, the state-of-the-art works in SNN-based AD systems still focus on proposing network models that can achieve high accuracy, and they have not systematically studied the roles of SNN parameters when used for learning event-based automotive data. Therefore, we still lack understanding of how to effectively develop SNN models for AD systems. Toward this, we propose a novel methodology to systematically study and analyze the impact of SNN parameters considering event-based automotive data, then leverage this analysis for enhancing SNN developments. To do this, we first explore different settings of SNN parameters that directly affect the learning mechanism (i.e., batch size, learning rate, neuron threshold potential, and weight decay), then analyze the accuracy results. Afterward, we propose techniques that jointly improve SNN accuracy and reduce training time. Experimental results show that our methodology can improve the SNN models for AD systems than the state-of-the-art, as it achieves higher accuracy (i.e., 86%) for the NCARS dataset, and it can also achieve iso-accuracy (i.e., ~85% with standard deviation less than 0.5%) while speeding up the training time by 1.9x. In this manner, our research work provides a set of guidelines for SNN parameter enhancements, thereby enabling the practical developments of SNN-based AD systems., Comment: To appear at the 18th International Conference on Control, Automation, Robotics and Vision (ICARCV), December 2024, Dubai, UAE

Robotic technologies have been an indispensable part for improving human productivity since they have been helping humans in completing diverse, complex, and intensive tasks in a fast yet accurate and efficient way. Therefore, robotic technologies have been deployed in a wide range of applications, ranging from personal to industrial use-cases. However, current robotic technologies and their computing paradigm still lack embodied intelligence to efficiently interact with operational environments, respond with correct/expected actions, and adapt to changes in the environments. Toward this, recent advances in neuromorphic computing with Spiking Neural Networks (SNN) have demonstrated the potential to enable the embodied intelligence for robotics through bio-plausible computing paradigm that mimics how the biological brain works, known as "neuromorphic artificial intelligence (AI)". However, the field of neuromorphic AI-based robotics is still at an early stage, therefore its development and deployment for solving real-world problems expose new challenges in different design aspects, such as accuracy, adaptability, efficiency, reliability, and security. To address these challenges, this paper will discuss how we can enable embodied neuromorphic AI for robotic systems through our perspectives: (P1) Embodied intelligence based on effective learning rule, training mechanism, and adaptability; (P2) Cross-layer optimizations for energy-efficient neuromorphic computing; (P3) Representative and fair benchmarks; (P4) Low-cost reliability and safety enhancements; (P5) Security and privacy for neuromorphic computing; and (P6) A synergistic development for energy-efficient and robust neuromorphic-based robotics. Furthermore, this paper identifies research challenges and opportunities, as well as elaborates our vision for future research development toward embodied neuromorphic AI for robotics., Comment: To appear at the 18th International Conference on Control, Automation, Robotics and Vision (ICARCV), December 2024, Dubai, UAE

This study introduces the Quantum Federated Neural Network for Financial Fraud Detection (QFNN-FFD), a cutting-edge framework merging Quantum Machine Learning (QML) and quantum computing with Federated Learning (FL) for financial fraud detection. Using quantum technologies' computational power and the robust data privacy protections offered by FL, QFNN-FFD emerges as a secure and efficient method for identifying fraudulent transactions within the financial sector. Implementing a dual-phase training model across distributed clients enhances data integrity and enables superior performance metrics, achieving precision rates consistently above 95%. Additionally, QFNN-FFD demonstrates exceptional resilience by maintaining an impressive 80% accuracy, highlighting its robustness and readiness for real-world applications. This combination of high performance, security, and robustness against noise positions QFNN-FFD as a transformative advancement in financial technology solutions and establishes it as a new benchmark for privacy-focused fraud detection systems. This framework facilitates the broader adoption of secure, quantum-enhanced financial services and inspires future innovations that could use QML to tackle complex challenges in other areas requiring high confidentiality and accuracy.

In this study, we explore the innovative domain of Quantum Federated Learning (QFL) as a framework for training Quantum Machine Learning (QML) models via distributed networks. Conventional machine learning models frequently grapple with issues about data privacy and the exposure of sensitive information. Our proposed Federated Quantum Neural Network (FedQNN) framework emerges as a cutting-edge solution, integrating the singular characteristics of QML with the principles of classical federated learning. This work thoroughly investigates QFL, underscoring its capability to secure data handling in a distributed environment and facilitate cooperative learning without direct data sharing. Our research corroborates the concept through experiments across varied datasets, including genomics and healthcare, thereby validating the versatility and efficacy of our FedQNN framework. The results consistently exceed 86% accuracy across three distinct datasets, proving its suitability for conducting various QML tasks. Our research not only identifies the limitations of classical paradigms but also presents a novel framework to propel the field of QML into a new era of secure and collaborative innovation., Comment: Accepted for presentation at IJCNN 2024

Recent advancements in quantum computing have led to the development of hybrid quantum neural networks (HQNNs) that employ a mixed set of quantum layers and classical layers, such as Quanvolutional Neural Networks (QuNNs). While several works have shown security threats of classical neural networks, such as adversarial attacks, their impact on QuNNs is still relatively unexplored. This work tackles this problem by designing AdvQuNN, a specialized methodology to investigate the robustness of HQNNs like QuNNs against adversarial attacks. It employs different types of Ansatzes as parametrized quantum circuits and different types of adversarial attacks. This study aims to rigorously assess the influence of quantum circuit architecture on the resilience of QuNN models, which opens up new pathways for enhancing the robustness of QuNNs and advancing the field of quantum cybersecurity. Our results show that, compared to classical convolutional networks, QuNNs achieve up to 60\% higher robustness for the MNIST and 40\% for FMNIST datasets., Comment: 7 pages, 6 figures

In current noisy intermediate-scale quantum devices, hybrid quantum-classical neural networks (HQNNs) represent a promising solution that combines the strengths of classical machine learning with quantum computing capabilities. Compared to classical deep neural networks (DNNs), HQNNs present an additional set of hyperparameters, which are specific to quantum circuits. These quantum-specific hyperparameters, such as quantum circuit depth, number of qubits, type of entanglement, number of shots, and measurement observables, can significantly impact the behavior of the HQNNs and their capabilities to learn the given task. In this paper, we investigate the impact of these variations on different HQNN models for image classification tasks, implemented on the PennyLane framework. We aim to uncover intuitive and counter-intuitive learning patterns of HQNN models within granular levels of controlled quantum perturbations, to form a sound basis for their correlation to accuracy and training time. The outcome of our study opens new avenues for designing efficient HQNN algorithms and builds a foundational base for comprehending and identifying tunable hyperparameters of HQNN models that can lead to useful design implementation and usage., Comment: Accepted at the 3rd International Conference on Emergent Quantum Technologies (ICEQT'24), July 2024

Hybrid Quantum-Classical Machine Learning (ML) is an emerging field, amalgamating the strengths of both classical neural networks and quantum variational circuits on the current noisy intermediate-scale quantum devices. This paper performs an extensive comparative analysis between different hybrid quantum-classical machine learning algorithms, namely Quantum Convolution Neural Network, Quanvolutional Neural Network and Quantum ResNet, for image classification. The experiments designed in this paper focus on different Quantum ML (QML) algorithms to better understand the accuracy variation across the different quantum architectures by implementing interchangeable quantum circuit layers, varying the repetition of such layers and their efficient placement. Such variations enable us to compare the accuracy across different architectural permutations of a given hybrid QML algorithm. The performance comparison of the hybrid models, based on the accuracy, provides us with an understanding of hybrid quantum-classical convergence in correlation with the quantum layer count and the qubit count variations in the circuit., Comment: Accepted at the 3rd International Conference on Emergent Quantum Technologies (ICEQT'24), July 2024

The Continuous Learning (CL) paradigm consists of continuously evolving the parameters of the Deep Neural Network (DNN) model to progressively learn to perform new tasks without reducing the performance on previous tasks, i.e., avoiding the so-called catastrophic forgetting. However, the DNN parameter update in CL-based autonomous systems is extremely resource-hungry. The existing DNN accelerators cannot be directly employed in CL because they only support the execution of the forward propagation. Only a few prior architectures execute the backpropagation and weight update, but they lack the control and management for CL. Towards this, we design a hardware architecture, TinyCL, to perform CL on resource-constrained autonomous systems. It consists of a processing unit that executes both forward and backward propagation, and a control unit that manages memory-based CL workload. To minimize the memory accesses, the sliding window of the convolutional layer moves in a snake-like fashion. Moreover, the Multiply-and-Accumulate units can be reconfigured at runtime to execute different operations. As per our knowledge, our proposed TinyCL represents the first hardware accelerator that executes CL on autonomous systems. We synthesize the complete TinyCL architecture in a 65 nm CMOS technology node with the conventional ASIC design flow. It executes 1 epoch of training on a Conv + ReLU + Dense model on the CIFAR10 dataset in 1.76 s, while 1 training epoch of the same model using an Nvidia Tesla P100 GPU takes 103 s, thus achieving a 58x speedup, consuming 86 mW in a 4.74 mm2 die.

Quantum Computing (QC) claims to improve the efficiency of solving complex problems, compared to classical computing. When QC is integrated with Machine Learning (ML), it creates a Quantum Machine Learning (QML) system. This paper aims to provide a thorough understanding of the foundational concepts of QC and its notable advantages over classical computing. Following this, we delve into the key aspects of QML in a detailed and comprehensive manner. In this survey, we investigate a variety of QML algorithms, discussing their applicability across different domains. We examine quantum datasets, highlighting their unique characteristics and advantages. The survey also covers the current state of hardware technologies, providing insights into the latest advancements and their implications for QML. Additionally, we review the software tools and simulators available for QML development, discussing their features and usability. Furthermore, we explore practical applications of QML, illustrating how it can be leveraged to solve real-world problems more efficiently than classical ML methods. This paper serves as a valuable resource for readers seeking to understand the current state-of-the-art techniques in the QML field, offering a solid foundation to embark on further exploration and development in this rapidly evolving area.

Machine learning (ML) is widely used today, especially through deep neural networks (DNNs), however, increasing computational load and resource requirements have led to cloud-based solutions. To address this problem, a new generation of networks called Spiking Neural Networks (SNN) has emerged, which mimic the behavior of the human brain to improve efficiency and reduce energy consumption. These networks often process large amounts of sensitive information, such as confidential data, and thus privacy issues arise. Homomorphic encryption (HE) offers a solution, allowing calculations to be performed on encrypted data without decrypting it. This research compares traditional DNNs and SNNs using the Brakerski/Fan-Vercauteren (BFV) encryption scheme. The LeNet-5 model, a widely-used convolutional architecture, is used for both DNN and SNN models based on the LeNet-5 architecture, and the networks are trained and compared using the FashionMNIST dataset. The results show that SNNs using HE achieve up to 40% higher accuracy than DNNs for low values of the plaintext modulus t, although their execution time is longer due to their time-coding nature with multiple time-steps.

Pretrained language models (PLMs) are today the primary model for natural language processing. Despite their impressive downstream performance, it can be difficult to apply PLMs to new languages, a barrier to making their capabilities universally accessible. While prior work has shown it possible to address this issue by learning a new embedding layer for the new language, doing so is both data and compute inefficient. We propose to use an active forgetting mechanism during pretraining, as a simple way of creating PLMs that can quickly adapt to new languages. Concretely, by resetting the embedding layer every K updates during pretraining, we encourage the PLM to improve its ability of learning new embeddings within a limited number of updates, similar to a meta-learning effect. Experiments with RoBERTa show that models pretrained with our forgetting mechanism not only demonstrate faster convergence during language adaptation but also outperform standard ones in a low-data regime, particularly for languages that are distant from English., Comment: NeurIPS 2023 Final Version

Transformers' compute-intensive operations pose enormous challenges for their deployment in resource-constrained EdgeAI / tinyML devices. As an established neural network compression technique, quantization reduces the hardware computational and memory resources. In particular, fixed-point quantization is desirable to ease the computations using lightweight blocks, like adders and multipliers, of the underlying hardware. However, deploying fully-quantized Transformers on existing general-purpose hardware, generic AI accelerators, or specialized architectures for Transformers with floating-point units might be infeasible and/or inefficient. Towards this, we propose SwiftTron, an efficient specialized hardware accelerator designed for Quantized Transformers. SwiftTron supports the execution of different types of Transformers' operations (like Attention, Softmax, GELU, and Layer Normalization) and accounts for diverse scaling factors to perform correct computations. We synthesize the complete SwiftTron architecture in a $65$ nm CMOS technology with the ASIC design flow. Our Accelerator executes the RoBERTa-base model in 1.83 ns, while consuming 33.64 mW power, and occupying an area of 273 mm^2. To ease the reproducibility, the RTL of our SwiftTron architecture is released at https://github.com/albertomarchisio/SwiftTron., Comment: To appear at the 2023 International Joint Conference on Neural Networks (IJCNN), Queensland, Australia, June 2023

Capsule Networks (CapsNets) are able to hierarchically preserve the pose relationships between multiple objects for image classification tasks. Other than achieving high accuracy, another relevant factor in deploying CapsNets in safety-critical applications is the robustness against input transformations and malicious adversarial attacks. In this paper, we systematically analyze and evaluate different factors affecting the robustness of CapsNets, compared to traditional Convolutional Neural Networks (CNNs). Towards a comprehensive comparison, we test two CapsNet models and two CNN models on the MNIST, GTSRB, and CIFAR10 datasets, as well as on the affine-transformed versions of such datasets. With a thorough analysis, we show which properties of these architectures better contribute to increasing the robustness and their limitations. Overall, CapsNets achieve better robustness against adversarial examples and affine transformations, compared to a traditional CNN with a similar number of parameters. Similar conclusions have been derived for deeper versions of CapsNets and CNNs. Moreover, our results unleash a key finding that the dynamic routing does not contribute much to improving the CapsNets' robustness. Indeed, the main generalization contribution is due to the hierarchical feature learning through capsules., Comment: To appear at the 2023 International Joint Conference on Neural Networks (IJCNN), Queensland, Australia, June 2023

While the way intermediate representations are generated in encoder-decoder sequence-to-sequence models typically allow them to preserve the semantics of the input sentence, input features such as formality might be left out. On the other hand, downstream tasks such as translation would benefit from working with a sentence representation that preserves formality in addition to semantics, so as to generate sentences with the appropriate level of social formality -- the difference between speaking to a friend versus speaking with a supervisor. We propose a sequence-to-sequence method for learning a formality-aware representation for Japanese sentences, where sentence generation is conditioned on both the original representation of the input sentence, and a side constraint which guides the sentence representation towards preserving formality information. Additionally, we propose augmenting the sentence representation with a learned representation of formality which facilitates the extraction of formality in downstream tasks. We address the lack of formality-annotated parallel data by adapting previous works on procedural formality classification of Japanese sentences. Experimental results suggest that our techniques not only helps the decoder recover the formality of the input sentence, but also slightly improves the preservation of input sentence semantics.

Prior work shows that it is possible to expand pretrained Masked Language Models (MLMs) to new languages by learning a new set of embeddings, while keeping the transformer body frozen. Despite learning a small subset of parameters, this approach is not compute-efficient, as training the new embeddings requires a full forward and backward pass over the entire model. We propose mini-model adaptation, a compute-efficient alternative that builds a shallow mini-model from a fraction of a large model's parameters. New language-specific embeddings can then be efficiently trained over the mini-model and plugged into the aligned large model for rapid cross-lingual transfer. We explore two approaches to learn mini-models: MiniJoint, which jointly pretrains the primary model and the mini-model using a single transformer with a secondary MLM head at a middle layer; and MiniPost, where we start from a regular pretrained model, build a mini-model by extracting and freezing a few layers, and learn a small number of parameters on top. Experiments on XNLI, MLQA and PAWS-X show that mini-model adaptation matches the performance of the standard approach using 2.3x less compute on average., Comment: Findings of ACL 2023 Camera Ready