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.

Direct collocation methods are powerful tools to solve trajectory optimization problems in robotics. While their resulting trajectories tend to be dynamically accurate, they may also present large kinematic errors in the case of constrained mechanical systems, i.e., those whose state coordinates are subject to holonomic or nonholonomic constraints, like loop-closure or rolling-contact constraints. These constraints confine the robot trajectories to an implicitly-defined manifold, which complicates the computation of accurate solutions. Discretization errors inherent to the transcription of the problem easily make the trajectories drift away from this manifold, which results in physically inconsistent motions that are difficult to track with a controller. This paper reviews existing methods to deal with this problem and proposes new ones to overcome their limitations. Current approaches either disregard the kinematic constraints (which leads to drift accumulation) or modify the system dynamics to keep the trajectory close to the manifold (which adds artificial forces or energy dissipation to the system). The methods we propose, in contrast, achieve full drift elimination on the discrete trajectory, or even along the continuous one, without artificial modifications of the system dynamics. We illustrate and compare the methods using various examples of different complexity.

Coronavirus disease 2019 (COVID-19) has spread all over the world for three years, but medical facilities in many areas still aren't adequate. There is a need for rapid COVID-19 diagnosis to identify high-risk patients and maximize the use of limited medical resources. Motivated by this fact, we proposed the deep learning framework MEDNC for automatic prediction and diagnosis of COVID-19 using computed tomography (CT) images. Our model was trained using two publicly available sets of COVID-19 data. And it was built with the inspiration of transfer learning. Results indicated that the MEDNC greatly enhanced the detection of COVID-19 infections, reaching an accuracy of 98.79% and 99.82% respectively. We tested MEDNC on a brain tumor and a blood cell dataset to show that our model applies to a wide range of problems. The outcomes demonstrated that our proposed models attained an accuracy of 99.39% and 99.28%, respectively. This COVID-19 recognition tool could help optimize healthcare resources and reduce clinicians' workload when screening for the virus.

We describe our contribution to the SemEVAl 2023 AfriSenti-SemEval shared task, where we tackle the task of sentiment analysis in 14 different African languages. We develop both monolingual and multilingual models under a full supervised setting (subtasks A and B). We also develop models for the zero-shot setting (subtask C). Our approach involves experimenting with transfer learning using six language models, including further pertaining of some of these models as well as a final finetuning stage. Our best performing models achieve an F1-score of 70.36 on development data and an F1-score of 66.13 on test data. Unsurprisingly, our results demonstrate the effectiveness of transfer learning and fine-tuning techniques for sentiment analysis across multiple languages. Our approach can be applied to other sentiment analysis tasks in different languages and domains.

Multi-agent Reinforcement learning (MARL) studies the behaviour of multiple learning agents that coexist in a shared environment. MARL is more challenging than single-agent RL because it involves more complex learning dynamics: the observations and rewards of each agent are functions of all other agents. In the context of MARL, Real-Time-Strategy (RTS) games represent very challenging environments where multiple players interact simultaneously and control many units of different natures all at once. In fact, RTS games are so challenging for the current RL methods, that just being able to tackle them with RL is interesting. This project provides the end-to-end experience of applying RL in the Lux AI v2 Kaggle competition, where competitors design agents to control variable-sized fleets of units and tackle a multi-variable optimization, resource gathering, and allocation problem in a 1v1 scenario against other competitors. We use a centralized approach for training the RL agents, and report multiple design decisions along the process. We provide the source code of the project: https://github.com/

Despite these WFSA algorithms We refer readers to [8, 9] for a comprehensive introduction to being well-known in the literature, new challenges finite state automata and related algorithms. A weighted finite arise from performance characteristics of modern architectures, state automaton (WFSA) A = (Σ, Q, i, f, E) over a semiring and from nuances in automatic differentiation. We describe (K,,, 0, 1) is specified by a finite alphabet Σ, a finite set a suite of generally applicable techniques employed in of states Q, an initial state i Q, a final state f, and a finite LAST to address these challenges, and demonstrate their effectiveness set of arcs E Q (Σ {ǫ}) K Q (ǫ denotes the with benchmarks on TPUv3 and V100 GPU.

We propose an experimental scheme for performing sensitive, high-precision laser spectroscopy studies on fast exotic isotopes. By inducing a step-wise resonant ionization of the atoms travelling inside an electric field and subsequently detecting the ion and the corresponding electron, time- and position-sensitive measurements of the resulting particles can be performed. Using a Mixture Density Network (MDN), we can leverage this information to predict the initial energy of individual atoms and thus apply a Doppler correction of the observed transition frequencies on an event-by-event basis. We conduct numerical simulations of the proposed experimental scheme and show that kHz-level uncertainties can be achieved for ion beams produced at extreme temperatures ($> 10^8$ K), with energy spreads as large as $10$ keV and non-uniform velocity distributions. The ability to perform in-flight spectroscopy, directly on highly energetic beams, offers unique opportunities to studying short-lived isotopes with lifetimes in the millisecond range and below, produced in low quantities, in hot and highly contaminated environments, without the need for cooling techniques. Such species are of marked interest for nuclear structure, astrophysics, and new physics searches.

The paper considers the DIverse MultiPLEx (DIMPLE) network model, introduced in Pensky and Wang (2021), where all layers of the network have the same collection of nodes and are equipped with the Stochastic Block Models. In addition, all layers can be partitioned into groups with the same community structures, although the layers in the same group may have different matrices of block connection probabilities. The DIMPLE model generalizes a multitude of papers that study multilayer networks with the same community structures in all layers, as well as the Mixture Multilayer Stochastic Block Model (MMLSBM), where the layers in the same group have identical matrices of block connection probabilities. While Pensky and Wang (2021) applied spectral clustering to the proxy of the adjacency tensor, the present paper uses Sparse Subspace Clustering (SSC) for identifying groups of layers with identical community structures. Under mild conditions, the latter leads to the strongly consistent between-layer clustering. In addition, SSC allows to handle much larger networks than methodology of Pensky and Wang (2021), and is perfectly suitable for application of parallel computing.

We propose a security verification framework for cryptographic protocols using machine learning. In recent years, as cryptographic protocols have become more complex, research on automatic verification techniques has been focused on. The main technique is formal verification. However, the formal verification has two problems: it requires a large amount of computational time and does not guarantee decidability. We propose a method that allows security verification with computational time on the order of linear with respect to the size of the protocol using machine learning. In training machine learning models for security verification of cryptographic protocols, a sufficient amount of data, i.e., a set of protocol data with security labels, is difficult to collect from academic papers and other sources. To overcome this issue, we propose a way to create arbitrarily large datasets by automatically generating random protocols and assigning security labels to them using formal verification tools. Furthermore, to exploit structural features of protocols, we construct a neural network that processes a protocol along its series and tree structures. We evaluate the proposed method by applying it to verification of practical cryptographic protocols.

The study aimed to evaluate the regression models' performance in predicting the cost of medical insurance. The Three (3) Regression Models in Machine Learning namely Linear Regression, Gradient Boosting, and Support Vector Machine were used. The performance will be evaluated using the metrics RMSE (Root Mean Square), r2 (R Square), and K-Fold Cross-validation. The study also sought to pinpoint the feature that would be most important in predicting the cost of medical insurance.The study is anchored on the knowledge discovery in databases (KDD) process. (KDD) process refers to the overall process of discovering useful knowledge from data. It show the performance evaluation results reveal that among the three (3) Regression models, Gradient boosting received the highest r2 (R Square) 0.892 and the lowest RMSE (Root Mean Square) 1336.594. Furthermore, the 10-Fold Cross-validation weighted mean findings are not significantly different from the r2 (R Square) results of the three (3) regression models. In addition, Exploratory Data Analysis (EDA) using a box plot of descriptive statistics observed that in the charges and smoker features the median of one group lies outside of the box of the other group, so there is a difference between the two groups. It concludes that Gradient boosting appears to perform better among the three (3) regression models. K-Fold Cross-Validation concluded that the three (3) regression models are good. Moreover, Exploratory Data Analysis (EDA) using a box plot of descriptive statistics ceases that the highest charges are due to the smoker feature.