Top Research Papers on Embedded Systems

An account for determining when an AI system can be said to embody certain values is proposed, which understands embodied values as the result of design activities intended to embed those values in such systems.

Deep learning inference on embedded devices is a burgeoning field with myriad applications because tiny embedded devices are omnipresent. But we must overcome major challenges before we can benefit from this opportunity. Embedded processors are severely resource constrained. Their nearest mobile counterparts exhibit at least a 100 -- 1,000x difference in compute capability, memory availability, and power consumption. As a result, the machine-learning (ML) models and associated ML inference framework must not only execute efficiently but also operate in a few kilobytes of memory. Also, the embe...

A lightweight object detection model, which is developed based on Mobilenet-v2, which achieves up to 75.9% mAP in the VOC dataset and can be applied in embedded systems with limited computational resources.

This study shows the feasibility of deployment of the customized model on cheap and power-efficient embedded hardware without compromising mean average detection accuracy (83.64%) and achieved frame rate up to 30 fps even for the difficult scenarios such as overlapping apples, complex background, less exposure of apple due to leaves and branches.

An overview of TSN in automotive applications is provided and the recent technological developments relevant to the adoption of T SNP are discussed, which points at the open challenges and future research directions.

This work focuses on the induced point dipole formulation of polarizable QM/MM approaches and shows how efficient and linear scaling implementations have allowed their application to the modeling of complex biosystems, including Born-Oppenheimer dynamics, enhanced sampling techniques and nonadiabatic descriptions.

This article comprehensively reviews the emerging concept of Internet of Intelligent Things (IoIT), adopting an integrated perspective centred on the areas of embedded systems, edge computing, and machine learning. With rapid developments in these areas, new solutions are emerging to address previously unsolved problems, demanding novel research and development paradigms. In this sense, this article aims to fulfil some important research gaps, laying down the foundations for cutting-edge research works following an ever-increasing trend based on embedded devices powered by compressed artificia...

Renewable energy is projected to play an important role in reducing greenhouse gas emissions and in realising the climate change goals. Large scale development of variable renewable energy, which is regarded as non-dispatchable, requires additional power system quality services such as voltage regulation, frequency regulation and inertial response. Energy storage provides an important means to supply these services but there are many uncertainties in terms of technology, market readiness, economics, and regulatory requirements. The aim of this study is to undertake a global state-of-the-art re...

Some of the major problems existing today in AI-powered embedded systems are presented, highlighting possible solutions and research directions to support them, increasing their security, safety, and time predictability.

In this survey, existing contextual embedding models, cross-lingual polyglot pre-training, the application of contextual embeddings in downstream tasks, model compression, and model analyses are reviewed.

Overall, EHR-based predictive models offer promising results for improving clinical outcomes, although several gaps in the literature remain, and most study designs were observational.

A survey of dynamic network embedding inspects the data model, representation learning technique, evaluation and application of current related works and derives common patterns from them and builds a taxonomy that refines the category hierarchy by typical learning models.

This paper proposes a random walk and word embedding based ontology embedding method named OWL2Vec*, which encodes the semantics of an OWL ontology by taking into account its graph structure, lexical information and logical constructors.

This work develops three classes of attacks to systematically study information that might be leaked by embeddings, and extensively evaluates the attacks on various state-of-the-art embedding models in the text domain.

This paper proposes Automated Concatenation of Embeddings (ACE) to automate the process of finding better concatenations of embeddings for structured prediction tasks, based on a formulation inspired by recent progress on neural architecture search.

This work presents a simple yet effective unsupervised network embedding method for attributed multiplex network called DMGI, inspired by Deep Graph Infomax (DGI) that maximizes the mutual information between local patches of a graph, and the global representation of the entire graph.

This paper proposes a cross-batch memory (XBM) mechanism that memorizes the embeddings of past iterations, allowing the model to collect sufficient hard negative pairs across multiple mini-batches - even over the whole dataset.

The unified embedding framework developed to model semantic embeddings for personalized search, and the system to serve embedding-based retrieval in a typical search system based on an inverted index are introduced.

Dr Donald Bailey starts with introductory material considering the problem of embedded image processing, and how some of the issues may be solved using parallel hardware solutions. Field programmable gate arrays (FPGAs) are introduced as a technology that provides flexible, fine-grained hardware that can readily exploit parallelism within many image processing algorithms. A brief review of FPGA programming languages provides the link between a software mindset normally associated with image processing algorithms, and the hardware mindset required for efficient utilization of a parallel hardwar...

The proposed method is out of the self-expressive framework, scales to the sample size linearly, and is applicable to arbitrarily large datasets and online clustering scenarios, and the clustering accuracy is much higher than its competitors.

While BERT is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning (Reimers and Gurevych, 2019), BERT based cross-lingual sentence embeddings have yet to be explored. We systematically investigate methods for learning multilingual sentence embeddings by combining the best methods for learning monolingual and cross-lingual representations including: masked language modeling (MLM), translation language modeling (TLM) (Conneau and Lample, 2019), dual encoder translation ranking (Guo et al., 2018), and additive margin softma...

A group of AI engineers, ethicists and social scientists suggest embedding ethicists into the development team as one way of improving the consideration of ethical issues during AI development.

This work proposes a graph neural network-based representation learning framework for heterogeneous hypergraphs, an extension of conventional graphs, which can well characterize multiple non-pairwise relations and shows that relationships beyond pairwise are also advantageous in the spammer detection.

This analysis provides a comprehensive comparison of embedding-based LP methods, extending the dimensions of analysis beyond what is commonly available in the literature.

The Contextualized Language and Knowledge Embedding (CoLAKE) is proposed, which jointly learns contextualized representation for both language and knowledge with the extended MLM objective, and achieves surprisingly high performance on a synthetic task called word-knowledge graph completion, which shows the superiority of simultaneously contextualizing language andknowledge representation.

It is shown that introducing a pre-trained multilingual language model dramatically reduces the amount of parallel training data required to achieve good performance by 80%, and a model that achieves 83.7% bi-text retrieval accuracy over 112 languages on Tatoeba is released.

Through the benchmarking of 33 models on MTEB, it is found that no particular text embedding method dominates across all tasks, suggesting that the field has yet to converge on a universal text embeddedding method and scale it up sufficiently to provide state-of-theart results on all embedding tasks.

This paper proposes two novel deep neural network models for sarcasm detection, namely ACE 1 and ACE 2, which extend the architecture of BERT by incorporating both affective and contextual features.

It is argued that deterministic functions are not sufficiently powerful to capture one-to-many correspondences and proposed Probabilistic Cross-Modal Embedding (PCME), where samples from the different modalities are represented as probabilistic distributions in the common embedding space.

The main goal of graph embedding methods is to pack every node's properties into a vector with a smaller dimension, hence, node similarity in the original complex irregular spaces can be easily quantified in the embedded vector spaces using standard metrics.

Language Embedded Radiance Fields (LERFs) are proposed, a method for grounding language embeddings from off-the-shelf models like CLIP into NeRF, which enable these types of open-ended language queries in 3D.

This survey conducts an in-depth survey on important compression and acceleration techniques that help adapt deep learning models to mobile and embedded devices, which are specifically classify as pruning, quantization, model distillation, network design strategies, and low-rank factorization.

The core of DualE lies a specific design of dual-quaternion-based multiplication, which universally models relations as the compositions of a series of translation and rotation operations.

Computer vision is a key technique to make agricultural machinery smart. Deep neural network has achieved great success in computer vision. How to use it at a small size, low cost, low power consumption device with high accuracy and speed on strawberry harvesting machinery has drawn much research attention. Since the infield situation has reduced number of objects and that they are easier to be distinguished from the background compared to other computer vision datasets, the huge neural network structure can be simplified in order to speed up the detection inference without penalizing the dete...

The continuous increase of greenhouse gas emission and the growing cost of fossil fuels are two motive forces to utilize energy sources more effectively. The agricultural greenhouses are developed to supply the food resources safely and with the higher quality, especially in off-season periods of year. In developing countries like Iran, the fuel usage of the greenhouse is very high, and there is a large margin for improving heating systems. The waste heat recovery system (HRS) in two forms of with and without embedded phase change material (PCM) added to the existing heating system, and their ...

Position encoding recently has shown effective in the transformer architecture. It enables valuable supervision for dependency modeling between elements at different positions of the sequence. In this paper, we first investigate various methods to integrate positional information into the learning process of transformer-based language models. Then, we propose a novel method named Rotary Position Embedding(RoPE) to effectively leverage the positional information. Specifically, the proposed RoPE encodes the absolute position with a rotation matrix and meanwhile incorporates the explicit relative...

This paper surveys how the NLP field has evolved from rule-based, statistical to more context-sensitive learned representations and covers the history of text representations from the 1970s and onwards, from regular expressions to the latest vector representations used to encode the raw text data.

A straightforward, sterically less demanding synthetic approach involving hybridization between two [6]helicene subunits and a perylene diimide (PDI) scaffold is presented, affording peryleneDiimide-embedded double [8]he Alicenes (PD8Hs) which represent the highest double carbohelicenes reported thus far.

The proposed hybrid GZSL framework with contrastive embedding, named CE-GZSL, is evaluated, and the results show that the CEGZSL method can outperform the state-of-the-arts by a significant margin on three datasets.

This paper proves that a suitable ESN, trained on a series of measurements of an invertible dynamical system, induces a C1 map from the dynamicalSystem's phase space to the ESN's reservoir space, and proves that the Echo State Map is generically an embedding with positive probability.