Learning-based Solution Research Articles

A Primer on Seq2Seq Models for Generative Chatbots

The recent spread of Deep Learning-based solutions for Artificial Intelligence and the development of Large Language Models has pushed forwards significantly the Natural Language Processing area. The approach has quickly evolved in the last ten years, deeply affecting NLP, from low-level text pre-processing tasks –such as tokenisation or POS tagging– to high-level, complex NLP applications like machine translation and chatbots. This article examines recent trends in the development of open-domain data-driven generative chatbots, focusing on the Seq2Seq architectures. Such architectures are compatible with multiple learning approaches, ranging from supervised to reinforcement and, in the last years, allowed to realise very engaging open-domain chatbots. Not only do these architectures allow to directly output the next turn in a conversation but, to some extent, they also allow to control the style or content of the response. To offer a complete view on the subject, we examine possible architecture implementations as well as training and evaluation approaches. Additionally, we provide information about the openly available corpora to train and evaluate such models and about the current and past chatbot competitions. Finally, we present some insights on possible future directions, given the current research status.

Geomechanical reservoir modelling with Thermodynamics-based Artificial Neural Networks (TANNs)

Geological subsurface storage is a promising strategy for large-scale, cost-efficient energy storage systems. Stress sensitivity has notably influence on the long-term stability and serviceability of subsurface reservoirs set for energy storage use. The evolution of the field stress of stress-sensitive reservoirs and the associated structural deformation, often a consequence of fluid production and injection, can cause significant changes to the reservoir stress-dependent properties [1]. These properties include pore and void compressibility, and porosity. For fractured and faulted reservoirs, field stress variation may also impact fracture conductivity, alter pre-existing fractures, and reactivate faults [4].
 Current geomechanical reservoir simulators incorporate the evolution of the reservoir mechanical state by means of material constitutive modelling, typically within a Finite Element Modelling (FEM) framework. In efforts to honour the heterogeneity and multi-scale nature of subsurface reservoirs, multi-scale solution methods are commonly adopted. For a reservoir multi-scale simulation, the micro-scale problem must be iteratively solved for different input parameters, render-ing the solution method computationally exhaustive, particularly for two-way coupled problems.
 The combination of machine learning-based solution methods and FEM frameworks can address the computational inefficiency of conventional multi-scale reservoir modelling schemes. In principle, a machine learning algorithm trained to learn the constitutive behaviour of a material can replace in-built material constitutive models in a FEM framework. We resort to Thermodynamics-based Artificial Neural Networks (TANNs), a physics-based data-driven machine learning algorithm introduced in [5] for material constitutive modelling. TANNs have been shown to guarantee the thermodynamic consistency of learnt material constitutive models. The first and second laws of thermodynamics are directly encoded in the architecture of TANNs through the definition of two scalar functions, an energy potential and a dissipation function, and the computation of their differentials [3].
 TANNs can be incorporated in FEM tools, in what is referred to as TANNxFEM in [6]. We present the application of the TANNxFEM framework to geomechanical reservoir modelling. This work aims to introduce a computationally efficient reservoir modelling framework, for which uncertainty quantification is possible. The proposed framework can maximise the use of information inherently contained in high-dimensional data, and significantly reduce computation demands necessary for accurate statistical evaluation, while warranting physical and geological realism.
 TANNs are first trained on analytical material data, computed by numerical integration of an incremental, thermodynamically consistent material model presented in [2]. The trained TANNs are then imported into a user material subroutine for the finite element package Abaqus. Through an input file, the parameters of the trained neural network are read as a set of material properties. The material subroutine then evaluates the incorporated trained neural network to construct the stress and elasticity tensors necessary for large-scale finite element reservoir simulation. The above proposed framework is validated against a large-scale finite element simulation with a user material subroutine implementing the constitutive model used to first generate the training data for TANNs. The free-energy, dissipation, and the stress-strain response of the two finite element simulations are compared for model verification.

Detecting Malicious Accounts in Online Developer Communities Using Deep Learning

Online developer communities like GitHub allow a massive number of developers to collaborate. However, the openness of the communities makes them vulnerable to different types of malicious attacks, since attackers can easily join these communities and interact with legitimate users. In this work, we propose GitSec, a deep learning-based solution for detecting malicious accounts in online developer communities. GitSec distinguishes malicious accounts from legitimate ones based on the account profiles, dynamic activity characteristics, as well as social interactions. First, GitSec introduces two user activity sequences and applies a parallel neural network design with an attention mechanism to process the sequences. Second, GitSec constructs two graphs to represent the interactions between users according to their repository operations. Especially, graph neural networks and structural hole theory are employed to deal with the two constructed graphs. Third, GitSec makes use of the descriptive features to enhance the detection performance. The final judgement is made by a decision maker implemented by a supervised machine learning-based classifier. Based on the real-world data of GitHub users, our comprehensive evaluations show that GitSec achieves a better performance than state-of-the-art solutions, with an AUC value of 0.916.

A review of the machine learning datasets in mammography, their adherence to the FAIR principles and the outlook for the future

The increasing rates of breast cancer, particularly in emerging economies, have led to interest in scalable deep learning-based solutions that improve the accuracy and cost-effectiveness of mammographic screening. However, such tools require large volumes of high-quality training data, which can be challenging to obtain. This paper combines the experience of an AI startup with an analysis of the FAIR principles of the eight available datasets. It demonstrates that the datasets vary considerably, particularly in their interoperability, as each dataset is skewed towards a particular clinical use-case. Additionally, the mix of digital captures and scanned film compounds the problem of variability, along with differences in licensing terms, ease of access, labelling reliability, and file formats. Improving interoperability through adherence to standards such as the BIRADS criteria for labelling and annotation, and a consistent file format, could markedly improve access and use of larger amounts of standardized data. This, in turn, could be increased further by GAN-based synthetic data generation, paving the way towards better health outcomes for breast cancer.

Pothole recognition using convolution neural networks and transfer learning

Potholes have been and still are a huge problem for every walk of life. There are many deaths and accidents reported daily due to that very problem. For that reason, pothole recognition comes into the picture. To maintain and preserve a road, it is vital to detect potholes. It also helps in the prevention of accidents. Roads play an important part in day-to-day transportation for every person around the world. But the quality of roads decreases drastically due to the way of usage and aging. The existing methods take much time and manpower to repair the damaged areas. The entire process is slowing down just because an expert team is checking whether there is a pothole at the reported location or not. So, if we automate the process of detection of potholes from a set of images reported from a particular location and appropriately alerting the authorities with the amount of damage, the process speeds up exponentially. We must solve the major problem of pothole recognition by using machine learning algorithms. This paper will discuss a convolution neural network-based and a transfer learning-based solution for pothole recognition.

Champion-level drone racing using deep reinforcement learning

First-person view (FPV) drone racing is a televised sport in which professional competitors pilot high-speed aircraft through a 3D circuit. Each pilot sees the environment from the perspective of their drone by means of video streamed from an onboard camera. Reaching the level of professional pilots with an autonomous drone is challenging because the robot needs to fly at its physical limits while estimating its speed and location in the circuit exclusively from onboard sensors1. Here we introduce Swift, an autonomous system that can race physical vehicles at the level of the human world champions. The system combines deep reinforcement learning (RL) in simulation with data collected in the physical world. Swift competed against three human champions, including the world champions of two international leagues, in real-world head-to-head races. Swift won several races against each of the human champions and demonstrated the fastest recorded race time. This work represents a milestone for mobile robotics and machine intelligence2, which may inspire the deployment of hybrid learning-based solutions in other physical systems.

Deep Learning-based Human Pose Estimation: A Survey

Human pose estimation aims to locate the human body parts and build human body representation (e.g., body skeleton) from input data such as images and videos. It has drawn increasing attention during the past decade and has been utilized in a wide range of applications including human-computer interaction, motion analysis, augmented reality, and virtual reality. Although the recently developed deep learning-based solutions have achieved high performance in human pose estimation, there still remain challenges due to insufficient training data, depth ambiguities, and occlusion. The goal of this survey article is to provide a comprehensive review of recent deep learning-based solutions for both 2D and 3D pose estimation via a systematic analysis and comparison of these solutions based on their input data and inference procedures. More than 260 research papers since 2014 are covered in this survey. Furthermore, 2D and 3D human pose estimation datasets and evaluation metrics are included. Quantitative performance comparisons of the reviewed methods on popular datasets are summarized and discussed. Finally, the challenges involved, applications, and future research directions are concluded. A regularly updated project page is provided: https://github.com/zczcwh/DL-HPE .

A deep learning-based solution for digitization of invoice images with automatic invoice generation and labelling

A deep learning-based solution for digitization of invoice images with automatic invoice generation and labelling

Attentional Composition Networks for Long-Tailed Human Action Recognition

The problem of long-tailed visual recognition has been receiving increasing research attention. However, the long-tailed distribution problem remains underexplored for video-based visual recognition. To address this issue, in this article we propose a compositional learning based solution for video-based human action recognition. Our method, named Attentional Composition Networks (ACN), first learns verb-like and preposition-like components, then shuffles these components to generate samples for the tail classes in the feature space to augment the data for the tail classes. Specifically, during training, we represent each action video by a graph that captures the spatial-temporal relations (edges) among detected human/object instances (nodes). Then, ACN utilizes the position information to decompose each action into a set of verb and preposition representations using the edge features in the graph. After that, the verb and preposition features from different videos are combined via an attention structure to synthesize feature representations for tail classes. This way, we can enrich the data for the tail classes and consequently improve the action recognition for these classes. To evaluate the compositional human action recognition, we further contribute a new human action recognition dataset, namely NEU-Interaction (NEU-I). Experimental results on both Something-Something V2 and the proposed NEU-I demonstrate the effectiveness of the proposed method for long-tailed, few-shot, and zero-shot problems in human action recognition. Source code and the NEU-I dataset are available at https://github.com/YajieW99/ACN .

Deep learning-based scene understanding for autonomous robots: a survey

Autonomous robots are a hot research subject within the fields of science and technology, which has a big impact on social-economic development. The ability of the autonomous robot to perceive and understand its working environment is the basis for solving more complicated issues. In recent years, an increasing number of artificial intelligence-based methods have been proposed in the field of scene understanding for autonomous robots, and deep learning is one of the current key areas in this field. Outstanding gains have been attained in the field of scene understanding for autonomous robots based on deep learning. Thus, this paper presents a review of recent research on the deep learning-based scene understanding for autonomous robots. This survey provides a detailed overview of the evolution of robotic scene understanding and summarizes the applications of deep learning methods in scene understanding for autonomous robots. In addition, the key issues in autonomous robot scene understanding are analyzed, such as pose estimation, saliency prediction, semantic segmentation, and object detection. Then, some representative deep learning-based solutions for these issues are summarized. Finally, future challenges in the field of the scene understanding for autonomous robots are discussed.

An efficient deep learning-based solution for network intrusion detection in wireless sensor network

An efficient deep learning-based solution for network intrusion detection in wireless sensor network

From one field to another—Unsupervised domain adaptation for semantic segmentation in agricultural robotics

In traditional arable crop fields, tractors treat the whole field uniformly applying large quantities of herbicides and pesticides for weed control and plant protection. Autonomous robots, instead, offer the potential to provide a per-plant treatment, thus turning weed control and plant protection environment-friendly. To this end, an autonomous robot has to reliably distinguish crops, weeds, and soil under a diverse range of environmental conditions using its onboard sensors. Such recognition ability forms the basis for targeted plant-specific treatments in the form of spot applications. Basically, all such perception systems used today rely on some form of machine learning technique. However, current learning-based solutions often show a performance decay when applied under new field conditions. This is a major bottleneck for real-world application and finally commercial adoption. In this paper, we propose a simple yet effective approach to unsupervised domain adaptation for semantic segmentation systems so that an existing segmentation pipeline can be adapted to different fields, different robots, and different crops. Our system yields a high segmentation performance in new target fields without the need for extra manual annotations. It exploits only annotations from the source domain, i.e., the original field used for training the robot’s vision system. Our thorough evaluation shows that our approach achieves high accuracy when transferring an existing segmentation system to different environmental conditions, different plant species, and different robotic systems.

Machine learning-based solutions for resource management in fog computing

Machine learning-based solutions for resource management in fog computing

Simplified review on cyber security threats detection in IoT environment using deep learning approach

Wearable technology, sensor networks, and home utilities are just a few of the businesses where the Internet of Things (IoT) is spreading quickly. With the development of the IoT, billions of gadgets are now connected to the internet and exchanging data. The proliferation of IoT devices has increased the number of IoT-based cyberattacks. In 2016 a massive denial of service (DDOS) cyber-attack was lunched utilizing infected internet of things devices a major website including Netflix and CNN was shutdown. Therefore, new ways for recognizing threats posed by hacked IoT nodes must be developed to overcome this concern. In that same context, ML and DL approaches are the best appropriate investigative control solution against IoT device-based intrusions. The point of the study is to offer a complete grasp of the IoT system-relevant technologies, standards, architecture, and the increasing dangers from corrupted IoT gadgets and an introduction to intrusion detection systems. Additionally, this research focuses on deep learning-based solutions for identifying IoT devices susceptible to cyber-attacks. The detection rate provided by deep learning algorithms shows promising results which reached 99% detection accuracy in some cases.

AutoSteer: Learned Query Optimization for Any SQL Database

This paper presents AutoSteer, a learning-based solution that automatically drives query optimization in any SQL database that exposes tunable optimizer knobs. AutoSteer builds on the Bandit optimizer (Bao) and extends it with new capabilities (e.g., automated hint-set discovery) to minimize integration effort and facilitate usability in both monolithic and disaggregated SQL systems. We successfully applied AutoSteer on PostgreSQL, PrestoDB, Spark-SQL, MySQL, and DuckDB - five popular open-source database engines with diverse query optimizers. We then conducted a detailed experimental evaluation with public benchmarks (JOB, Stackoverflow, TPC-DS) and a production workload from Meta's PrestoDB deployments. Our evaluation shows that AutoSteer can not only outperform these engines' native query optimizers (e.g., up to 40% improvements for PrestoDB) but can also match the performance of Bao-for-PostgreSQL with reduced human supervision and increased adaptivity, as it replaces Bao's static, expert-picked hint-sets with those that are automatically discovered. We also provide an open-source implementation of AutoSteer together with a visual tool for interactive use by query optimization experts.

Joint Velocity and Spectrum Optimization in Urban Air Transportation System via Multi-Agent Deep Reinforcement Learning

The emerging concepts of Urban Air Mobility (UAM) and Advanced Air Mobility (AAM) open a new paradigm for urban air transportation. One big challenge is that new aerial vehicles (AV) will quickly saturate the already crowded aviation spectrum, which is an essential resource to ensure reliable communications for safe air operations. In this paper, we consider an air transportation system where multiple AVs are operated to transport passengers or cargo from different sources to destinations along their pre-defined paths. During the flight, the minimum communication Quality of Service (QoS) must be achieved at all times to ensure flight safety. Our objective is to minimize the total mission completion time by jointly optimizing the velocities and spectrum allocation for all AVs. We formulate the optimization problem as a multi-stage Markov game where the optimization variables are coupled together. A multi-agent deep reinforcement learning VD3QN algorithm is proposed to enable cooperative learning among AVs. Additionally, we propose a heuristic greedy algorithm (HGA) and an orthogonal multiple access (OMA) solution as baseline solutions. Extensive simulation results show that our learning-based solution outperforms the baseline solutions under different network configurations.

Computer-vision-powered Automatic Waste Sorting Bin: a Machine Learning-based Solution on Waste Management

The mainstream waste separation methods – recycle bins and manual separation – are time consuming, prone to error, and human labor intensive. Robots have demonstrated their superior efficiency at repetitive task over human’s time to time again. This research implements and evaluate a computer-vision-powered Automatic Waste Sorting Bin, which is capable to classify the waste types in a short time with high efficiency. Controlled and classified using a Raspberry pi and a camera, the bin can detect the waste type and drop it in the right bin accordingly. The dataset on which the model is trained on is relatively small. An image of the waste is first captured with the camera, and then analyzed using a YOLOv5 model. Parameters that yield the optimal result are 150 epochs with a YOLOv5l with an accuracy of 93.33 %. To improve the model’s performance, we experimented with different epoch settings and measured the results. The setup proposed with this paper provides an automated solution to replace the mainstream methods of waste separation. This paper provides a low-cost and flexible solution that can be easily replicated and trained with a larger dataset to acquire a better result. Future research may reference this paper as a proof of concept and/or an implementation for a streamlined solution.

SUShe: simple unsupervised shadow removal

Shadow removal is an important problem in computer vision, since the presence of shadows complicates core computer vision tasks, including image segmentation and object recognition. Most state-of-the-art shadow removal methods are based on complex deep learning architectures, which require training on a large amount of data. In this paper a novel and efficient methodology is proposed aiming to provide a simple solution to shadow removal, both in terms of implementation and computational cost. The proposed methodology is fully unsupervised, based solely on color image features. Initially, the shadow region is automatically extracted by a segmentation algorithm based on Electromagnetic-Like Optimization. Superpixel-based segmentation is performed and pairs of shadowed and non-shadowed regions, which are nearest neighbors in terms of their color content, are identified as parts of the same object. The shadowed part of each pair is relighted by means of histogram matching, using the content of its non-shadowed counterpart. Quantitative and qualitative experiments on well-recognized publicly available benchmark datasets are conducted to evaluate the performance of proposed methodology in comparison to state-of-the-art methods. The results validate both its efficiency and effectiveness, making evident that solving the shadow removal problem does not necessarily require complex deep learning-based solutions.

Deep learning based Metal Artifact Reduction in X-ray Computed Tomography

Due to the presence of metal implants, computed tomography(CT) images of patients undergoing CT scans produce severe metal artifacts. In recent years, metal artifact reduction (MAR) algorithms have been developed at a high speed, and deep learning-based MAR algorithms have proved to be one of the very effective methods. However, based on the fact that most of the current deep learning-based solutions utilize simulated data for supervised training, these models are difficult to be directly applied in clinical settings. In addition, current MAR schemes still face considerable challenges in reduce streak artifacts and dark band artifacts from metal artifacts. In this paper, we propose an improved unsupervised network model based on CycleGAN suitable for small-sample training of medical images. In order to enhance the extraction of image features, the generator in this paper adopts the U-Net network model, which is commonly used in medical image processing, and the residual connection and attention modules are added between each layer of the U-Net network. In addition, in order to better repair the dark band artifacts, the input of the network is processed by mixing the original image and the prior image, and in order to make the output image retain as much as possible the effective information of the original image, this paper adds the correlation constraints of the prior image and the original image to the output of the network. Experimentally, it is proved that the method in this paper has very good ability to reduce metal artifacts.

A Comprehensive Analysis of Machine Learning- and Deep Learning-Based Solutions for DDoS Attack Detection in SDN

A Comprehensive Analysis of Machine Learning- and Deep Learning-Based Solutions for DDoS Attack Detection in SDN