A deep reinforcement learning approach to fair distributed dynamic spectrum access

Wed 27 Jul 2022 14:34:59 AEST

This paper investigates the task how to achieve fairness in distributed dynamic spectrum access (DSA). Specifically, we consider a cognitive radio network scenario with multiple primary users (PUs) and secondary users (SUs). Each PU operates in a licensed channel. We assume that there is no coordination between PUs and SUs, and no coordination among SUs. The key challenges for SUs are to: (1) avoid collisions with PUs, (2) avoid collisions with other SUs, (3) fair access of spectrum resources in an uncoordinated system, (4) deal with different PU activity patterns, (5) deal with spectrum sensing errors. To address these challenges, we propose a deep reinforcement learning (DRL) approach and an associated reward function to achieve fair access to spectrum resources. Specifically, we use the method of Dueling Double Deep Q-Networks with Prioritised Experience Replay (D3QN-PER) as DRL algorithm for each SU. In our simulation experiments, we demonstrate that the proposed approach performs better than existing DRL methods.

Semi-supervised manifold alignment using parallel deep autoencoders

Wed 26 Oct 2022 14:26:21 AEDT

The aim of manifold learning is to extract low-dimensional manifolds from high-dimensional data. Manifold alignment is a variant of manifold learning that uses two or more datasets that are assumed to represent different high-dimensional representations of the same underlying manifold. Manifold alignment can be successful in detecting latent manifolds in cases where one version of the data alone is not sufficient to extract and establish a stable low-dimensional representation. The present study proposes a parallel deep autoencoder neural network architecture for manifold alignment and conducts a series of experiments using a protein-folding benchmark dataset and a suite of new datasets generated by simulating double-pendulum dynamics with underlying manifolds of dimensions 2, 3 and 4. The dimensionality and topological complexity of these latent manifolds are above those occurring in most previous studies. Our experimental results demonstrate that the parallel deep autoencoder performs in most cases better than the tested traditional methods of semi-supervised manifold alignment. We also show that the parallel deep autoencoder can process datasets of different input domains by aligning the manifolds extracted from kinematics parameters with those obtained from corresponding image data.

Recognition of emotion from speech using evolutionary cepstral coefficients

Wed 23 Nov 2022 14:17:23 AEDT

An optimal representation of acoustic features is an ongoing challenge in automatic speech emotion recognition research. In this study, we proposed Cepstral coefficients based on evolutionary filterbanks as emotional features. It is difficult to guarantee that an individual optimized filterbank provides the best representation for emotion classification. Consequently, we employed six HMM-based binary classifiers that used a specific filterbank, which was optimized by a genetic algorithm to categorize the data into seven emotion classes. These optimized classifiers were applied in a hierarchical manner and outperformed conventional Mel Frequency Cepstral Coefficients in terms of overall emotion classification accuracy. The proposed method using evolutionary-based Cepstral coefficients achieved a weighted average recall of 87.29% on the Berlin database while the same approach but using conventional Cepstral features achieved only 79.63%.

Uncertainty Estimation in the Neural Model for Aeromagnetic Compensation

Wed 17 Aug 2022 11:59:14 AEST

Measuring the performance of an aeromagnetic compensation system is usually difficult. The standard deviation of the signal has been used as an index in the industry. While the standard deviation is drawn from frequency statistics, it cannot represent the performance on a single sampling point. On the other hand, as the true geomagnetic intensity is unknown, the signal's deviation is actually an approximate measurement of the residual error. This letter first analyzes the traditional neural model for aeromagnetic compensation to reveal the fact that the model can only estimate the expectation of interference. Then, we introduce a stochastic hidden variable to predict the standard deviation synchronously. The proposed model is derived from variational inference and trained as a stochastic gradient variational Bayes estimator. Simulations are performed to show the correlation between the true residual error and the estimated standard deviation.

Comparing ellipse detection and deep neural networks for the identification of drinking glasses in images

Wed 14 Sep 2022 09:12:37 AEST

This study compares a deep learning approach with the traditional computer vision method of ellipse detection on the task of detecting semi-transparent drinking glasses filled with water in images. Deep neural networks can, in principle, be trained until they exhibit excellent performance in terms of detection accuracy. However, their ability to generalise to different types of surroundings relies on large amounts of training data, while ellipse detection can work in any environment without requiring additional data or algorithm tuning. Two deep neural networks trained on different image data sets containing drinking glasses were tested in this study. Both networks achieved high levels of detection accuracy, independently of the test image resolution. In contrast, the ellipse detection method was less consistent, greatly depending on the visibility of the top and bottom of the glasses, and water levels. The method detected the top of the glasses in less than half of the cases, at lower resolutions; and detection results were even worse for the water level and bottom of the glasses, in all resolutions.

Fast evolution of CNN architecture for image classification

Wed 14 Dec 2022 15:56:32 AEDT

The performance improvement of Convolutional Neural Network (CNN) in image classification and other applications has become a yearly event. Generally, two factors are contributing to achieving this envious success: stacking of more layers resulting in gigantic networks and use of more sophisticated network architectures, e.g. modules, skip connections, etc. Since these state-of-the-art CNN models are manually designed, finding the most optimized model is not easy. In recent years, evolutionary and other nature-inspired algorithms have become human competitors in designing CNN and other deep networks automatically. However, one challenge for these methods is their very high computational cost. In this chapter, we investigate if we can find an optimized CNN model in the classic CNN architecture and if we can do that automatically at a lower cost. Towards this aim, we present a genetic algorithm for optimizing the number of blocks and layers and some other network hyperparameters in classic CNN architecture. Experimenting with CIFAR10, CIFAR100, and SVHN datasets, it was found that the proposed GA evolved CNN models which are competitive with the other best models available.

CyTex: Transforming speech to textured images for speech emotion recognition

Wed 13 Sep 2023 10:09:53 AEST

Speech emotion recognition is an important aspect of emotional state recognition in human–machine interaction. Approaches using speech-to-image transforms have become popular in recent years because they can utilise deep neural network models that have proven to be successful in the image processing domain. In this paper, we propose a new speech-to-image transform, CyTex, that maps the raw speech signal directly to a textured image by using calculations based on the fundamental frequency of each speech frame. The textured RGB images resulting from the CyTex transform can then be classified using standard deep neural network models for the recognition of different classes of emotion. Using this approach, we can report an improvement of classification accuracies over the previous state-of-the-art results by 0.81% for the Emo-DB database, and also by 0.5% for the IEMOCAP database.

Global urban cityscape - unsupervised clustering exploration of human activity and mobility infrastructure

Wed 13 Mar 2024 08:50:08 AEDT

It is widely accepted that cities cultivate innovation and are the engines of productivity. The identification of strengths and weaknesses will enchant social mobility providing equal opportunities for all. The study at hand investigates the relationship between social mobility and transportation planning in 1,860 central urban areas across the globe. Datamining processes combining open-sourced, automated, and crowdsourced information from four major pillars of social mobility (demographics, human activity, transport infrastructure, and environmental quality) are used to describe each location. Next, unsupervised clustering algorithms are used to analyse the extracted information, in order to identify similar characteristics and patterns among urban areas. The process, which comprises an objective framework for the analysis of urban environments, resulted in four major types of central areas, that represent similar patterns of human activity and transport infrastructure.

Foreign exchange trading with support vector machines

Wed 11 Apr 2018 18:15:20 AEST

This paper analyzes and examines the general ability of Support Vector Machine (SVM) models to correctly predict and trade daily EUR exchange rate directions. Seven models with varying kernel functions are considered. Each SVM model is benchmarked against traditional forecasting techniques in order to ascertain its potential value as out-of-sample forecasting and quantitative trading tool. It is found that hyperbolic SVMs perform well in terms of forecasting accuracy and trading results via a simulated strategy. This supports the idea that SVMs are promising learning systems for coping with nonlinear classification tasks in the field of financial time series applications.

The Machine Intelligence Hex Project

Wed 11 Apr 2018 17:07:21 AEST

Hex is a challenging strategy board game for two players. To enhance students' progress in acquiring understanding and practical experience with complex machine intelligence and programming concepts we developed the Machine Intelligence Hex (MIHex) project. The associated undergraduate student assignment is about designing and implementing Hex players and evaluating them in an automated tournament of all programs developed by the class. This article surveys educational aspects of the MIHex project. Additionally, fundamental techniques for game programming as well as specific concepts for Hex board evaluation are reviewed. The MIHex game server and possibilities of tournament organisation are described. We summarise and discuss our experiences from running the MIHex project assignment over four consecutive years. The impact on student motivation and learning benefits are evaluated using questionnaires and interviews.

Sustaining residential social space: a visual and spatial analysis of the nearly urban

Wed 11 Apr 2018 15:57:28 AEST

Residential development within heritage conservation areas is regulated by Development Control Plans (DCP), which provide guidelines about the shape and form that new houses, alterations and additions should take (DIPNR 2004). By understanding that the visual amenity of streets within a city plays an important role in creating a sense of place and community for its citizens (Lynch 1960) they attempt to sustain, through regulation, an urban pattern that has become valued by the community. This paper, being the third in a series of papers delivered at ANZAScA conferences, will build on an established interdisciplinary approach, utilising architectural knowledge and computer imaging to evaluate the visual character of detached housing within a heritage conservation area. The visual environment is assessed using computer software developed to locate the visual boundaries within a view or elevation. Aerial views of the streets and houses are also analysed using the established image segmentation methods, extending the spatial assessment of a streetscape beyond the purely visual information derived from walking through it.

A small spiking neural network with LQR control applied to the acrobot

Wed 11 Apr 2018 14:37:33 AEST

This paper presents the results of a computer simulation which, combined a small network of spiking neurons with linear quadratic regulator (LQR) control to solve the acrobot swing-up and balance task. To our knowledge, this task has not been previously solved with spiking neural networks. Input to the network was drawn from the state of the acrobot, and output was torque, either directly applied to the actuated joint, or via the switching of an LQR controller designed for balance. The neural network’s weights were tuned using a (μ + λ)-evolution strategy without recombination, and neurons’ parameters, were chosen to roughly approximate biological neurons.

Application of intelligent systems for news analytics

Wed 11 Apr 2018 14:28:36 AEST

The chapter starts with an overview of existing text mining systems whose main purpose is predicting equity price movements on the financial markets. In general, these systems transform the input text to a so-called sentiment score, a numerical value equivalent to the opinion of an analyst on the influence of the news text to the further development of the regarded stock. In the second part it is explored how the sentiment score relates to some of the relevant macroeconomic variables. It is suggested that raw sentiment score can be transformed to reveal sentiment reversals, and such transformed indicator relates better to future returns. As an example the project FINDS is presented as an integrated system that consists of a module that performs sentiment extraction from the financial news, a benchmark module for comparison between different classification engines, and a visualization module used for the representation of the sentiment data to the end users, thus supporting the traders in analysing news and making buy and sell decisions.

Line segmentation: a computational technique for architectural image analysis

Wed 11 Apr 2018 14:01:23 AEST

Planar methods have typically dominated the computational analysis of architectural and urban space and form. For example, the majority of space syntax research that has been undertaken over the last three decades has been concerned almost exclusively with architectural and urban plans. In contrast, analytical methods that consider the formal, or visual, qualities of architectural facades, or images of buildings, are not only rare, but only a few have ever been repeated and adequately tested. The present paper outlines a new method-derived from the Hough Transform algorithm - for the dissolution of architectural images into segmented lines that can be counted and charted, and that can have their spatial orientation determined. This method for investigating the visual qualities of buildings is demonstrated in an analysis of a series of images of suburban houses. The proposed method, line segmentation, is potentially significant because it is a method not commonly used for the quantitative analysis of the formal and textural character of real buildings, it is repeatable, and it delivers consistent results if a simple procedure is followed.

Kernel methods in finance

Wed 11 Apr 2018 13:54:13 AEST

The first part of the present chapter provides some theoretical background and explains in the next section the general idea of kernel machines and kernelisation. Then the three fundamental machine learning paradigms dimensionality reduction, regression, and classification as well as associated questions of kernel and parameter selection are addressed. The chapter's second part gives a survey of typical questions and tasks arising in finance applications and how kernel methods have been applied to solve them. Finally follows a brief overview of relevant software toolboxes.

A computational analysis of fractal dimensions in the architecture of Eileen Gray

Wed 11 Apr 2018 12:26:11 AEST

This paper is the first investigation of the fractal dimensions of five of the house designs of Eileen Gray: a prominent architect working mainly in France between 1922 and and 1956. In this paper, a computational variation of the "box-counting approach" (used to determine fractal dimension) is applied to a multi-dimensional review of the houses of Gray. As a contemporary of Le Corbusier, Gray is a significant architect for such an analysis. This research is important because it expands the set of examples of early twentieth century architects who have been analyzed using the method. This paper provides a computer-assisted mathematical analysis of characteristic visual complexity in five houses designs by Eileen Gray.

Motivated reinforcement learning for improved head actuation of humanoid robots

Wed 11 Apr 2018 12:14:35 AEST

The ability of an autonomous agent to self-localise within its local environment is critically dependent on its ability to make accurate observations of static, salient features. This notion has driven considerable research into the development and improvement of feature extraction and object recognition algorithms, both within RoboCup and the robotics community at large. Instead, this paper focuses on a rarely-considered issue; determining an optimal policy for actuating a robot’s head, to ensure it observes regions of the environment that will maximise the positional nformation provided. The complexity of this task is magnified by a number of common computational issues; specifically high dimensionality state spaces and noisy environmental observations. This paper details the application of motivated reinforcement learning to partially overcome these issues, leading to an 11% improvement (relative to the null case of uniformly distributed actuation policies) in self-localisation for an agent trained online for less than 1 hour. The method is demonstrated as a viable method for improving self-localisation in robotics, without the need for further optimisation of object recognition or tuning of probabilistic filters.

The visual qualities of liveable streets

Wed 11 Apr 2018 10:06:57 AEST

The visual character of buildings is often associated with the style of its construction - a set of visual characteristics that a group of buildings might share. These characteristics include the relationship of the parts of the building to each other and to the building as a whole, the use of ornament and visible textures , and the scale of elements within the composition. Alexander (Alexander, 2002) and Rapoport (Rapoport, 1990) have discussed how these visual features affect urban coherence, and the way that people behave in public spaces. Using algorithms developed within robotic research that enable a computer to interpret a visual environment (similar to those who used the medicine and facial recognition for instance), this paper outlines how algorithms can be used to study the visual properties of the built environment. the paper discussed how the urban texture of houses within the streetscape, at the scale of the individual, relates to the urban visual character of precincts within the city. The architectural drawings of a house designed by Le Corbusier and the street frontages of post war and federation houses are visually analysed using Archimage, software developed by the authors for this type of analysis.

A multi-agent system with reinforcement learning for railway traffic management

Wed 07 Sep 2022 14:48:05 AEST

Australia is the world's largest coal exporter and in New South Wales, more than 87% of the coal is transported through railways. One of the strategies to increase throughput is to use complex computational techniques for train scheduling optimisation. This paper applies artificial intelligence techniques to the scheduling of coal trains in the context of the Hunter Valley Coal Chain Coordinator (HVCCC). We modelled a reduced version of the HVCCC network using a Multi-Agent System consisting of a centralized dispatcher agent and multiple train agents. The network has 9 single track sections, 6 passing loops, and 3 loading/dumping loops, in a closed tree design. Three methods are tested and analysed: (a) First-in-First-Out (FIFO), (b) Random Walk and (c) Reinforcement Learning (RL) by Q-learning algorithm. The dispatcher agent is responsible for traffic management of the trains and oversees the entire rail network. Our results show that the Q-learning algorithm outperforms the FIFO strategy in all test cases, between a minimum of 3.5% and a maximum of 15.2%. Different congestion scenarios were also compared and the RL technique presented the best values for the objective of minimizing the total dwell time of all trains in the network. Considering the results achieved by the Q-learning algorithm, this work presents a first step towards the application of intelligent agents for the traffic management of HVCCC's network.

Topological Dynamics of Functional Neural Network Graphs During Reinforcement Learning

Wed 06 Mar 2024 10:45:24 AEDT

This study investigates the topological structures of neural network activation graphs, with a focus on detecting higher-order Betti numbers during reinforcement learning. The paper presents visualisations of the neurotopological dynamics of reinforcement learning agents both during and after training, which are useful for different dynamics analyses which we explore in this work. Two applications are considered: frame-by-frame analysis of agent neurotopology and tracking per-neuron presence in cavity boundaries over training steps. The experimental analysis suggests that higher-order Betti numbers found in a neural network’s functional graph can be associated with learning more complex behaviours.

Violence Detection in Real-Life Audio Signals Using Lightweight Deep Neural Networks

Tue 28 Nov 2023 16:02:13 AEDT

The automatic detection of violent behavior in acoustic data has become an important research area due to its growing application potential in various surveillance and behavior monitoring tasks where it causes fewer privacy issues than video. In this paper, we propose two deep learning approaches of different characteristics to classify speech data containing violent behavior against data that only includes patterns of non-violent behavior. The first approach is based on conventional deep neural networks, while the second approach uses lightweight deep neural networks. Both utilize Mel-spectrogram images of speech signals as input to fine-tuned models. Our best lightweight model's classification accuracy is about 8% better than the previous state-of-the-art result using the same benchmark data. Lightweight models have fewer parameters and require fewer computing resources than conventional deep models. This can be an important advantage when being deployed on mobile or edge devices.

Cognitive Radio Spectrum Sensing and Prediction Using Deep Reinforcement Learning

Tue 26 Jul 2022 14:14:26 AEST

In this paper, we propose to use deep reinforcement learning (DRL) for the task of cooperative spectrum sensing (CSS) in a cognitive radio network. We selected a recently proposed offline DRL method called conservative Q-learning (CQL) due to its ability to learn complex data distributions efficiently. The task of CSS is performed as follows. Each secondary user (SU) performs local sensing and using CQL algorithm, determines the presence of licensed user for current and k-1 future timeslots. These results are forwarded to the fusion centre where another CQL algorithm is operating that generates a global decision for the current and k-1 future timeslots. Then, SUs do not perform sensing for the next k-1 timeslots to save energy. The proposed CSS mechanism can significantly increase the licensed user detection accuracy and the data transmissions by SUs. In addition, it reduces the sensing results transmission overhead. The proposed solution is tested with a stochastic traffic load model for different activity patterns. Our simulation results show that the proposed problem formulation using the CQL algorithm can achieve similar detection accuracy as other state-of-the-art methods for CSS while significantly reducing the computation time.

Improving the reliability of implicit averaging methods using new conditional operators for robust optimization

Tue 23 Aug 2022 08:27:15 AEST

In the field of robust optimization, the robustness of a solution is confirmed using a robustness indicator. In the literature, such an indicator uses explicit or implicit averaging techniques. One of the main drawbacks of the implicit averaging techniques is unreliability since they only use the sampled points generated by an optimization algorithm. In this paper, we propose a set of conditional operators for comparing solutions based on the number of sampled solutions in their neighbourhoods, thereby making reliable decisions during the process of robust optimization. This technique is integrated into the Particle Swarm Optimization (PSO) to update GBEST and PBESTs reliably, and the designed robust PSO algorithm is applied to a number of case studies. A set of extensive experiments shows that the proposed technique prevents an algorithm that relies on implicit averaging technique from making risky decisions and thus proven beneficial in finding robust solutions.

Synthetic Data Generation and Deep Learning for the Topological Analysis of 3D Data

Tue 21 May 2024 09:37:32 AEST

This research uses deep learning to estimate the topology of manifolds represented by sparse, unordered point cloud scenes in 3D. A new labelled dataset was synthesised to train neural networks and evaluate their ability to estimate the genus of these manifolds. This data used random homeomorphic deformations to provoke the learning of visual topological features. We demonstrate that deep learning models could extract these features and discuss some advantages over existing topological data analysis tools that are based on persistent homology. Semantic segmentation was used to provide additional geometric information in conjunction with topological labels. Common point cloud multi-layer perceptron and transformer networks were both used to compare the viability of these methods. The experimental results of this pilot study support the hypothesis that, with the aid of sophisticated synthetic data generation, neural networks can perform segmentation-based topological data analysis. While our study focused on simulated data, the accuracy achieved suggests a potential for future applications using real data.

Optimisation of 2D U-Net model components for automatic prostate segmentation on MRI

Tue 18 Jan 2022 11:49:18 AEDT

In this paper, we develop an optimised state-of-the-art 2D U-Net model by studying the effects of the individual deep learning model components in performing prostate segmentation. We found that for upsampling, the combination of interpolation and convolution is better than the use of transposed convolution. For combining feature maps in each convolution block, it is only beneficial if a skip connection with concatenation is used. With respect to pooling, average pooling is better than strided-convolution, max, RMS or L2 pooling. Introducing a batch normalisation layer before the activation layer gives further performance improvement. The optimisation is based on a private dataset as it has a fixed 2D resolution and voxel size for every image which mitigates the need of a resizing operation in the data preparation process. Non-enhancing data preprocessing was applied and five-fold cross-validation was used to evaluate the fully automatic segmentation approach. We show it outperforms the traditional methods that were previously applied on the private dataset, as well as outperforming other comparable state-of-the-art 2D models on the public dataset PROMISE12.

Parallel LSTM Architectures for Non-Intrusive Load Monitoring in Smart Homes

Tue 16 Aug 2022 16:37:40 AEST

Non-Intrusive Load Monitoring (NILM) is becoming popular as an appliance monitoring technique that can extract detailed power consumption information related to the different appliances used in a household. NILM utilises time series analysis methods to disaggregate signals of operating appliances from a single point in houses. Based on that it can help to provide energy consumption advice for household owners. NILM is known to be a challenging task from a computational aspect. This study proposes deep neural network models with a Parallel Long Short-Term Memory Topology (PLT) and evaluates them on the public REDD dataset. The new models' experimental evaluation results exhibit a significant improvement in four main state-based and energy-based metrics in comparison to previous results obtained using deep denoising autoencoders.

A smart meter firmware update strategy through network coding for AMI network

Tue 14 Nov 2023 14:54:59 AEDT

With the introduction of communication infrastructure into the traditional power grids, smart power grids are emerging to meet the future electricity demands. In smart grid, advanced metering infrastructure (AMI) is one of the main components that enables bi-directional communication between home area networks and utility providers. In an AMI network, one of the crucial operations is to update the firmware of the smart meters. In this paper, we propose a new forwarding strategy for the firmware updates in AMI network. Our simulation results show that the completion time of the smart meter firmware update process can be reduced significantly by using the proposed new strategy.

Monocular ORB-SLAM on a humanoid robot for localization purposes

Tue 14 Jan 2020 12:16:45 AEDT

This work presents an application of ORB-SLAM in an iGus bipedal humanoid robotic platform. The method was adapted from its original implementation into the framework used by the NUbots robotic soccer team and used for localization purposes. The paper presents a description of the challenges to implement the adaptation, as well as several tests where the method's performance is analyzed to determine its suitability for further development and use on medium sized humanoid robots. To conduct the tests, we determined the robot's real location using a high-accuracy, camera-based infrared tracking system. Two experiments were performed to estimate the robustness of the method to the vibration and constant camera wobbling inherent to a bipedal walk and its ability to deal with the kidnapped robot problem. The tests indicate that ORB-SLAM is suitable for implementation into a medium sized humanoid robot in situations comparable to a robotic soccer environment, and requires relatively low computational resources, leaving enough CPU power for other tasks. Additionally, since ORB-SLAM is robust to the difficulties associated with humanoid motion, we conclude that it provides a good SLAM algorithm to enhance with features specific to the humanoid robotic platform.

Stereo Visual Mesh for Generating Sparse Semantic Maps at High Frame Rates

Tue 13 Feb 2024 12:01:29 AEDT

The Visual Mesh is an input transform for deep learning that allows depth independent object detection at very high frame rates. The present study introduces a Visual Mesh based stereo vision method for sparse stereo semantic segmentation. A dataset of simulated 3D scenes was generated and used for training to show that the method is capable of processing high resolution stereo inputs to generate both left and right sparse semantic maps. The new stereo method demonstrated better classification accuracy than the corresponding monocular approach. The high frame rates and high accuracy may make the proposed approach attractive to fast-paced on-board robot or IoT applications.

A comparative analysis of deep neural network architectures for sentence classification using genetic algorithm

Tue 11 Jun 2024 16:10:17 AEST

Because of the number of different architectures, numerous settings of their hyper-parameters and disparity among their sizes, it is difficult to equitably compare various deep neural network (DNN) architectures for sentence classification. Evolutionary algorithms are emerging as a popular method for the automatic selection of architectures and hyperparameters for DNNs whose generalisation performance is heavily impacted by such settings. Most of the work in this area is done in the image domain, leaving text analysis, another prominent application domain of deep learning, largely absent. Besides, literature presents conflicting claims regarding the superiority of one DNN architecture over others in the context of sentence classification. To address this issue, we propose a genetic algorithm (GA) for optimising the architectural and hyperparameter settings in different DNN types for sentence classification. To enable the representation of the wide variety of architectures and hyperparameters utilised in DNNs, we employed a generalised and flexible encoding scheme in our GA. Our study involves optimising two convolutional and three recurrent architectures to ensure a fair and unbiased evaluation of their performance. Furthermore, we explore the effects of using F1 score versus accuracy as a performance metric during evolutionary optimisation of those architectures. Our results, using ten datasets, show that, in general, the architectures and hyperparameters evolved using the F1 score tended to outperform those evolved using accuracy and in the case of CNN and BiLSTM the results were significant in statistical measures. Of the five architectures considered, the GA-evolved gated recurrent unit (GRU) performed the strongest overall, achieving good generalisation performance while using relatively few trainable parameters, establishing GRU as the preferred architecture for the sentence classification task. The optimised architectures exhibited comparable performance with the state-of-the-art, given the large difference in trainable parameters.

Enhanced embeddings in zero-shot learning for environmental audio

Tue 09 Apr 2024 15:28:34 AEST

Zero-shot learning is a scenario in machine learning where the classes used in the training and test sets are disjoint. This work considers zero-shot learning for environmental audio and improves results by enhancing audio and word embeddings. Previous works use the VGGish model for audio embeddings, and textual class labels are often used as input for word embedding networks such as Word2Vec. This study instead uses a modified YAMNet network to obtain semantic audio embeddings for zero-shot learning. Moreover, part of this study involves adding linguistic devices, such as synonyms, semantic broadening and onomatopoeia, to the input of the word embeddings. With these two modifications, top-1 accuracy is increased on average by over five percentage points compared to the state-of-the-art on ESC-50. This emerging area of research has applications in robot awareness, security systems and wildlife conservation in situations where no data is available for some classes.

Optimization of Robot Movements Using Genetic Algorithms and Simulation

Tue 08 Nov 2022 09:32:21 AEDT

This work describes the optimization of two robot movements in the context of the Humanoid league competition at RoboCup. A multi-objective genetic algorithm (MOGA) was used in conjunction with the real-time physics simulator Gazebo. The motivation for this work was that the NUbots team, from the University of Newcastle, lacked a simulation platform for their soccer-playing robots. Gazebo was the preferred choice of simulator, offering built-in compatibility with the Robot Operating System (ROS). The NUbots robot software, however, uses a proprietary message-passing framework in place of ROS. This work thus describes the pathway to use Gazebo with non-ROS compliant applications. In addition, it describes how MOGA can be used to optimize complex movements in an efficient manner. The two robot movements optimized were a kick script and the walk engine. For the kick script, the resulting optimal configuration improved the kick distance by a factor of six, with 50% less torso sway. For the walk engine, the forward speed increased by 50%, with 38% less torso sway, compared to the manually-tuned walk engine.

Sliding window Bag-of-Visual-Words for low computational power robotics scene matching

Tue 01 Feb 2022 14:58:34 AEDT

In this paper, we introduce a new method, based on a sliding window geometrical extension to Bag-of-Visual-Words (called swBOVW) intended for application to low computational power robots. Benchmarked against RANSAC as a geometric validator to BOVW, three implementations of this technique are presented to improve either the performance or the computational cost. The three implementations are: as a replacement to RANSAC as a geometric validator; as a supplement to RANSAC; and as a replacement to traditional BOVW when the number of images in the database can be reduced. Seeking to utilise some of the geometric information ignored by traditional BOVW, this technique is developed from the use of sub-regions in Spatial Pyramids, and applied to the matching of whole images. This technique is applied in the context of humanoid robotic soccer to the problem of field end symmetry, and provides geometric validation along the horizontal axis of images. When applied, the technique has been able to either halve the cases of unresolved image queries, or halve the computational cost required to achieve comparable results to the benchmark.

Learning to See Topological Properties in 4D Using Convolutional Neural Networks

Thu 25 Jul 2024 09:46:52 AEST

Topology describes the essential structure of a space, and in 4D, a larger variety of topologically distinct manifolds can be embedded versus 2D or 3D. The present study investigates an end-to-end visual approach, which couples data generation software and convolutional neural networks (CNNs) to estimate the topology of 4D data. A synthetic 4D training data set is generated with the use of several manifolds, and then labelled with their associated Betti numbers by using techniques from algebraic topology. Several approaches to implementing a 4D convolution layer are compared. Experiments demonstrate that already a basic CNN can be trained to provide estimates for the Betti numbers associated with the number of one-, two-, and three-dimensional holes in the data. Some of the intricacies of topological data analysis in the 4D setting are also put on view, including aspects of persistent homology.

Robust multi-objective optimization using conditional pareto optimal dominance

Thu 24 Nov 2022 12:00:48 AEDT

Robust optimization of real-world problems is essential to reduce the significant negative impact of uncertainties and noises present in the environment. Uncertainties in the decision variables are often handled using explicit or implicit averaging methods, in which the fitness of a solution isjudged based on the objective values of neighbouring solutions. Explicit averaging methods are highly reliable but require additional objective function evaluation, which can significantly increases the overall computational cost of an optimization process. On the other hand, implicit averaging techniques are computationally cheap, yet they suffer from low reliability since they use the history of search in a population-based optimization algorithm. This work proposes a conditional Pareto optimal dominance to improve the reliability of robust optimization methods that use implicit averaging methods. The proposed method is applied to Multi-Objective Particle Swarm optimisation. Empirical study with a benchmark suite shows the benefit of the proposed conditional Pareto optimal dominance in locating robust solutions in multi-objective problems.

Testing the robustness of manifold learning on examples of thinned-out data

Thu 17 Jun 2021 14:21:18 AEST

Manifold learning can only be successful if enough data is available. If the data is too sparse, the geometrical and topological structure of the manifold extracted from the data cannot be recognised and the manifold collapses. In this paper we used data from a simulated two-dimensional double pendulum and tested how well several manifold learning methods could extract the expected manifold, a two-dimensional torus. The experiments were repeated while the data was down sampled in several ways to test the robustness of the different manifold learning methods. We also developed a neural network-based deep autoencoder for manifold learning and demonstrated that it performed in most of our test cases similarly or better than traditional methods such as principal component analysis and isomap.

Subliminal Faces

Thu 16 Nov 2023 12:19:44 AEDT

The dataset comprises 103 jpg images of house facades or related building elements. The facade images were selected to have enough structure so that some of them could be associated with abstract face-like patterns. For example, some windows may be interpreted as eyes, some doors as noses, and so on.

A decentralised multi-agent system for rail freight traffic management

Thu 11 Aug 2022 11:26:17 AEST

The world’s largest coal export operation is located in New South Wales, Australia. The state has more than 87% of the coal transportation done through railways, and one of the strategies to increase throughput is the use of sophisticated computational techniques for rail traffic optimisation. The current state of the art shows a lack of practical applications, thus making scalability, decentralisation and real-world commitment three key research directions. Towards that, this research presents a simulation-based machine learning approach for the railway traffic management problem, in the context of the Hunter Valley Coal Chain (HVCC). We modelled trains, load points and terminals as autonomous intelligent agents that interact, learn and act independently—thus constituting a multi-agent system (MAS). The MAS is implemented on top of a rail network simulation model currently in use at the HVCC. The model is adapted as a decentralised partially-observed Markov decision process environment that allows multi-agent learning via a genetic algorithm. We present experiments with scenarios based on the actual rail network data, which show that the MAS outperforms the heuristic approach embedded in the HVCC simulation tool by up to 81% (in terms of the schedule’s total dwell time). Further to those experiments, a comparison analysis evaluates the relevance of specific state features (e.g. track length, train conflicts, etc.). Finally, an important outcome was that the agents have learned to overcome very complex traffic situations that appear in train scheduling operations and that sometimes result in unnecessarily long dwell times. This type of high level learning represents a significant step forward in the use of complex computational techniques for rail transportation problems. [Final citation details to be advised.]

A novel u-shaped transfer function for binary particle swarm optimisation

Thu 09 Sep 2021 14:13:32 AEST

Particle swarm optimisation (PSO), one of the most elegant algorithms in the field of nature-inspired optimisation, has many variants for solving different types of problems. One of these variants is binary particle swarm optimisation (BPSO), which is suitable for solving combinatorial optimisation problems. A main component of BPSO is the transfer function that maps continuous velocity values to probability values which in turn are used to update particle positions. Transfer function has a significant impact on the performance of BPSO algorithm. This paper proposes a novel transfer function with tunable parameters that allows different U-shaped transfer functions. For evaluating the proposed transfer functions, a set of benchmark functions and 0/1 knapsack problems are employed. The results show that the U-shaped transfer functions can significantly improve the performance of BPSO. It is also shown that the BPSO algorithms equipped with U-shaped transfer functions provide superior results compared to the existing transfer functions in the literature.

Voxel-based supervised machine learning of peripheral zone prostate cancer using noncontrast multiparametric MRI

Thu 01 Sep 2022 13:18:31 AEST

Purpose: The aim of this study was to develop and assess the performance of supervised machine learning technique to classify magnetic resonance imaging (MRI) voxels as cancerous or noncancerous using noncontrast multiparametric MRI (mp-MRI), comprised of T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), and advanced diffusion tensor imaging (DTI) parameters. Materials and methods: In this work, 191 radiomic features were extracted from mp-MRI from prostate cancer patients. A comprehensive set of support vector machine (SVM) models for T2WI and mp-MRI (T2WI + DWI, T2WI + DTI, and T2WI + DWI + DTI) were developed based on novel Bayesian parameters optimization method and validated using leave-one-patient-out approach to eliminate any possible overfitting. The diagnostic performance of each model was evaluated using the area under the receiver operating characteristic curve (AUROC). The average sensitivity, specificity, and accuracy of the models were evaluated using the test data set and the corresponding binary maps generated. Finally, the SVM plus sigmoid function of the models with the highest performance were used to produce cancer probability maps. Results: The T2WI + DWI + DTI models using the optimal feature subset achieved the best performance in prostate cancer detection, with the average AUROC , sensitivity, specificity, and accuracy of 0.93 ± 0.03, 0.85 ± 0.05, 0.82 ± 0.07, and 0.83 ± 0.04, respectively. The average diagnostic performance of T2WI + DTI models was slightly higher than T2WI + DWI models (+3.52%) using the optimal radiomic features. Conclusions: Combination of noncontrast mp-MRI (T2WI, DWI, and DTI) features with the framework of a supervised classification technique and Bayesian optimization method are able to differentiate cancer from noncancer voxels with high accuracy and without administration of contrast agent. The addition of cancer probability maps provides additional functionality for image interpretation, lesion heterogeneity evaluation, and treatment management.

Software development in the City Evolutions Project

Sat 24 Mar 2018 08:02:21 AEDT

The goal of the City Evolutions Project is to establish interactive systems and games to entertain users. Because of the existing variabilities in the system and potential reuse for similar systems in this domain, the software system is designed and developed in a reuse-based way, i.e. Software Product Line Engineering (SPLE). SPLE is a reuse-based software approach with reusable software assets in order to maximise software reuse. In SPLE, it is very important to maintain the requirement traceability from software family establishment to individual product derivation. In this paper, we describe our experience of applying SPLE approach in the City Evolutions project and propose an approach to enhance the requirements traceability in SPLE.

A call to scale up biodiversity monitoring from idiosyncratic, small-scale programmes to coordinated, comprehensive and continuous monitoring across large scales

Mon 24 Jul 2023 11:52:50 AEST

Conservation managers cannot manage what they don't know about, yet our existing biodiversity monitoring is idiosyncratic and small in scale. One of Australia's commitments to the Convention for Biological Diversity in 2015 was the creation of a national biodiversity monitoring programme. This has not yet occurred despite the urgent need to monitor common and threatened species, as highlighted by the challenges of determining the biodiversity impacts of the Black Summer fires of 2019/20. In light of improvements to automation, miniaturisation and powering devices, the world urgently needs to scale-up biodiversity monitoring to become coordinated, comprehensive and continuous across large scales. We propose the BIOMON project that could achieve this where individual sensor nodes use machine learning models to identify biodiversity via sound or photos onboard. This could be coupled with abiotic data on temperature and humidity, plus factors such as bushfire smoke. Nodes would be set within networks that transmit the results back to a central cloud repository where robust analyses are conducted and provided free to the public (along with the raw data). Network arrays could be set up across entire continents to measure the change in biodiversity. No one has achieved this yet, and significant challenges remain associated with training the algorithms, low power cellular network coverage, sensor power versus memory trade-offs, and sensor network placement. Much work is still needed to achieve these goals; however we are living in the 21st Century and such lofty goals cannot be achieved unless we start working towards them.

Teaching advanced computing technologies to managers, engineers and other professionals

Mon 23 Sep 2019 13:50:39 AEST

Context: Nowadays most businesses in Australia maintain a website to advertise their products and services and some to also conduct online sales and payments. Many have an additional Facebook page or utilise other on-line tools and phone apps. Data is now considered to be one of the most valuable assets of companies, and many are using Cloud services for safe storage and data processing. With the rapid growth of the volume and diversity of collected data, and its increasing importance for strategic planning, many businesses have started using big data analytics techniques to extract knowledge from data to support their decisions. All of the above technologies require formal security management or otherwise they can become a weak link which can create security vulnerabilities and expose company assets. While some of the computing tasks and their associated security can be outsourced to specialised IT companies, it is crucial that business managers and professionals have sufficient conceptual understanding of these topics in order to make quality decisions and ensure the survival of their business in the modern world. Purpose The question studied it this paper is how best to teach advanced computing technologies to business managers, engineers and other professionals with different prior experience and educational backgrounds. Approach: We describe a curriculum for teaching the most relevant computing technologies to professionals working in various branches of industry and business, mostly based on a Graduate Certificate in Advanced Computing Technologies for Business, currently being taught in an intensive weekend mode. A similar program could also be taught online with one intensive face-to-face session per course. Results: The limited analysis possible during the first run of the program indicates that the delivered courses were well-received by the students who have been able to successfully master the curriculum presented in an intensive weekend delivery mode. Conclusions: While virtually all Australian universities have programs for training IT professionals, apart for a few isolated courses, few universities provide opportunities for other graduates to learn the most relevant computing technologies. Such programs should be developed and included in engineering and business coursework Masters programs and made available to middle and senior business managers.

Estimating Betti numbers using deep learning

Mon 20 Nov 2023 15:54:52 AEDT

This paper proposes an efficient computational approach for estimating the topology of manifold data as it may occur in applications. For two- or three-dimensional point cloud data, the computation of Betti numbers using persistent homology tools can already be computationally very expensive. We propose an alternative approach that employs deep learning to estimate Betti numbers of manifolds approximated by point clouds. A critical aspect in this new approach is the generation of suitable synthetic training data of scalable topological complexity. Once deep neural networks are trained on this data, inference can be computationally efficient and robust to noise. The pilot results of our study for two- and three-dimensional data support the hypothesis that deep convolutional neural networks can estimate Betti numbers of simulated data that has a topological complexity beyond immediate human visual comprehension. The approach could be generalised beyond estimating the numbers of holes, cavities and tunnels in low-dimensional manifolds to counting high-dimensional holes in high-dimensional data.

Designing charm: harnessing the affective power of form in robotic development

Mon 19 Aug 2024 19:45:59 AEST

This paper presents the initial response to the news from Sony that they will cease production of the Aibo robot, which has up to now, provided the player base for the Robocup kague. The design of a new player has all the classic issues of product development but two design considerations stand out l. That the new design enhances the scholarly research objectives that have been a driving force behind the RoboCup. 2.That the new design can step up to the challenge of wearing the mantle Aibo has won for itself as an ambassador for Science and Engineering. The design must convey a positive and engaging image through its form as it moves about the field. It is photographed and presented to a far wider audience than simply die-hard robotic fans. Without losing sight of the academic objectives, can the David Beckham factor be captured, technically brilliant and cute.

Data-Driven Low-Complexity Detection in Grant-Free NOMA for IoT

Mon 17 Jun 2024 10:26:51 AEST

This article proposes a low-complexity data-driven multiuser detector for grant-free nonorthogonal multiple access (GF-NOMA), which has gained significant interest in Internet of Things (IoT). IoT traffic is predominantly sporadic, where devices become active whenever they have data to transmit. The conventional grant-access procedure for requesting a transmission slot every time results in significant signaling overhead and latency. In power domain GF-NOMA, multiple devices can be preallocated the same channel resource, but different power levels. Whenever a device has data, it starts transmission directly using the allocated power level without any grant request. While this significantly reduces the signaling overhead, the access point has to perform the complex task of identifying the active devices and decoding their data. Conventional receivers for power domain NOMA fail in such GF scenarios and the typical solution is to limit transmissions to be packet-synchronized and add carefully chosen pilots in every packet to facilitate activity detection. However, in fairly static IoT networks with low-complexity devices and small packet sizes, this represents a significant overhead and reduces efficiency. In this work we solve the GF-NOMA detection problem without these constraints, by analyzing the boundaries of the received constellation points in power domain GF-NOMA for all activation combinations at once. A low-complexity decision tree-based receiver is proposed, which performs as well as the maximum likelihood-based benchmark receiver, and better than traditional data-driven detectors for GF-NOMA. Comprehensive simulation results demonstrate the performance of the proposed detector in terms of its detection efficiency and parameter learning with minimal training data.

Bulk anatomical density based dose calculation for patient-specific quality assurance of MRI-only prostate radiotherapy

Mon 16 Aug 2021 16:29:49 AEST

Prostate cancer treatment planning can be performed usingmagnetic resonance imaging(MRI) only with sCT scans. However, sCT scans are computer generated from MRI and therefore robust, efficient, and accurate patient-specific quality assurance for dosimetric verification are required. Bulk anatomical density (BAD) maps can be generated based on anatomical contours derived from the MRI image. This study investigates and optimizes the BAD map approach for sCT quality assurance with a large patient CT and MRI dataset. 3D T2-weighted MRI and full density CT images of 54 patients were used to create BAD maps with different tissue class combinations. Mean Hounsfield units (HU) of Fat (F: below −30 HU), the entire Tissue [T: excluding bone (B)], and Muscle (M: excluding bone and fat) were derived from the CT scans. CT based BAD maps (BADBT,CT and BADBMF,CT) and a conventional bone and water bulk-density method (BADBW,CT) were compared to full CT calculations with bone assignments to 366 HU (measured) and 288 HU (obtained from literature). Optimal bulk densities of Tissue for BADBT,CT and Bone for BADBMF,CT were derived to provide zero mean isocenter dose agreement to the CT plan. Using the optimal densities, the dose agreement of BADBT,CT and BADBMF,CT to CT was redetermined. These maps were then created for the MRI dataset using auto-generated contours and dose calculations compared to CT. The average mean density of Bone, Fat, Muscle, and Tissue were 365.5 ± 62.2, −109.5 ± 12.9, 23.3 ± 9.7, and −46.3 ± 15.2 HU, respectively. Comparing to other bulk-density maps, BADBMF,CT maps provided the closest dose to CT. Calculated optimal mean densities of Tissue and Bone were −32.7 and 323.7 HU, respectively. The isocenter dose agreement of the optimal density assigned BADBT,CT and BADBMF,CT to full density CT were 0.10 ± 0.65% and 0.01 ± 0.45%, respectively. The isocenter dose agreement of MRI generated BADBT,MR and BADBMF,MR to full density CT were −0.15 ± 0.90% and −0.16 ± 0.65%, respectively. The BAD method with optimal bulk densities can provide accurate and efficient patient-specific quality assurance for dose calculations in MRI-only radiotherapy.

Affective analysis of visual scenes using face pareidolia and scene-context

Mon 14 Nov 2022 08:49:01 AEDT

This study presents a new computer vision approach to perform affective analysis of a scene or object. The approach utilises a simulation of the phenomenon of face pareidolia that can be described as the perception of non-existent faces, for example, in random textures, clouds or rock formations. The emergence of face pareidolia in product designs and natural scenes can modulate affective perception of our everyday experiences without our conscious awareness. We propose a new deep learning method to simulate the face pareidolia ability of humans and predict associated emotional responses in two-dimensional valence and arousal space. Starting from a face detector that was trained on images of human faces, we propose a novel cross-domain weakly supervised three-step progressive domain adaptation approach to simulate face pareidolia by fine-tuning the human face detector on three types of synthetically generated sample images. Our approach fuses two deep network models, one model for predicting valence and arousal from abstract and minimal face-like patterns producing face pareidolia, and a second for recognising overall moods and context associated with the scene. To evaluate our approach, we constructed a new dataset containing instance-level annotations of face pareidolia occurrences as well as valence and arousal emotional values associated with the overall scene. The quantitative and qualitative experimental results of the present study demonstrate that our approach can outperform other state-of-the-art methods in face pareidolia detection as well as in predicting associated emotions.

Evaluation of colour models for computer vision using cluster validation techniques

Mon 13 Nov 2023 08:51:40 AEDT

Computer vision systems frequently employ colour segmentation as a step of feature extraction. This is particularly crucial in an environment where important features are colour-coded, such as robot soccer. This paper describes a method for determining an appropriate colour model by measuring the compactness and separation of clusters produced by the k-means algorithm. RGB, HSV, YC b Cr and CIE L*a*b* colour models are assessed for a selection of artificial and real images, utilising an implementation of the Dunn's-based cluster validation index. The effectiveness of the method is assessed by qualitatively comparing the relative correctness of the segmentation to the results of the cluster validation. Results demonstrate a significant variation in segmentation quality among colour spaces, and that YC b Cr is the best choice for the DARwIn-OP platform tested.

Evolutionary Hyperparameter Optimisation for Sentence Classification

Mon 11 Jul 2022 13:55:35 AEST

The performance that Deep Neural Networks can achieve on a specific task is impacted significantly by the hyperparameters selected. In order to compare the performance of various Deep Neural Network architectures on sentence classification, an optimised set of hyperparameters needs to be found for each architecture. In this work we use a simple Genetic Algorithm to optimise the hyperparameters of three different architectures and we evaluate their performance on a suite of sentence classification benchmarks. We found that a single Genetic Algorithm is capable of optimising a variety of different architectures and that the evolved configurations found can compete with those chosen by experts while using fewer overall trainable parameters. The three architectures tested are a recurrent neural network and two types of convolutional networks with a large difference in complexity. Of the three architectures, optimised for sentence classification, a simple Convolutional Neural Network was the overall best performer consistently achieving good performance while using very few trainable parameters.

Adjacent network for semantic segmentation of liver CT scans

Mon 09 Sep 2019 11:49:33 AEST

Fully convolutional neural networks have shown remarkable success in performing semantic segmentation. The use of convolutional layers for the entire architecture and skip connections to combine different resolution features or predictions have been adopted in successful networks, such as U-Net and DenseNet. However, these models employ several max-pooling layers that cause the network to lose spatial information and require them to mimic an autoencoder architecture to perform semantic segmentation at the original input resolution. In this paper, we propose a network that extracts features automatically with convolutional layers, like the fully convolutional neural network, but retains the spatial information of each of the extracted features. It then utilises the extracted features to make predictions with an efficient upsampling method. We evaluate the network performance on a liver segmentation task where it performs with comparable accuracy to other state-of-the-art networks while being much smaller in terms of the number of parameters as well as faster in computation time.

The Impact of Image Resolution on Facial Expression Analysis with CNNs

Mon 07 Nov 2022 09:51:17 AEDT

While deep learning has achieved state-of-the-art results on many computer vision tasks it is still challenged when interpreting human facial expressions, namely: poor generalisation ability of models across datasets, failure to account for individual differences in similar emotional states, and inability to recognise compound facial expressions and low-intensity or subtle emotional states. This study analyses how the resolution of face images that are input to various Convolutional Neural Network (CNN) models impacts on their ability to recognise compound and low-intensity emotions. Several high-resolution facial expression databases were combined to compile a simple dataset containing high-intensity emotions and a complex data set consisting of compound and low-intensity emotions. In the experiments, standard pre-trained CNN models that were further fine-tuned achieved higher validation accuracies than CNN models that were trained from scratch on the simple data set. However, when tested on the complex data the models trained from scratch generalised better than the refined pre-trained models. Using a technique of output visualisation we could show how our high-resolution CNN models were able to generalise to the complex data where they utilised small facial features that previously were not detectable.

From face recognition to facial pareidolia: Analysing hiddenneuron activations in cnns for cross-depiction recognition

Mon 07 Nov 2022 09:51:07 AEDT

The imagination of non-existent faces in random patterns, clouds and rock formations is known as facial pareidolia. We show that facial pareidolia also occurs naturally in a standard Convolutional Neural Network (CNN) trained on face recognition. For achieving this we propose a new method to analyse CNNs that combines feature visualisation and dimensionality reduction methods to cluster the hidden neuron activations in convolutional layers into groups with discriminative roles. The main contributions of the present paper are 1.) an approach that uses a CNN trained on human face detection for facial pareidolia simulation without any additional training on a target image set of abstract facial patterns and 2.) a novel way of improving the generalisation capacity of a CNN for cross-depiction recognition and domain adaptation scenarios using features learned by hidden neurons.

End-to-end speech emotion recognition based on time and frequency information using deep neural networks

Fri 24 Mar 2023 16:04:46 AEDT

We propose a speech emotion recognition system based on deep neural networks, operating on raw speech data in an end-to-end manner to predict continuous emotions in arousal-valence space. The model is trained using time and frequency information of speech recordings of the publicly available part of the multi-modal RECOLA database. We use the Concordance Correlation Coefficient (CCC) as it was proposed by the Audio-Visual Emotion Challenges to measure the similarity between the network prediction and gold-standard. The CCC prediction results of our model outperform the results achieved by other state-of-the-art end-to-end models. The innovative aspect of our study is an end-to-end approach to using data that previously was mostly used by approaches involving combinations of pre-processing or post-processing. Our study used only a small subset of the RECOLA dataset and obtained better results than previous studies that used the full dataset.

Joint Optimization of Topology and Hyperparameters of Hybrid DNNs for Sentence Classification

Fri 20 Oct 2023 09:24:32 AEDT

Deep Neural Networks (DNN) require specifically tuned architectures and hyperparameters when being applied to any given task. Nature-inspired algorithms have been successfully applied for optimising various hyperparameters in different types of DNNs such as convolutional and recurrent for sentence classification. Hybrid networks, which contain multiple types of neural architectures have more recently been used for sentence classification in order to achieve better performance. However, the inclusion of hybrid architectures creates numerous possibilities of designing the network and those sub-networks also need fine-tuning. At present these hybrid networks are designed manually and various organisation attempts are noticed. In order to understand the benefit and the best design principle of such hybrid DNNs for sentence classification, in this work we used an Evolutionary Algorithm (EA) to optimise the topology and various hyperparameters in different types of layers within the network. In our experiments, the proposed EA designed the hybrid networks by using a single dataset and evaluated the evolved networks on multiple other datasets to validate their generalisation capability. We compared the EA-designed hybrid networks with human-designed hybrid networks in addition to other EA-optimised and expert-designed non-hybrid architectures.

4D 128 cube topological dataset

Fri 17 Nov 2023 11:41:11 AEDT

The materials in this repository comprise a 4D topological dataset and two 4D visualisations in video format. The dataset can be downloaded directly as a series of twenty 1.5GB .zip files. More details can be found in the README.txt file. This data is also available via mediated access in the following formats: A single file approximately 30GB .zip file, or a series of ten 3GB .zip files.

All code and data released with this supplementary material uses the following license: Creative Commons Attribution 4.0 International (CC BY 4.0) http://creativecommons.org/licenses/by/4.0 This license permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original authors & provide a link to the Creative Commons license, and indicate if changes were made.

Please contact researchdata@newcastle.edu.au to access the data in alternative formats.

Investigation of a water equivalent depth method for dosimetric accuracy evaluation of synthetic CT

Fri 15 Sep 2023 17:46:38 AEST

Purpose: To provide a metric that reflects the dosimetric utility of the synthetic CT (sCT) and can be rapidly determined. Methods: Retrospective CT and atlas-based sCT of 62 (53 IMRT and 9 VMAT) prostate cancer patients were used. For image similarity measurements, the sCT and reference CT (rCT) were aligned using clinical registration parameters. Conventional image similarity metrics including the mean absolute error (MAE) and mean error (ME) were calculated. The water equivalent depth (WED) was automatically determined for each patient on the rCT and sCT as the distance from the skin surface to the treatment plan isocentre at 36 equidistant gantry angles, and the mean WED difference (ΔWED¯) between the two scans was calculated. Doses were calculated on each scan pair for the clinical plan in the treatment planning system. The image similarity measurements and ΔWED- were then compared to the isocentre dose difference(ΔDiso) between the two scans. Results: While no particular relationship to dose was observed for the other image similarity metrics, the ME results showed a linear trend against ΔDiso with R2 = 0.6, and the 95 % prediction interval for ΔDiso between −1.2 and 1 %. The ΔWED¯results showed an improved linear trend (R2 = 0.8) with a narrower 95 % prediction interval from −0.8 % to 0.8 %. Conclusion: ΔWED¯ highly correlates with ΔDiso for the reference and synthetic CT scans. This is easy to calculate automatically and does not require time-consuming dose calculations. Therefore, it can facilitate the process of developing and evaluating new sCT generation algorithms.

DQR: deep q-routing in software defined networks

Fri 14 Oct 2022 08:50:48 AEDT

In this paper, we investigate the task of quality of service (QoS) routing in software defined networks (SDN). We consider delay, bandwidth, loss, and cost as QoS parameters. We propose a new deep reinforcement learning solution for greedy online QoS routing in SDN and call it Deep Q-Routing (DQR). DQR utilises a dueling deep Q-network with prioritised experience replay to compute a path for any source-destination pair request in the presence of multiple QoS metrics. In contrast to existing DRL-based routing methods, the proposed DQR method regards the task of routing as a discrete control problem and uses a reward function comprising weighted QoS parameters. Our simulation results show that DQR substantially improves end-to-end throughput compared to other existing learning based methods.

A bio-inspired clustering model for anomaly detection in the mining industry

Fri 10 Nov 2023 15:25:20 AEDT

Being able to detect anomalies for predicting machine breakdown is of critical importance in the mining industry. These anomalies are usually found in unlabelled sensor data, and therefore unsupervised models represent the preferred choice for the task. In this chapter, we propose the use of a bio-inspired clustering model based on the self-organizing map (SOM) to detect anomalies in real-world data provided by Joy Global, a manufacturer of high-productivity mining solutions. The proposed SOM is compared to two other well-known clustering models, namely k-means and fuzzy c-means. Simulation experiments using grease cycle data from the manufacturer show that the SOM is able to detect a more reasonable number of anomalies than k-means and fuzzy c-means. Based on real scenarios given by Joy Global, we devise a simple way to prevent machine failures by triggering alarms through the anomalies detected, and the SOM is again shown to be more capable of identifying incidents of potential machine breakdown compared to the other two clustering models.

Classifying coke using CT scans and landmark multidimensional scaling

Fri 08 Sep 2023 15:05:41 AEST

One factor that limits development of fundamental research on the influence of coke microstructure on its strength is the difficulty in quantifying the way that microstructure is both classified and distributed in three dimensions. To support such fundamental studies, this study evaluated a novel volumetric approach for classifying small (approx. 450 μm3) blocks of coke microstructure from 3D computed tomography scans. An automated process for classifying microstructure blocks was described. It is based on Landmark Multi-Dimensional Scaling and uses the Bhattacharyya metric and k-means clustering. The approach was evaluated using 27 coke samples across a range of coke with different properties and reliably identified 6 ordered class of coke microstructure based on the distribution of voxel intensities associated with structural density. The lower class (1–2) subblocks tend to be dominated by pores and thin walls. Typically, there is an increase in wall thickness and reduced pore sizes in the higher classes. Inert features are also likely to be seen in higher classes (5–6). In general, this approach provides an efficient automated means for identifying the 3D spatial distribution of microstructure in CT scans of coke.

Fast automatic optimisation of CNN architectures for image classification using genetic algorithm

Fri 04 Oct 2019 13:51:54 AEST

Convolutional Neural Networks (CNNs) are currently the most prominent deep neural network models and have been used with great success for image classification and other applications. The performance of CNNs depends on their architecture and hyperparameter settings. Early CNN models like LeNet and AlexNet were manually designed by experienced researchers. The empirical design and optimisation of a new CNN architecture require a lot of expertise and can be very time-consuming. In this paper, we propose a genetic algorithm that can, for a given image processing task, efficiently explore a defined space of potentially suitable CNN architectures and simultaneously optimise their hyperparameters. We named this fast automatic optimisation model fast-CNN and employed it to find competitive CNN architectures for image classification on CIFAR10. In a series of comparative simulation experiments we could demonstrate that the network designed by fast-CNN achieved nearly as good accuracy as some of the other best network models available but fast-CNN took significantly less time to evolve. The trained fast-CNN network model also generalised well to CIFAR100.

Multimodal emotion recognition based on speech and physiological signals using deep neural networks

Fri 03 Jun 2022 15:35:27 AEST

A suitable combination of data in a multimodal emotion recognition model allows conveying and combining each channel's information to achieve a better recognition of the encoded emotion than would be possible using only a single modality and channel. In this paper, we focus on combining speech and physiological signals to predict the arousal and valence levels of the emotional states of a person. We designed a neural network that can use the information from raw audio signals, electrocardiograms, heart rate variability, electro-dermal activity, and skin conductance levels, to predict emotional states. The proposed deep neural network architecture works as an end-to-end process, which means, neither any pre-processing of the input data nor post-processing of the prediction of the network was applied. Using the data of the modalities available in the publicly accessible part of the RECOLA database, we achieved results comparable to other state-of-the-art approaches.