Macquarie University

Semantic relation prediction aims to mine the implicit relationships between objects in heterogeneous graphs, which consist of different types of objects and different types of links. In real-world scenarios, new semantic relations constantly emerge and they typically appear with only a few labeled data. Since a variety of semantic relations exist in multiple heterogeneous graphs, the transferable knowledge can be mined from some existing semantic relations to help predict the new semantic relations with few labeled data. This inspires a novel problem of few-shot semantic relation prediction across heterogeneous graphs. However, the existing methods cannot solve this problem because they not only require a large number of labeled samples as input, but also focus on a single graph with a fixed heterogeneity. Targeting this novel and challenging problem, in this paper, we propose a Meta-learning based Graph neural network for Semantic relation prediction, named MetaGS. Firstly, MetaGS decomposes the graph structure between objects into multiple normalized subgraphs, then adopts a two-view graph neural network to capture local heterogeneous information and global structure information of these subgraphs. Secondly, MetaGS aggregates the information of these subgraphs with a hyper-prototypical network, which can learn from existing semantic relations and adapt to new semantic relations. Thirdly, using the well-initialized two-view graph neural network and hyper-prototypical network, MetaGS can effectively learn new semantic relations from different graphs while overcoming the limitation of few labeled data. Extensive experiments on three real-world datasets have demonstrated the superior performance of MetaGS over the state-of-the-art methods.

Sequential recommendation, which aims to recommend next item that the user will likely interact in a near future, has become essential in various Internet applications. Existing methods usually consider the transition patterns between items, but ignore the transition patterns between features of items. We argue that only the item-level sequences cannot reveal the full sequential patterns, while explicit and implicit feature-level sequences can help extract the full sequential patterns. Meanwhile, the item-level sequential recommendation also suffers from limited supervised signal issues. In this paper, we propose a novel model Feature-level Deeper Self-Attention Network with Contrastive Learning (FDSA-CL) for sequential recommendation. Specifically, FDSA-CL first integrates various heterogeneous features of items into feature-level sequences with different weights through a vanilla attention mechanism. After that, FDSA-CL applies separated self-attention blocks on item-level sequences and feature-level sequences, respectively, to model item transition patterns and feature transition patterns. Moreover, we propose contrastive learning and item feature recommendation tasks to capture the embedding commonality and further utilize the beneficial interaction among the two levels, so as to alleviate the sparsity of the supervised signal and extract the most critical information. Finally, we jointly optimize the above tasks. We evaluate the proposed model using two real-world datasets and experimental results show that our model significantly outperforms the state-of-the-art approaches.

There are increasing concerns about malicious attacks on autonomous vehicles. In particular, inaudible voice command attacks pose a significant threat as voice commands become available in autonomous driving systems. How to empirically defend against these inaudible attacks remains an open question. Previous research investigates utilizing deep learning-based multimodal fusion for defense, without considering the model uncertainty in trustworthiness. As deep learning has been applied to increasingly sensitive tasks, uncertainty measurement is crucial in helping improve model robustness, especially in mission-critical scenarios. In this paper, we propose the Multimodal Fusion Framework (MFF) as an intelligent security system to defend against inaudible voice command attacks. MFF fuses heterogeneous audio-vision modalities using VGG family neural networks and achieves the detection accuracy of 92.25% in the comparative fusion method empirical study. Additionally, extensive experiments on audio-vision tasks reveal the model’s uncertainty. Using Expected Calibration Errors, we measure calibration errors and Monte-Carlo Dropout to estimate the predictive distribution for the proposed models. Our findings show empirically to train robust multimodal models, improve standard accuracy and provide a further step toward interpretability. Finally, we discuss the pros and cons of our approach and its applicability for Advanced Driver Assistance Systems.

Geoid anomalies offer crucial information on the internal density structure of the Earth, and thus, on its constitution and dynamic state. In order to interpret geoid undulations in terms of depth, magnitude and lateral extension of density anomalies in the lithosphere and upper mantle, the effects of lower mantle density anomalies need to be removed from the full geoid (thus obtaining the residual ’upper mantle geoid’). However, how to achieve this seemingly simple filtering exercise has eluded consensus for decades in the solid Earth community. While there is wide agreement regarding the causative masses of degrees > 10 in spherical harmonic expansions of the upper mantle geoid, those contributing to degrees < 7‐8 remain ambiguous. Here we use spherical harmonic analysis and recent tomography and density models from joint seismic‐geodynamic inversions to derive a representative upper mantle geoid, including the contributions from low harmonic degrees. We show that the upper mantle geoid contains important contributions from degrees 5 and 6 and interpret the causative masses as arising from the coupling between the long‐wavelength lithospheric structure and the sublithospheric upper mantle convection pattern. Importantly, the contributions from degrees 3 < l < 8 do not show a simple power‐law behaviour (e.g. Kaula’s rule), which precludes the use of standard filtering techniques in the spectral domain. Our upper mantle geoid model will be useful in studies of i) lithospheric structure, ii) dynamic topography and mantle viscosity, iii) lithosphere‐asthenosphere interactions and iv) the global stress field within the lithosphere and its associated hazards.

Background and aims The prominent cognitive-behavioral model of hoarding posits that information processing deficits contribute to hoarding disorder. Although individuals with hoarding symptoms consistently self-report attentional and impulsivity difficulties, neuropsychological tests have inconsistently identified impairments. These mixed findings may be the result of using different neuropsychological tests, tests with poor psychometric properties, and/or testing individuals in a context that drastically differs from their own homes. Methods One hundred twenty-three participants (hoarding = 63; control = 60) completed neuropsychological tests of sustained attention, focused attention, and response inhibition in cluttered and tidy environments in a counterbalanced order. Results Hoarding participants demonstrated poorer sustained attention and response inhibition than the control group (CPT-3 Omission and VST scores) and poorer response inhibition in the cluttered environment than when in the tidy environment (VST scores). CPT-3 Detectability and Commission scores also indicated that hoarding participants had greater difficulty sustaining attention and inhibiting responses than the control group; however, these effect sizes were just below the lowest practically meaningful magnitude. Posthoc exploratory analyses demonstrated that fewer than one-third of hoarding participants demonstrated sustained attention and response inhibition difficulties and that these participants reported greater hoarding severity and greater distress in the cluttered room. Discussion and conclusions Given these findings and other studies showing that attentional difficulties may be a transdiagnostic factor for psychopathology, future studies will want to explore whether greater sustained attention and response inhibition difficulties in real life contexts contribute to comorbidity and functional impairment in hoarding disorder.

In this paper, an efficient, coplanar waveguide (CPW)-fed printed circular ring fractal ultra-wideband (UWB) antenna is presented for biomedical applications. In UWB technology, short-range wireless communication is possible with low transceiving power, a characteristic that is particularly advantageous in the context of microwave and millimeter-wave (mmWave) medical imaging. In the proposed antenna configuration, the UWB response is achieved by introducing wedged slots in the radiating patch, designed on a low-loss substrate. A CPW partial ground plane is truncated from the edges to optimize the antenna impedance. Experimental results indicate the antenna’s robust performance across the frequency range of 3.2–20 GHz. The well-matched measured and simulated results confirm our contribution’s employability. Furthermore, a time-domain study offers valuable insights into how the antenna responds to transient signals, highlighting its responsiveness and adaptability to biomedical applications.

Minimally invasive skull base approaches relying on microscopes, endoscopes, and/or exoscopes have become a surgical workhorse for pathologies of the anterior skull base. Surgical robotics has recently been adopted for applications not limited to oncological operations in the abdominal/pelvic and head and neck areas and has tremendous potential in skull base surgery in the anterior cranial fossa. In this chapter, we present a brief historical perspective on anterior cranial fossa operations followed by an in-depth exploration of current developments in robotic surgery as applied to pathologies of the anterior cranial fossa. Both clinical and cadaveric studies are included for discussion of anatomical approaches and surgical indications. Finally, current limitations and future directions are explored to identify avenues that could one day allow robotic surgery to become the workhorse of skull base surgery within the anterior cranial fossa.

Secure environmental contexts are crucial for hominin interpretation and comparison. The discovery of a Denisovan individual and associated fauna at Tam Ngu Hao 2 (Cobra) Cave, Laos, dating back to 164–131 ka, allows for environmental comparisons between this (sub)tropical site and the Palearctic Denisovan sites of Denisova Cave (Russia) and Baishiya Karst Cave (China). Denisovans from northern latitudes foraged in a mix of forested and open landscapes, including tundra and steppe. Using stable isotope values from the Cobra Cave assemblage, we demonstrate that, despite the presence of nearby canopy forests, the Denisovan individual from Cobra Cave primarily consumed plants and/or animals from open forests and savannah. Using faunal evidence and proxy indicators of climates, results herein highlight a local expansion of rainforest at ~ 130 ka, raising questions about how Denisovans responded to this local climate change. Comparing the diet and habitat of the archaic hominin from Cobra Cave with those of early Homo sapiens from Tam Pà Ling Cave (46–43 ka), Laos, it appears that only our species was able to exploit rainforest resources.

Australian tidal wetlands differ in important respects to better studied northern hemisphere systems, an artefact stable to falling sea levels over millennia. A network of Surface Elevation Table-Marker Horizon (SET-MH) monitoring stations has been established across the continent to assess accretionary and elevation responses to sea-level rise. This network currently consists of 289 SET-MH installations across all mainland Australian coastal states and territories. SET-MH installations are mostly in mangrove forests but also cover a range of tidal marsh and supratidal forest ecosystems. Mangroves were found to have higher rates of accretion and elevation gain than all the other categories of tidal wetland, a result attributable to their lower position within the tidal frame (promoting higher rates of accretion) higher biomass (with potentially higher rates of root growth), and lower rates of organic decomposition. While Australian tidal marshes in general show an increase in elevation over time, in 80% of locations, this was lower than the rate of sea-level rise. High rates of accretion did not translate into high rates of elevation gain, because the rate of subsidence in the shallow substrate increased with higher accretion rates ( r ² = 0.87). The Australian SET-MH network, already in many locations spanning two decades of measurement, provides an important benchmark against which to assess wetland responses to accelerating sea-level rise in the decades ahead.

Indoor intrusion detection is a critical task for home security. Previous works in intrusion detection suffer from the problems such as blind spots in non-line-of-sight (NLOS) areas, restricted device locations, massive offline training required, and privacy concern. In this paper, we design and implement an omnidirectional indoor intrusion detection system, named AudioGuard , using only a pair of speaker and microphone. AudioGuard is able to detect both line-of-sight (LOS) and NLOS intrusions. Our observation of acoustic signal propagation in an indoor environment shows that there exist abundant multipath reflections and human movement introduces Doppler shift in echo signals. We hence capture periodical Doppler shift caused by intruder's walking motion to detect intrusion. Specifically, we first extract the Doppler shift embedded in echo signals, we then propose a periodicity polarization method to cancel out the impact of the change of radial angle and the distance on periodicity of Doppler shift. Finally, we detect intrusion by measuring periodicity of Doppler shift over time. Extensive experiments show that AudioGuard achieves a miss report rate of 0% and 1.75% for LOS and NLOS intrusion, respectively, and a false alarm rate of 4.17%.

Electrocardiogram (ECG) monitoring has been widely explored in detecting and diagnosing cardiovascular diseases due to its accuracy, simplicity, and sensitivity. However, medical- or commercial-grade ECG monitoring devices can be costly for people who want to monitor their ECG on a daily basis. These devices typically require several electrodes to be attached to the human body which is inconvenient for continuous monitoring. To enable low-cost measurement of ECG signals with off-the-shelf devices on a daily basis, in this paper, we propose a novel ECG sensing system that uses acceleration data collected from a smartphone. Our system offers several advantages over previous systems, including low cost, ease of use, location and user independence, and high accuracy. We design a two-tiered denoising process, comprising SWT and Soft-Thresholding, to effectively eliminate interference caused by respiration, body, and hand movements. Finally, we develop a multi-level deep learning recovery model to achieve efficient, real-time and user-independent ECG measurement on commercial mobile phones. We conduct extensive experiments with 30 participants (with nearly 36,000 heartbeat samples) under a user-independent scenario. The average errors of the PR interval, QRS interval, QT interval, and RR interval are 12.02 ms, 16.9 ms, 16.64 ms, and 1.84 ms, respectively. As a case study, we also demonstrate the strong capability of our system in signal recovery for patients with common heart diseases, including tachycardia, bradycardia, arrhythmia, unstable angina, and myocardial infarction.

Ant species exhibit behavioural commonalities when solving navigational challenges for successful orientation and to reach goal locations. These behaviours rely on a shared toolbox of navigational strategies that guide individuals under an array of motivational contexts. The mechanisms that support these behaviours, however, are tuned to each species' habitat and ecology with some exhibiting unique navigational behaviours. This leads to clear differences in how ant navigators rely on this shared toolbox to reach goals. Species with hybrid foraging structures, which navigate partially upon a pheromone-marked column, express distinct differences in their toolbox, compared to solitary foragers. Here, we explore the navigational abilities of the Western Thatching ant (Formica obscuripes), a hybrid foraging species whose navigational mechanisms have not been studied. We characterise their reliance on both the visual panorama and a path integrator for orientation, with the pheromone's presence acting as a non-directional reassurance cue, promoting continued orientation based on other strategies. This species also displays backtracking behaviour, which occurs with a combination of unfamiliar terrestrial cues and the absence of the pheromone, thus operating based upon a combination of the individual mechanisms observed in solitarily and socially foraging species. We also characterise a new form of goalless orientation in these ants, an initial retreating behaviour that is modulated by the forager's path integration system. The behaviour directs disturbed inbound foragers back along their outbound path for a short distance before recovering and reorienting back to the nest.

This study explores the changing patterns of the length of stay (LOS) at Australian residential aged care facilities during 2008–2018 and likely trends up to 2040. The expected LOS was estimated via the hazard function of exiting from such a facility and its heterogeneity by residents’ sociodemographic characteristics using an improved Cox regression model. Data were sourced from the Australian Institute of Health and Welfare. In-sample modelling results reveal that the estimated LOS differed by age (in general, shorter for older groups), marital status (longer for the widowed) and sex (longer for females). In addition, the estimated LOS increased slowly from 2008–2009 to 2016–2017 but declined steadily thereafter. Out-of-sample predictions suggest that the declining trend of the estimated LOS will continue until 2040 and that the longest LOS (approximately 37 months) will be observed among widowed females aged 50–79 years. Relative uncertainty measures are provided. The results portray the current changing landscape and the future trend of residential aged care use in Australia, which can inform the development of optimised residential aged care policies to support ageing Australians more effectively.

Malware authors often use cryptographic tools such as XOR encryption and block ciphers like AES to obfuscate part of the malware to evade detection. Use of cryptography may give the impression that these obfuscation techniques have some provable guarantees of success. In this paper, we take a closer look at the use of cryptographic tools to obfuscate malware. We first find that most techniques are easy to defeat (in principle), since the decryption algorithm and the key is shipped within the program. In order to clearly define an obfuscation technique’s potential to evade detection we propose a principled definition of malware obfuscation, and then categorize instances of malware obfuscation that use cryptographic tools into those which evade detection and those which are detectable. We find that schemes that are hard to de-obfuscate necessarily rely on a construct based on environmental keying. We also show that cryptographic notions of obfuscation, e.g., indistinghuishability and virtual black box obfuscation, may not guarantee evasion detection under our model. However, they can be used in conjunction with environmental keying to produce hard to de-obfuscate version of programs.

The present study asked whether oral vocabulary training can facilitate reading in a second language (L2). Fifty L2 speakers of English received oral training over three days on complex novel words, with predictable and unpredictable spellings, composed of novel stems and existing suffixes (i.e., vishing, vishes, vished). After training, participants read the novel word stems for the first time (i.e., trained and untrained), embedded in sentences, and their eye movements were monitored. The eye-tracking data revealed shorter looking times for trained than untrained stems, and for stems with predictable than unpredictable spellings. In contrast to monolingual speakers of English, the interaction between training and spelling predictability was not significant, suggesting that L2 speakers did not generate orthographic skeletons that were robust enough to affect their eye-movement behaviour when seeing the trained novel words for the first time in print.

Two or more independent species lineages can fuse through an evolutionary transition to form a single lineage, such as in the case of eukaryotic cells, lichens, and coral. The fusion of two or more independent lineages requires intermediary steps of increasing selective interdependence between these lineages. We argue a precursory selective regime of such a transition can be Multilevel Selection 1 (MLS1). We propose that intraspecies MLS1 can be extended to ecological multispecies arrangements. We develop a trait group selection (MLS1) model applicable to multispecies mutualistic interactions. We then explore conditions under which such a model could apply to mutualistic relationships between pollinators and plants. We propose that MLS1 could drive transitions towards higher interdependency between mutualists and stabilise obligate mutualisms in the face of invasion by cheater variants. This represents a radical extension of multilevel selection theory, applying it to the evolution of multispecies populations, and indicating new avenues for researching ecological community evolution.