BACKGROUND:Inter-segment joint angles can be obtained from inertial measurement units (IMUs); however, accurate 3D joint motion measurement, which requires sensor fusion and signal processing, sensor alignment with segments and joint axis calibration, can be challenging to achieve. RESEARCH QUESTION:Can an artificial neural network modeling framework be used for direct, real-time conversion of IMU data to joint angles during walking and running, and how does sensor number, location on the body and gait speed impact prediction accuracy? METHODS:Thirty healthy adult participants performed gait experiments in which kinematics data were obtained from self-placed IMUs and video motion analysis, the reference standard for joint kinematics. Data were collected during walking at 0.5 m/s, 1.0 m/s and 1.5 m/s, as well as during running at 2.0 m/s and 3.0 m/s. A generative adversarial network was trained and used to predict lower limb joint angles at all gait speeds using IMU data only, and prediction accuracy assessed using all combinations of sensors. RESULTS:Joint angle prediction accuracy was strongly dependent on the number and location of sensors, as well as walking and running speed. A single IMU could be used to predict sagittal plane joint angles at either the hip, knee or ankle during walking with RMS errors below 4.0°, though highest 3D joint motion accuracy was obtained with two or three IMUs for a given joint. SIGNIFICANCE:This study reports a modeling framework for direct conversion of IMU data to joint angles without signal processing or joint calibration. The findings suggest that combinations of up to four IMUs reproduce hip, knee and ankle joint kinematics simultaneously during walking and running with highest accuracy. The findings may be useful in maximizing accuracy of IMU-based motion measurements of the lower limb joints in applications such as remote monitoring of movement, sports training, and in rehabilitation.

The ability to discover meaningful, accurate, and concise mathematical equations that describe datasets is valuable across various domains. Equations offer explicit relationships between variables, enabling deeper insights into underlying data patterns. Most existing equation discovery methods rely on genetic programming, which iteratively searches the equation space but is often slow and prone to overfitting. By representing equations as directed acyclic graphs, we leverage the use of graph neural networks to learn the underlying semantics of equations, and generate new, previously unseen equations. Although graph generative models have been shown to be successful in discovering new types of graphs in many fields, there application in discovering equations remains largely unexplored. In this work, we propose Graph-EQ, a deep graph generative model designed for efficient equation discovery. Graph-EQ uses a conditional variational autoencoder (CVAE) to learn a rich latent representation of the equation space by training it on a large corpus of equations in an unsupervised manner. Instead of directly searching the equation space, we employ Bayesian optimization to efficiently explore this learned latent space. We show that the encoder-decoder architecture of Graph-Eq is able to accurately reconstruct input equations. Moreover, we show that the learned latent representation can be sampled and decoded into valid equations, including new and previously unseen equations in the training data. Finally, we assess Graph-Eq's ability to discover equations that best fit a dataset by exploring the latent space using Bayesian optimization. Latent space exploration is done on 20 dataset with known ground-truth equations, and Graph-Eq is shown to successfully discover the grountruth equation in the majority of datasets.

Existing neural architecture representation learning methods focus on continuous representation learning, typically using Variational Autoencoders (VAEs) to map discrete architectures onto a continuous Gaussian distribution. However, sampling from these spaces often leads to a high percentage of invalid or duplicate neural architectures, likely due to the unnatural mapping of inherently discrete architectural space onto a continuous space. In this work, we revisit architecture representation learning from a fundamentally discrete perspective. We propose Arch-VQ, a framework that learns a discrete latent space of neural architectures using a Vector-Quantized Variational Autoencoder (VQ-VAE), and models the latent prior with an autoregressive transformer. This formulation yields discrete architecture representations that are better aligned with the underlying search space while decoupling representation learning from prior modeling. Across NASBench-101, NASBench-201, and DARTS search spaces, Arch-VQ improves the quality of generated architectures, increasing the rate of valid and unique generations by 22

Inertial measurement units (IMUs) are wearable motion sensing devices that support low-cost human kinematics measurement outside of the laboratory; however, the impact of IMU motion measurement errors on estimates of upper limb muscle and joint force using musculoskeletal modeling remains poorly understood. The aims of this study were to measure upper limb kinematics using IMUs and optical motion analysis, and evaluate the differences in shoulder musculoskeletal model estimates of muscle and joint forces when using IMU-based motion analysis and optical motion analysis data. Thirty healthy adults performed activities of daily living at fast and slow speeds while upper limb motion measurement was simultaneously acquired using an optical motion analysis system and self-placed IMUs. Kinematics measurement accuracy using IMUs was highest in humerothoracic joint elevation and scapular lateral rotation, and lowest in humerothoracic plane of elevation and humeral axial rotation. Musculoskeletal model calculations of a muscle’s force were most sensitive to IMU motion measurement errors about the degree of freedom primarily actuated by that muscle. For example, subscapularis force calculations were more sensitive to errors in humeral axial rotation than humeral elevation. Consequently, errors in humeral axial rotation motion measurement affected infraspinatus and subscapularis muscle forces, and subsequently, estimates of the compressive component of the glenohumeral joint force. During fast flexion, a mean difference of 4.5 %BW in the compressive joint force was observed when using IMU motion measurement compared to optical motion analysis data (p < 0.001). The findings of this study provide guidance in the use of IMUs for muscle and joint force calculation during activities of daily living.

ABSTRACTLongitudinal studies that continuously generate data enable the capture of temporal variations in experimentally observed parameters, facilitating the interpretation of results in a time-aware manner. We propose IL-VIS (Incrementally Learned Visualizer), a new machine learning pipeline that incrementally learns and visualizes a progression trajectory representing the longitudinal changes in longitudinal studies. At each sampling time point in an experiment, IL-VIS generates a snapshot of the longitudinal process on the data observed thus far, a new feature that is beyond the reach of classical static models. We first verify the utility and correctness of IL-VIS using simulated data, for which the true progression trajectories are known. We find that it accurately captures and visualizes the trends and (dis)similarities between high-dimensional progression trajectories. We then apply IL-VIS to longitudinal Multi-Electrode Array data from brain cortical organoids when exposed to different levels of Quinolinic Acid, a metabolite contributing to many neuroinflammatory diseases including Alzheimer’s disease, and its blocking antibody. We uncover valuable insights into the organoids’ electrophysiological maturation and response patterns over time under these conditions.

Neural growth is the process of growing a small neural network to a large network and has been utilized to accelerate the training of deep neural networks. One crucial aspect of neural growth is determining the optimal growth timing. However, few studies investigate this systematically. Our study reveals that neural growth inherently exhibits a regularization effect, whose intensity is influenced by the chosen policy for growth timing. While this regularization effect may mitigate the overfitting risk of the model, it may lead to a notable accuracy drop when the model underfits. Yet, current approaches have not addressed this issue due to their lack of consideration of the regularization effect from neural growth. Motivated by these findings, we propose an under/over fitting risk-aware growth timing policy, which automatically adjusts the growth timing informed by the level of potential under/overfitting risks to address both risks. Comprehensive experiments conducted using CIFAR-10/100 and ImageNet datasets show that the proposed policy achieves accuracy improvements of up to 1.3% in models prone to underfitting while achieving similar accuracies in models suffering from overfitting compared to the existing methods.

The integration of Artificial Intelligence (AI) has catalysed a paradigm shift in agricultural practices, revolutionising critical aspects of contemporary farming methodologies. In recent years, discrimination has become a significant topic of controversy in the use of AI-based systems. As AI becomes increasingly integrated into agriculture, the potential for bias within these systems raises significant concerns, as biases can affect farming practices, crop management, resource allocation, and the broader agricultural economy. Mitigating these issues is vital to ensure equitable access to technology, enhance farming practices for all, and foster sustainable and inclusive food production systems. This manuscript provides an extensive review of the literature on biases and fairness in AI focusing on the field of agriculture and the need for a holistic and inclusive approach to mitigate AI biases, ensuring that AI advancements benefit a broad spectrum of farming communities while promoting sustainability and equity.

Regional solar power forecasting, which involves predicting the total power generation from all rooftop photovoltaic (PV) systems in a region holds significant importance for various stakeholders in the energy sector to ensure a stable electricity supply. However, the vast amount of solar power generation and weather time series from geographically dispersed locations that need to be considered in the forecasting process makes accurate regional forecasting challenging. Therefore, previous studies have limited the focus to either forecasting a single time series (i.e., aggregated time series) which is the addition of all solar generation time series in a region, disregarding the location-specific weather effects or forecasting solar generation time series of each PV site (i.e., individual time series) independently using location-specific weather data, resulting in a large number of forecasting models. In this work, we propose two new deep-learning-based regional forecasting methods that can effectively leverage both types of time series (aggregated and individual) with weather data in a region. We propose two hierarchical temporal convolutional neural network architectures (HTCNN A1 and A2) and two new strategies to adapt HTCNNs for regional solar power forecasting. In the first strategy, we explore generating a regional forecast using a single HTCNN. In the second, we divide the region into multiple sub-regions based on weather information and train separate HTCNNs for each sub-region; the forecasts of each sub-region are then added to generate a regional forecast. The proposed work is evaluated using a large dataset collected over a year from 101 locations across Western Australia to provide a day ahead forecast at an hourly time resolution which involves forecasting a horizon of 18 h. We compare our approaches with well-known alternative methods, including long short-term memory networks and convolution neural networks, and show that proposed HTCNN-based approaches require fewer individually trained networks. Furthermore, the sub-region-based HTCNN-based approach achieves a forecast skill score of 40.2% and reduces a statistically significant forecast error by 6.5% compared to the best-performing counterpart. Our results indicate that the proposed approaches are well-suited for forecasting applications covering large regions containing many individual solar PV systems in different locations.

Traditional machine learning is generally treated as a black-box optimization problem and does not typically produce interpretable functions that connect inputs and outputs. However, the ability to discover such interpretable functions is desirable. In this work, we propose GINN-LP, an interpretable neural network to discover the form and coefficients of the underlying equation of a dataset, when the equation is assumed to take the form of a multivariate Laurent Polynomial. This is facilitated by a new type of interpretable neural network block, named the "power-term approximator block", consisting of logarithmic and exponential activation functions. GINN-LP is end-to-end differentiable, making it possible to use backpropagation for training. We propose a neural network growth strategy that will enable finding the suitable number of terms in the Laurent polynomial that represents the data, along with sparsity regularization to promote the discovery of concise equations. To the best of our knowledge, this is the first model that can discover arbitrary multivariate Laurent polynomial terms without any prior information on the order. Our approach is first evaluated on a subset of data used in SRBench, a benchmark for symbolic regression. We first show that GINN-LP outperforms the state-of-the-art symbolic regression methods on datasets generated using 48 real-world equations in the form of multivariate Laurent polynomials. Next, we propose an ensemble method that combines our method with a high-performing symbolic regression method, enabling us to discover non-Laurent polynomial equations. We achieve state-of-the-art results in equation discovery, showing an absolute improvement of 7.1% over the best contender, by applying this ensemble method to 113 datasets within SRBench with known ground-truth equations.

Dual-energy X-ray absorptiometry (DXA) scans are one of the most frequently used imaging techniques for calculating bone mineral density, yet calculating fracture risk using DXA image features is rarely performed. The objective of this study was to combine deep neural networks, together with DXA images and patient clinical information, to evaluate fracture risk in a cohort of adults with at least one known fall and age-matched healthy controls. DXA images of the entire body as, well as isolated images of the hip, forearm, and spine (1488 total), were obtained from 478 fallers and 48 non-faller controls. A modeling pipeline was developed for fracture risk prediction using the DXA images and clinical data. First, self-supervised pretraining of feature extractors was performed using a small vision transformer (ViT-S) and a convolutional neural network model (VGG-16 and Resnet-50). After pretraining, the feature extractors were then paired with a multilayer perceptron model, which was used for fracture risk classification. Classification was achieved with an average area under the receiver-operating characteristic curve (AUROC) score of 74.3%. This study demonstrates ViT-S as a promising neural network technique for fracture risk classification using DXA scans. The findings have future application as a fracture risk screening tool for older adults at risk of falls. © 2023 The Authors. JBMR Plus published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research.

Inertial measurement units (IMUs) have become the mainstay in human motion evaluation outside of the laboratory; however, quantification of 3-dimensional upper limb motion using IMUs remains challenging. The objective of this systematic review is twofold. Firstly, to evaluate computational methods used to convert IMU data to joint angles in the upper limb, including for the scapulothoracic, humerothoracic, glenohumeral, and elbow joints; and secondly, to quantify the accuracy of these approaches when compared to optoelectronic motion analysis. Fifty-two studies were included. Maximum joint motion measurement accuracy from IMUs was achieved using Euler angle decomposition and Kalman-based filters. This resulted in differences between IMU and optoelectronic motion analysis of 4° across all degrees of freedom of humerothoracic movement. Higher accuracy has been achieved at the elbow joint with functional joint axis calibration tasks and the use of kinematic constraints on gyroscope data, resulting in RMS errors between IMU and optoelectronic motion for flexion–extension as low as 2°. For the glenohumeral joint, 3D joint motion has been described with RMS errors of 6° and higher. In contrast, scapulothoracic joint motion tracking yielded RMS errors in excess of 10° in the protraction–retraction and anterior-posterior tilt direction. The findings of this study demonstrate high-quality 3D humerothoracic and elbow joint motion measurement capability using IMUs and underscore the challenges of skin motion artifacts in scapulothoracic and glenohumeral joint motion analysis. Future studies ought to implement functional joint axis calibrations, and IMU-based scapula locators to address skin motion artifacts at the scapula, and explore the use of artificial neural networks and data-driven approaches to directly convert IMU data to joint angles.

Catastrophic forgetting; the loss of old knowledge upon acquiring new knowledge, is a pitfall faced by deep neural networks in real-world applications. Many prevailing solutions to this problem rely on storing exemplars (previously encountered data), which may not be feasible in applications with memory limitations or privacy constraints. Therefore, the recent focus has been on Non-Exemplar based Class Incremental Learning (NECIL) where a model incrementally learns about new classes without using any past exemplars. However, due to the lack of old data, NECIL methods struggle to discriminate between old and new classes causing their feature representations to overlap. We propose NAPA-VQ: Neighborhood Aware Prototype Augmentation with Vector Quantization, a framework that reduces this class overlap in NECIL. We draw inspiration from Neural Gas to learn the topological relationships in the feature space, identifying the neighboring classes that are most likely to get confused with each other. This neighborhood information is utilized to enforce strong separation between the neighboring classes as well as to generate old class representative prototypes that can better aid in obtaining a discriminative decision boundary between old and new classes. Our comprehensive experiments on CIFAR-100, TinyImageNet, and ImageNet-Subset demonstrate that NAPA-VQ outperforms the State-of-the-art NECIL methods by an average improvement of 5%, 2%, and 4% in accuracy and 10%, 3%, and 9% in forgetting respectively. Our code can be found in https://github.com/TamashaM/NAPA-VQ.git.

High throughput technologies used in experimental biological sciences produce data with a vast number of variables at a rapid pace, making large volumes of high-dimensional data available. The exploratory analysis of such high-dimensional data can be aided by human interpretable low-dimensional visualizations. This work investigates how both discrete and continuous structures in biological data can be captured using the recently proposed dimensionality reduction method SONG, and compares the results with commonly used methods UMAP and PHATE. Using simulated and real-world datasets, we observe that SONG produces insightful visualizations by preserving various patterns, including discrete clusters, continuums, and branching structures in all considered datasets. More importantly, for datasets containing both discrete and continuous structures, SONG performs better at preserving both the structures compared to UMAP and PHATE. Furthermore, our quantitative evaluation of the three methods using downstream analysis validates the on par quality of the SONG's low-dimensional embeddings compared to the other methods.

Automatically converting text descriptions into images using transformer architectures has recently received considerable attention. Such advances have implications for many applied design disciplines across fashion, art, architecture, urban planning, landscape design and the future tools available to such disciplines. However, a detailed analysis capturing the capabilities of such models, specifically with a focus on the built environment, has not been performed to date. In this work, we investigate the capabilities and biases of such text-to-image methods as it applies to the built environment in detail. We use a systematic grammar to generate queries related to the built environment and evaluate resulting generated images. We generate 1020 different images and find that text to image transformers are robust at generating realistic images across different domains for this use-case. Generated imagery can be found at the github: https://github.com/sachith500/DALLEURBAN

Joint angle quantification from inertial measurement units (IMUs) is commonly performed using kinematic modelling, which depends on anatomical sensor placement and/or functional joint calibration; however, accurate three-dimensional joint motion measurement remains challenging to achieve. The aims of this study were firstly to employ deep neural networks to convert IMU data to ankle joint angles that are indistinguishable from those derived from motion capture-based inverse kinematics (IK) - the reference standard; and secondly, to validate the robustness of this approach across contrasting walking speeds in healthy individuals. Kinematics data were simultaneously calculated using IMUs and IK from 9 subjects during walking on a treadmill at 0.5 m/s, 1.0 m/s and 1.5 m/s. A generative adversarial network was trained using gait data at two of the walking speeds to predict ankle kinematics from IMU data alone for the third walking speed. There were significant differences between IK and IMU joint angle predictions for ankle eversion and internal rotation during walking at 0.5 m/s and 1.0 m/s (p < 0.001); however, no significant differences in joint angles were observed between the generative adversarial network prediction and IK at any speed or plane of joint motion (p < 0.05). The RMS difference in ankle joint kinematics between the generative adversarial network and IK for walking at 1.0 m/s was 3.8°, 2.1° and 3.5° for dorsiflexion, inversion and axial rotation, respectively. The modeling approach presented for real-time IMU to ankle joint angle conversion, which can be readily expanded to other joints, may provide enhanced IMU capability in applications such as telemedicine, remote monitoring and rehabilitation.

Nonparametric dimensionality reduction techniques, such as t-distributed Stochastic Neighbor Embedding (t-SNE) and uniform manifold approximation and projection (UMAP), are proficient in providing visualizations for data sets of fixed sizes. However, they cannot incrementally map and insert new data points into an already provided data visualization. We present self-organizing nebulous growths (SONG), a parametric nonlinear dimensionality reduction technique that supports incremental data visualization, i.e., incremental addition of new data while preserving the structure of the existing visualization. In addition, SONG is capable of handling new data increments, no matter whether they are similar or heterogeneous to the already observed data distribution. We test SONG on a variety of real and simulated data sets. The results show that SONG is superior to Parametric t-SNE, t-SNE, and UMAP in incremental data visualization. Especially, for heterogeneous increments, SONG improves over Parametric t-SNE by 14.98% on the Fashion MNIST data set and 49.73% on the MNIST data set regarding the cluster quality measured by the adjusted mutual information scores. On similar or homogeneous increments, the improvements are 8.36% and 42.26%, respectively. Furthermore, even when the abovementioned data sets are presented all at once, SONG performs better or comparable to UMAP and superior to t-SNE. We also demonstrate that the algorithmic foundations of SONG render it more tolerant to noise compared with UMAP and t-SNE, thus providing greater utility for data with high variance, high mixing of clusters, or noise.

Contemporary single-cell technologies produce data with a vast number of variables at a rapid pace, making large volumes of high-dimensional data available. The exploratory analysis of such high dimensional data can be aided by intuitive low dimensional visualizations. In this work, we investigate how both discrete and continuous structures in single cell data can be captured using the recently proposed dimensionality reduction method SONG, and compare the results with commonly used methods UMAP and PHATE. Using simulated and real-world datasets, we observed that SONG preserves a variety of patterns including discrete clusters, continuums, and branching structures. More importantly, SONG produced more/equally insightful visualizations compared to UMAP and PHATE in all considered datasets. We also quantitatively validate the high-dimensional pairwise distance preservation ability of these visualization methods in the low dimensional space for the generated visualizations.

Background Estimating the parameters that describe the ecology of viruses,particularly those that are novel, can be made possible using metagenomic approaches. However, the best-performing existing methods require databases to first estimate an average genome length of a viral community before being able to estimate other parameters, such as viral richness. Although this approach has been widely used, it can adversely skew results since the majority of viruses are yet to be catalogued in databases. Results In this paper, we present ENVirT, a method for estimating the richness of novel viral mixtures, and for the first time we also show that it is possible to simultaneously estimate the average genome length without a priori information. This is shown to be a significant improvement over database-dependent methods, since we can now robustly analyze samples that may include novel viral types under-represented in current databases. We demonstrate that the viral richness estimates produced by ENVirT are several orders of magnitude higher in accuracy than the estimates produced by existing methods named PHACCS and CatchAll when benchmarked against simulated data. We repeated the analysis of 20 metavirome samples using ENVirT, which produced results in close agreement with complementary in virto analyses. Conclusions These insights were previously not captured by existing computational methods. As such, ENVirT is shown to be an essential tool for enhancing our understanding of novel viral populations.

Customer churn prediction is one of the most useful areas of study in customer analytics. Due to the enormous amount of data available for such predictions, machine learning and data mining have been heavily used in this domain. There exist many machine learning algorithms directly applicable for the problem of customer churn prediction, and here, we attempt to experiment on a novel approach by using a cognitive learning based technique in an attempt to improve the results obtained by using a combination of supervised learning methods, with cognitive unsupervised learning methods. Keywords—Growing Self Organizing Maps, Kernel Methods,