When one or several classes are much less prevalent than another class (unbalanced data), class error rates and variable importances of the machine learning algorithm random forest can be biased, particularly when sample sizes are smaller, imbalance levels higher, and effect sizes of important variables smaller. Using simulated data varying in size, imbalance level, number of true variables, their effect sizes, and the strength of multicollinearity between covariates, we evaluated how eight versions of random forest ranked and selected true variables out of a large number of covariates despite class imbalance. The version that calculated variable importance based on the area under the curve (AUC) was least adversely affected by class imbalance. For the same number of true variables, effect sizes, and multicollinearity between covariates, the AUC variable importance ranked true variables still highly at the lower sample sizes and higher imbalance levels at which the other seven versions no longer achieved high ranks for true variables. Conversely, using the Hellinger distance to split trees or downsampling the majority class already ranked true variables lower and more variably at the larger sample sizes and lower imbalance levels at which the other algorithms still ranked true variables highly. In variable selection, a higher proportion of true variables were identified when covariates were ranked by AUC importances and the proportion increased further when the AUC was used as the criterion in forward variable selection. In three case studies, known species–habitat relationships and their spatial scales were identified despite unbalanced data.

In order to enable widespread integration of solar energy into the power system, there is an increasing need to reduce the uncertainty associated with solar power output which requires major improvements in solar irradiance forecasting. While most recent works have addressed short-term (minutes or hours ahead) forecasting, through this work, we propose using deep sequence learning models for forecasting at longer lead times such as a week in advance, as this can play a significant role in future power system storage applications. Along with point forecasts, we also produce uncertainty estimates through probabilistic prediction and showcase the potential of our machine learning frameworks for a new and important application of longer lead time forecasting in this domain. Our study on the SURFRAD data over seven US cities compares various deep sequence models and the results are encouraging, demonstrating their superior performance against most benchmarks from the literature and a current machine learning-based probabilistic prediction baseline (previously applied to short-term solar forecasting).

In this manuscript, a scheme for neural-learning-enhanced Cartesian Admittance control is presented for a robotic manipulator to deal with dynamic environments with moving remote center of motion (RCM) constraints. Although some research has been implemented to address fixed constrained motion, the dynamic moving movement constraint is still challenging. Indeed, the moving active RCM constraints generate uncertain disturbance on the robot tool shaft with unknown dynamics. The neural-learning-enhanced decoupled controller with disturbance optimisation is employed and implemented to maintain the performance under the kinematic uncertain and dynamic uncertain generated. In addition, the admittance Cartesian control method is introduced to control the robot, providing compliant behaviour to an external force in its operational space. In this proposed framework, a neural-learning-enhanced disturbance observer is investigated to calculate the external factor operating on the end effector premised on generalised momentum in order to ensure accuracy. Finally, the experiments are implemented using a redundant robot to validate the efficacy of the suggested approach with moving RCM constraints.

Outside of community-led design projects, most participatory design processes initiated by a company or organisation maintain or even strengthen power imbalances between the design organisation and the community on whose purported behalf they are designing, further increasing the absencing experience. Radical participatory design (RPD) is a radically relational answer to the coloniality inherent in participatory design where the community members’ disappointment is greater due to the greater expectations and presencing potential of a ‘participatory design’ process. We introduce the term RPD to show how research and design processes can be truly participatory to the root or core. Instead of treating participatory design as a method, a way of conducting a method, or a methodology, we introduce RPD as a meta-methodology, a way of doing any methodology. We explicitly describe what participation means and compare and contrast design processes based on the amount of participation, creating a typology of participation. We introduce ‘designer as community member’, ‘community member as designer,’ and ‘community member as facilitator’ models and provide characteristics for the meta-methodology of RPD.

We compared climatic relationships to insurance loss across the inland Pacific Northwest region of the United States, using a design matrix methodology, to identify optimum temporal windows for climate variables by county in relationship to wheat insurance loss due to drought. The results of our temporal window construction for water availability variables (precipitation, temperature, evapotranspiration, and the Palmer drought severity index [PDSI]) identified spatial patterns across the study area that aligned with regional climate patterns, particularly with regards to drought-prone counties of eastern Washington. Using these optimum time-lagged correlational relationships between insurance loss and individual climate variables, along with commodity pricing, we constructed a regression-based random forest model for insurance loss prediction and evaluation of climatic feature importance. Our cross-validated model results indicated that PDSI was the most important factor in predicting total seasonal wheat/drought insurance loss, with wheat pricing and potential evapotranspiration having noted contributions. Our overall regional model had a $ {R}^2 $ of 0.49, and a RMSE of $30.8 million. Model performance typically underestimated annual losses, with moderate spatial variability in terms of performance between counties.

There are many types of approaches for Paraphrase Identification (PI), an NLP task of determining whether a sentence pair has equivalent semantics. Traditional approaches mainly consist of unsupervised learning and feature engineering, which are computationally inexpensive. However, their task performance is moderate nowadays. To seek a method that can preserve the low computational costs of traditional approaches but yield better task performance, we take an investigation into neural network-based transfer learning approaches. We discover that by improving the usage of parameters efficiently for feature-based transfer, our research goal can be accomplished. Regarding the improvement, we propose a pre-trained task-specific architecture. The fixed parameters of the pre-trained architecture can be shared by multiple classifiers with small additional parameters. As a result, the computational cost left involving parameter update is only generated from classifier-tuning: the features output from the architecture combined with lexical overlap features are fed into a single classifier for tuning. Furthermore, the pre-trained task-specific architecture can be applied to natural language inference and semantic textual similarity tasks as well. Such technical novelty leads to slight consumption of computational and memory resources for each task and is also conducive to power-efficient continual learning. The experimental results show that our proposed method is competitive with adapter-BERT (a parameter-efficient fine-tuning approach) over some tasks while consuming only 16% trainable parameters and saving 69-96% time for parameter update.

There has been considerable work recently in the natural language community and elsewhere on Responsible AI. Much of this work focuses on fairness and biases (henceforth Risks 1.0), following the 2016 best seller: Weapons of Math Destruction. Two books published in 2022, The Chaos Machine and Like, Comment, Subscribe, raise additional risks to public health/safety/security such as genocide, insurrection, polarized politics, vaccinations (henceforth, Risks 2.0). These books suggest that the use of machine learning to maximize engagement in social media has created a Frankenstein Monster that is exploiting human weaknesses with persuasive technology, the illusory truth effect, Pavlovian conditioning, and Skinner’s intermittent variable reinforcement. Just as we cannot expect tobacco companies to sell fewer cigarettes and prioritize public health ahead of profits, so too, it may be asking too much of companies (and countries) to stop trafficking in misinformation given that it is so effective and so insanely profitable (at least in the short term). Eventually, we believe the current chaos will end, like the lawlessness in Wild West, because chaos is bad for business. As computer scientists, this paper will summarize criticisms from other fields and focus on implications for computer science; we will not attempt to contribute to those other fields. There is quite a bit of work in computer science on these risks, especially on Risks 1.0 (bias and fairness), but more work is needed, especially on Risks 2.0 (addictive, dangerous, and deadly).

This article sets out some of the analytical moves that are necessary to developing a distinctive area of research called postcolonial memory studies. A key barrier to synthesising insights from postcolonial and memory studies has been a reductive approach to analogue and digital technologies which operate as vehicles for memory. Three analytical moves are needed to decentre, or at the very least de-naturalise the technological narratives and ecologies of Europe and the US. Media memory studies needs to draw more effectively on postcolonial studies to position mediated memory as inextricably connected to the legacies of colonialism and empire; develop a much broader account of media infrastructures emerging from what is increasingly characterised as ‘global media studies’; make an empirical and analytical shift away from the primacy of digital communications technologies and to explore technologies, not just as artefacts but as knowledge generating cultural practices. The combined value of these three shifts in approaches to media and communications technologies in memory studies research has considerable potential for developing postcolonial media memory studies research which offers a thorough and empirically grounded analysis of the complex ways in which the legacies of colonialism shape and structure the ways in which practices and performances of remembering are mediated in contemporary social life. This shift towards postcolonial memory studies can be seen as part of the wider project of what Anna (Amza) Reading has in this volume called ‘rewilding memory’ by rethinking ‘the underlying ecologies of knowledge within studies of memory’.

Despite the importance of diverse expertise in helping solve difficult interdisciplinary problems, measuring it is challenging and often relies on proxy measures and presumptive correlates of actual knowledge and experience. To address this challenge, we propose a text-based measure that uses researcher’s prior work to estimate their substantive expertise. These expertise estimates are then used to measure team-level expertise diversity by determining similarity or dissimilarity in members’ prior knowledge and skills. Using this measure on 2.8 million team invented patents granted by the US Patent Office, we show evidence of trends in expertise diversity over time and across team sizes, as well as its relationship with the quality and impact of a team’s innovation output.

The 2011 policy pivot of the German government, from extending nuclear power plants terms to securing their shutdown for 2022, cannot be explained without looking at how the German political discourse network shifted in the months following Fukushima. This paper seeks to model and identify mechanisms that help explain how the two-mode network of political actors’ support for claims developed. We identify possible mechanisms to explain discourse dynamics from literature on political discourse and discourse networks, and extend homophily mechanisms to two-mode networks as “tertius” effects. We then introduce and employ a multimodal extension of dynamic network actor models to answer two questions key to how the discourse has evolved: which actors support claims more frequently and which claims they support. Our results indicate that mechanisms vary according to the discursive phase, but that powerful actors participate in the discourse more often, and actors tend to support claims that have already found support by cross-party coalitions. These are the two most prominent mechanisms that help to explain the dramatic nuclear policy change in Germany after Fukushima.