Birmingham City University Open Access Repository
Not a member yet
8823 research outputs found
Sort by
A Streaming Approach to Data Discrepancy Detection and Adaptation in Deep Neural Networks
No full textDeep learning has achieved remarkable success over the last ten years. However, keeping Deep Neural Networks (DNNs) up to date with changing data remains at the forefront of creating genuinely practical systems. To address some aspects of this challenge, this thesis studies the detection of streaming changing data and the subsequent adaptation of DNNs. The main challenges are to efficiently detect the changes and update the DNN promptly without forgetting the pertinent older information. This presents more of a challenge for DNNs than for other traditional machine learning models as DNNs take longer to train, require more data and suffer from catastrophic forgetting where previously learnt classes are forgotten when the DNN is adapted to the new data. DNNs are typically used with high dimensional unstructured data as opposed to lower dimensional structured data, which the streaming literature has focused upon thus far. Hence, DNN adaptation has not been widely studied in the streaming machine learning literature.
So far, clustering is the preferred method for detecting changes in data however, this can be slow as a number of instances must be received before a change is detected. A DNN is usually considered as a ’black box’ where, given an input, it provides outputs, but the specific process by which it arrived at the output is not easily discernible. Inside this ’black box’ are many artificial neurons which output values called activations. This thesis investigates if these activations can be used as a different representation of the input data and used as the input to streaming machine learning models to assist in detecting changing data and in DNN adaptation.
To address this, initially, this thesis proposes a method that handles outlier detection, adding an extra classification of ’unknown’ to DNNs, so known classes are classified as their class and the detected changed data is classified as ’unknown’. Activations are extracted, providing a unique dynamic trajectory of activations for use with a streaming machine learning clustering model to detect outliers and label them as unknown. It is shown that the DNN activations can be used as a different representation of the input data and used with streaming machine learning techniques in order to detect outliers. Experiments show that our method outperforms the other leading open-set classification methods by a minimum of 2% and a maximum of 30% on the F1-Score and is between 5 and 100 times faster.
The outlier detection method leads onto offering the second contribution, which proposes a solution that handles the scenario of novel classes appearing in a stream of data (known as concept evolution) and DNN adaptation. A novel method of extracting DNN activations is used with an accuracy volatility concept evolution detection method and DNN adaptation process. Experiments show that our method outperforms other leading methods with regards to accuracy when placed in the concept evolution scenario with limited true-labelled data. The results of the experiments are analysed based on accuracy, speed of inference and speed of adaptation. On accuracy, our method outperforms the next best adaptation method by 27% and the next best combined novel class detection and CNN adaptation method by 24%. On speed, our method is within 1.5ms of the fastest inference speed and within 1.6s of the fastest DNN adaptation speed.
Thirdly, this thesis proposes a method that handles the more advanced problem of concept drift detection and DNN adaptation in drift pattern scenarios. DNN activations from multiple hidden layers are used with our novel ensemble drift detection and DNN adaptation method. Experiments show that our method overall outperforms other leading methods of detecting concept drift and DNN adaptation in all drift scenarios. We compare with eleven other leading methods of drift detection, adaptation and combined detection and adaptation methods. Our method outperforms other leading drift detection methods by between 8% and 46% on F1-Score, and other leading drift detection and adaptation methods by between 5% and 20% on accuracy. Our method is within 1.1ms of the fastest inference speed and 7 times faster than other adaptation methods.
These three methods culminate to provide the overall contributions of this thesis, which are: (1) The application of methods to extract activations from DNNs, providing more features than the input data, termed by us as activation classification footprints; and (2) applying these footprints to our own drift detection and DNN adaptation methods in order to detect and adapt to outliers, concept evolution and concept drift. The use of DNN activations in streaming machine learning models to detect data changes and in DNN adaptation offers a unique perspective in this research area and is a step forward towards realising fully adaptive continuous deep learning systems
Evoking Episodic and Semantic Details with Instructional Manipulation: the Semantic Autobiographical Interview
No full textMost measures of naturalistic human memory instruct participants to recall personally-experienced episodes in narrative format. These narratives contain non-episodic details, such as general knowledge of the world, or personal knowledge about one’s life circumstances that are elevated with aging. As this non-episodic content is incidental to the instructions, it is difficult to interpret. We modified the widely used Autobiographical Interview (AI) to create a Semantic Autobiographical Interview (SAI) that explicitly targets personal semantic (P-SAI) and general semantic memories (G-SAI). We tested the SAI in young and older adults, alongside with the original AI. Older adults produced a higher proportion of off-task utterances (i.e., details not probed by instructions) across all sections of the interview. Specifically, older adults produced more autobiographical facts in the AI, more episodic and general semantic details in the P-SAI, and more self-knowledge in the G-SAI than did young adults. However, older adults also consistently produced more probed autobiographical facts than did young adults on the P-SAI. These findings suggest that the increased production of semantic details in ageing reflects a bias towards age differences in autobiographical recall that goes beyond episodic remembering, as reflected by an age-associated abundance of semantic details across sections of the interview, findings that are not accommodated by accounts of aging and memory emphasizing reduced cognitive control or compensation for episodic memory impairment
Brexit’s Last Hurrah: Wine By The Pint
No full textMore than 7 1/2 years after the June 2016 referendum, almost four years after the UK formally withdrew from the European Union, we finally have the confirmation of Brexit’s endpoint.
Wine by the pint
Raising the voices of AuDHD women and girls: exploring the co-occurring conditions of autism and ADHD
No full textThis is a Current Issue because neurodivergent women and girls have been left behind, missed by the medical profession for decades. ADHD and autism have historically been considered to be male conditions, with diagnosis being 4 times more likely for males than females. The tides are changing with more women becoming aware of, diagnosed with, and seeking a diagnosis of conditions such as ADHD and autism. The media response to this surge in diagnoses has often been disparaging, although attempts to widen awareness exist (including recent BBC documentary Unmasking My Autism). However, women face long waiting lists, a lack of pre- and post-diagnosis support, sex biases in diagnostic criteria, and are left to deal with the trauma of a lifetime of misdiagnoses, poor mental health, and internalisation of negative messages about their character, alone. This piece explores two issues; the gendered omission of women and girls from autism and ADHD diagnoses and the theoretical and practical implications of the co-occurring conditions. When I refer to gender, I am referring to the social structure of gender which produces shared meanings about gendered behaviours, norms, roles, relations and institutions
Liz Truss was in Maryland with an explanation for her historically-short, 50-day tenure as UK Prime Minister
No full textTruss — whose stay in 10 Downing Street was infamously shorter than the shelf life of a lettuce — did not mention the impending economic catastrophe which she wrought in only a few weeks. She did not reflect on how she presented Britons with soaring mortgages and energy costs, or how she lost the confidence of businesses and financial markets as well as that of many Conservative Party colleagues
Recycling carbon taxes for reindustrialisation: addressing structural rigidity and financialisation in natural resource exporting countries
No full textInclusion of developing and emerging countries in the low carbon transition agenda is imperative to meet climate goals, and policies should be tailored to their unique characteristics. Despite their significance, the structural specifics of these countries are frequently overlooked in lowcarbon transition models. In an effort to establish an appropriate framework for such analyses, this article formulates a Structural StockFlow Consistent (Structural SFC)model designed for open developing economies. This model categorizes production into three sectors: resource based exports, non-tradable goods and services, and other tradable sectors. While SFC models play a crucial role in emphasizing financial constraints, they frequently lack a multi-sectoral viewpoint and disregard structural specificities. Our model makes a dual contribution: (1) it offers a flexible framework capable of accommodating diverse country characteristics while balancing short-term demand with long term structural strategies, and (2) it underscores the inadequacy of relying solely on carbon pricing for economies deeply rooted in carbon-intensive sectors. By incorporating structurally distinct sectors within a genuinely monetary framework, the model enables us to comprehend the decisive role played by financial constraints arising from structural rigidities in shaping the dynamics of the low-carbon transition. Our findings show that the efficacy of carbon pricing is contingent on a country’s commercial, financial, and productionstructure. Inclusion of developing and emerging countries in the low carbon transition agenda is imperative to meet climate goals, and policies should be tailored to their unique characteristics. Despite their significance, the structural specifics of these countries are frequently overlooked in lowcarbon transition models. In an effort to establish an appropriate framework for such analyses, this article formulates a Structural StockFlow Consistent (Structural SFC)model designed for open developing economies. This model categorizes production into three sectors: resource based exports, non-tradable goods and services, and other tradable sectors. While SFC models play a crucial role in emphasizing financial constraints, they frequently lack a multi-sectoral viewpoint and disregard structural specificities. Our model makes a dual contribution: (1) it offers a flexible framework capable of accommodating diverse country characteristics while balancing short-term demand with long term structural strategies, and (2) it underscores the inadequacy of relying solely on carbon pricing for economies deeply rooted in carbon-intensive sectors. By incorporating structurally distinct sectors within a genuinely monetary framework, the model enables us to comprehend the decisive role played by financial constraints arising from structural rigidities in shaping the dynamics of the low-carbon transition. Our findings show that the efficacy of carbon pricing is contingent on a country’s commercial, financial, and productionstructure
Secondary traumatic stress and trauma informed practice in Higher Education students: an exploratory study
No full textThe helping professions have long understood that secondary traumatic stress and its counterpart’s burnout and compassion fatigue are a problem for workers in the field. However, less is known about the impact of the issue on students who have placements. This quantitative research study sought to explore if a convenience sample of 45 students on two programmes in the field were affected, using the Secondary Traumatic Stress Scale. The results have shown several non- significant results, suggesting that the number of weekly caring responsibility hours did not predict perceived STSS scores after placement, and that high scoring students have shown no significant difference in STSS scores before and after placement. Overall, we also found that the sub-sample of 10 students with caring responsibilities had higher STSS scores. The article discusses wellbeing in students generally, incorporating trauma informed perspectives. While no students in this study were affected, the discussion what can be done to better support students from an ecological perspective to protect and prepare them for their placements and future careers. Finally, this article calls for policy and practice in education and the curriculum of the professions to routinely incorporate awareness of the issues in training and supervision
Talking a Good Game: Identifying the Discrepancies of Football Coaches’ Beliefs and Actions in Player Selection
No full textCoach intuition plays a critical role in the selection of academy players. A coach’s beliefs about a player's current abilities and perceived potential are critical in deciding a player’s future. Therefore, this study attempted to gain insight towards each coach’s experience and beliefs in selecting players, before undertaking a hypothetical selection activity to understand whether coaches act on such knowledge. Twenty-four coaches recruited from 21 unique professional football (soccer) academies (nine Category 1, eight Category 2, and seven Category 3) took part in semi-structured interviews. The findings established that coach beliefs and actions differed, whereby coaches stated a wide range of holistic beliefs towards selection, yet the hypothetical scenario outlined a far narrower selection criteria applied in action. While several beliefs were reinforced, it was also clear that biases were also presented. Maturation-related bias, favoring the more mature players, explained a potential focus on specific physical qualities (speed) and the perceived potential of players. Additionally, a focus on current performance, over wider elements related to perceived future potential, was evident during the selection scenario. Moreover, while subjective input will remain a key contributor to the player selection process, objective assessments and the input of wider multidisciplinary staff should be utilized to help mitigate the above-mentioned issues
Shuttle Time for Seniors: The Impact of 8-Week Structured Badminton Training on Markers of Healthy Aging and Evaluation of Lived Experiences - A Quasi-Experimental Study
No full textBackground/Objectives: Engagement in sport offers the potential for improved physical and psychological well-being and has been shown to be beneficial for promoting healthy aging. Opportunities for older adults to (re)engage with sport are limited by a paucity of age-appropriate introductory sports intervention programs. As such, the study evaluated the efficacy of a newly designed 8-week badminton training program (Shuttle Time for Seniors) on markers of healthy aging and the lived experiences of participation. Methods: Forty-three older adults assigned to a control (N = 20) or intervention group (N = 23) completed pre–post assessment of physical and cognitive function, self-efficacy for exercise, and well-being. Focus groups were conducted for program evaluation and to understand barriers and enablers to sustained participation. Results: Those in the intervention group increased upper body strength, aerobic fitness, coincidence anticipation time, and self-efficacy for exercise. Objectively improved physical and cognitive functions were corroborated by perceived benefits indicated in thematic analysis. Shuttle Time for Seniors was perceived as appropriate for the population, where the age-appropriate opportunity to participate with likeminded people of similar ability was a primary motivator to engagement. Despite willingness to continue playing, lack of badminton infrastructure was a primary barrier to continued engagement. Conclusion: Shuttle Time for Seniors offered an important opportunity for older adults to (re)engage with badminton, where the physical and psychosocial benefits of group-based badminton improved facets important to healthy aging. Significance/Implications: Age-appropriate introductory intervention programs provide opportunity for older adults to (re)engage with sport. However, important barriers to long-term engagement need to be addressed from a whole systems perspective
Compromising allocation for optimising agri-food supply chain distribution network: a fuzzy stochastic programming approach
No full textThe management of Agri-food supply chains is a complex task, given the unique product characteristics, perishability, uncertain demand, and specific storage requirements. This research introduces an innovative approach to optimizing product allocation among producers, brokers, wholesalers, and retailers, focusing on minimizing transportation costs and network delivery time through multi-objective programming. To address uncertainties, supply and demand constraints are modelled using a gamma distribution, and the maximum likelihood estimation method determines their parameters with specified probabilities. The study conducts a case analysis to showcase the model’s practical effectiveness, and a numerical comparison with alternative approaches is included. The primary goal of this study is to enhance the efficiency of agri-food supply chain management practices, providing valuable insights for practitioners in the field, with a focus on cost reduction and improved delivery time