The automotive industry is inevitably experiencing a paradigm shift from fossil fuels to electric powertrain with significant technological breakthroughs in vehicle electrification. Emerging hybrid electric vehicles were one of the first steps towards cleaner and greener vehicles with a higher fuel economy and lower emission levels. The energy management strategy in hybrid electric vehicles determines the power flow pattern and significantly affects vehicle performance. Therefore, in this thesis, a learning-based strategy is proposed to address the energy management problem of a hybrid electric vehicle in various driving conditions. The idea of a deep recurrent neural network-based energy management strategy is proposed, developed, and evaluated. Initially, a hybrid electric vehicle model with a rule-based supervisory controller is constructed for this case study to obtain training data for the deep recurrent neural network and to evaluate the performance of the proposed energy management strategy. Secondly, due to its capabilities to remember historical data, a long short-term memory recurrent neural network is designed and trained to estimate the powertrain control variables from vehicle parameters. Extensive simulations are conducted to improve the model accuracy and ensure its generalization capability. Also, several hyper-parameters and structures are specifically tuned and debugged for this purpose. The novel proposed energy management strategy takes sequential data as input to capture the characteristics of both driver and controller behaviors and improve the estimation/prediction accuracy. The energy management controller is defined as a time-series problem, and a network predictor module is implemented in the system-level controller of the hybrid electric vehicle model. According to the simulation results, the proposed strategy and prediction model demonstrated lower fuel consumption and higher accuracy compared to other learning-based energy management strategies. / Thesis / Master of Applied Science (MASc)

No description available.

No description available.

No description available.

The ability to hold visual information in mind over a brief delay is critical for acquiring information and navigating a complex visual world. Despite the ubiquitous nature of visual short-term memory (VSTM) in our everyday lives, this system is fundamentally limited in capacity. Therefore, the potential to improve VSTM through training is a growing area of research. An emerging body of literature suggests that extensive experience playing action video games yields a myriad of perceptual and attentional benefits. Several lines of converging work provide evidence that action video game play influences VSTM as well. The current study utilized a training paradigm to examine whether action video games cause improvements to the quantity and/or the quality of information stored in VSTM and whether these VSTM advantages extend visual working memory (VWM). The results suggest that VSTM capacity is increased after action video game training, as compared to training on a control game, and that some limited improvement to VSTM precision occurs with action game training as well. The VSTM improvements seen in individuals trained on an action video game are not better accounted for by differences in motivation or engagement, differential expectations, or baseline differences in demographics as compared to the control group used. However, these findings do not appear to extend to measures of VWM, nor to verbal working memory. In sum, action video game training represents a potentially unique and engaging platform by which this severely capacity-limited VSTM system might be enhanced. / Psychology

Research suggests that novel word learning tasks engage both verbal short-term memory (STM) and lexical processing, and may serve as a potential treatment for word processing and functional language in aphasia (e.g., Gupta, Martin, Abbs, Schwartz, Lipinski, 2006; Tuomiranta, Grönroos, Martin, & Laine, 2014). The purpose of this study was to gain support for the hypotheses that novel word learning engages verbal STM and lexical access processes and can be used to promote improvements in these abilities in treatment of aphasia. We used a novel word learning task as a treatment with three participants: KT, UP, and CN, presenting with different types and severities of aphasia and predicted that treatment would result in (1) acquisition of trained novel words (2) improved verbal STM capacity and (3) improved access to and retrieval of real words. Twenty novel words were trained for 1 hour x 2 days/week x 4 weeks. Language and learning measures were administered pre- and post-treatment. All three participants showed receptive learning and some improvement on span tasks, while UP and CN demonstrated some expressive learning. KT also improved in performance on the Peabody Picture Vocabulary Test and the Philadelphia Naming Test. UP showed significant improvement on proportion Correct Information Units (CIUs) in discourse. CN showed some minimal improvement in narrative production for proportion CIUs and proportion of closed class words. These findings support that novel word learning treatment, which engages verbal STM processes and lexical retrieval pathways, can improve input lexical processing. Theoretically, this study provides further evidence for models that propose common mechanisms supporting novel word learning, short-term memory, and lexical processing. / Communication Sciences

The conceptual relation theory postulates that English noun-noun compound words (e.g., snowman) have an underlying predicate structure that is not present in the surface form, but is recovered during compound processing (e.g., man made of snow). The relational nature of constituent binding in compound words marks them as a linguistic construction that is distinct from both the simplex words (monomorphemic) and other complex words (derived and inflected words) previously examined in the context of verbal working memory. In short-term memory research, a growing body of evidence suggests that semantic properties of words influence verbal recall; however, such effects have not been examined in the context of compound conceptual relations. The present study investigated the possible effects of compound conceptual relations in verbal working memory via an immediate serial recall task. The task was designed to examine whether sharing of an individual relation leads to facilitative or inhibitory effects for compounds associated with that relation and, more generally, whether this semantic property of compound words contributes to their recollection from short-term memory. Evidence from the serial recall experiment suggested an effect of compound relation priming in working memory. Relational similarity between recall list items appeared to inhibit recall performance. The thesis discusses how this may be the result of increased competition between compound constituents as a result of heightened constituent-level activation during word recall. This effect was not observed in relations that appeared to be overly general, suggesting that the effect is only present when compound words are matched according to salient, sufficiently specified relations. / Thesis / Master of Science (MSc)

Short-term memory (STM) has demonstrated to be affected by serial order, involving the use of rhythm and entrainment to stimuli. However, less is known of the extent of this relationship and language learning, and the literature focuses on words rather than sentences. Moreover, the literature lacks an exploration of whether this relationship has a correlation with MMN responses. We had 30 participants (21 female) complete two sentence repetition tasks, a temporal rhythm accuracy task, and two temporal order judgment tasks. We also recorded the electroencephalograms (EEG) from 24 of the participants (17 female) while they listened to syllables differing by time of presentation and differing by consonant and vowel. We then correlated performance on these tasks to performance on a foreign-word learning (FWL) task. We hypothesized that the STM tasks would predict performance in the FWL task, and we explored whether temporal accuracy and word learning correlated with MMN responses to early stimuli. We found that only the foreign sentence repetition task significantly predicted performance in the FWL task. We also did not find any significant correlations with MMN responses and temporal accuracy and word learning abilities. Findings show that with previous exposure to a novel language, the prosodic pattern of the foreign language is stored temporarily in STM, which enhances learning of the foreign words. Further exploration is needed to understand the relationship of temporal order and language learning with cortical responses. / Thesis / Master of Science (MSc) / Spoken language is driven by rhythm and keeping track of this rhythm allows us to keep track of the order in which sounds in language are presented. Remembering the order of items requires the use of short-term memory. The better one is at repeating back the order of items, the better they are at learning new words. This thesis investigates the relationship between various short-term memory tasks (English nonword sentence repetition task, foreign sentence repetition task, temporal rhythm accuracy task, auditory judgment task, visual judgment task) and foreign-word learning. This thesis also explores whether there is a correlation between one’s brain responses to differing stimuli and a person’s ability to track the timing and order of items, as well as a person’s ability to learn new words. The results reveal that only the foreign sentence repetition task, using the same foreign language as the word learning task, significantly predicts one’s ability to accurately learn foreign words. The results did not show any significant interaction between one’s neural responses and rhythm or word learning. These results suggest that the ability to maintain the order of items in memory aids word learning, but further exploration is required with regards to non-verbal stimuli and neural responses. It is important to investigate individual differences in repetition tasks that require short-term memory, as this will aid in understanding normal language development and language acquisition.

Bitcoin has since its inception in 2009 seen its market capitalisation rise to a staggering 846 billion US Dollars making it the world’s leading cryptocurrency. This has attracted financial analysts as well as researchers to experiment with different models with the aim of developing one capable of predicting Bitcoin closing prices. The aim of this thesis was to examine how well the LSTM and the SVR models performed in predicting Bitcoin closing prices. As a measure of performance, the RMSE, NRMSE and MAPE were used as well as the Random walk without drift as a benchmark to further contextualise the performance of both models. The empirical results show that the Random walk without drift yielded the best results for both the RMSE and NRMSE scoring 1624.638 and 0.02525, respectively while the LSTM outperformed both the Random Walk without drift and the SVR model in terms of the MAPE scoring 0.0272 against 0.0274 for both the Random walk without drift and SVR, respectively. Given the performance of the Random Walk against both models, it cannot be inferred that the LSTM and SVR models yielded statistically significant predictions. / <p>Aaron Green</p>

Yes / Time-series anomaly detection receives increasing research interest given the growing number of data-rich application domains. Recent additions to anomaly detection methods in research literature include deep neural networks (DNNs: e.g., RNN, CNN, and Autoencoder). The nature and performance of these algorithms in sequence analysis enable them to learn hierarchical discriminative features and time-series temporal nature. However, their performance is affected by usually assuming a Gaussian distribution on the prediction error, which is either ranked, or threshold to label data instances as anomalous or not. An exact parametric distribution is often not directly relevant in many applications though. This will potentially produce faulty decisions from false anomaly predictions due to high variations in data interpretation. The expectations are to produce outputs characterized by a level of confidence. Thus, implementations need the Prediction Interval (PI) that quantify the level of uncertainty associated with the DNN point forecasts, which helps in making better-informed decision and mitigates against false anomaly alerts. An effort has been made in reducing false anomaly alerts through the use of quantile regression for identification of anomalies, but it is limited to the use of quantile interval to identify uncertainties in the data. In this paper, an improve time-series anomaly detection method called deep quantile regression anomaly detection (DQR-AD) is proposed. The proposed method go further to used quantile interval (QI) as anomaly score and compare it with threshold to identify anomalous points in time-series data. The tests run of the proposed method on publicly available anomaly benchmark datasets demonstrate its effective performance over other methods that assumed Gaussian distribution on the prediction or reconstruction cost for detection of anomalies. This shows that our method is potentially less sensitive to data distribution than existing approaches. / Petroleum Technology Development Fund (PTDF) PhD Scholarship, Nigeria (Award Number: PTDF/ ED/PHD/IAT/884/16)