19614 research outputs found
Sort by
2024 On the Verge Writing Contest Non-Fiction Honorary Mention
No full textAn award-winning work of non-fiction with the theme of equity, diversity, and human rights created by undergraduate student Alex Da Matta, selected by celebrity judge Thembelihle (Thembie) Moyo.UndergraduateReviewe
Integration of model predictive control and reinforcement learning for dynamic systems with application to robot manipulators
Get PDFThe last decade has witnessed great progress in the development of reinforcement learning (RL) across many applications, such as games and autonomous driving. RL is effective in solving control problems for complex systems whose dynamics are intractable to be accurately modeled. In an RL algorithm, the agent learns the optimal policy in terms of the maximum reward based on measurement samples from the interactions with the environment. To obtain the optimal policy, RL requires collecting sufficiently large number of samples, which is challenging in real-world applications, e.g., robotics, manufacturing, and so on. To tackle this problem, model predictive control-based RL (MPC-based RL) is proposed to improve the sample efficiency. In the MPC-based RL algorithm, a model is learned from collected samples, the learned model and MPC are utilized to predict trajectories over a specified prediction horizon, and an action is obtained through the RL algorithm by maximizing the cumulative reward. This thesis is devoted to the investigation of the MPC-based RL design and its application to robot manipulators.
In Chapter 2, an MPC-based deep RL framework for constrained linear systems with bounded disturbances is proposed. In the proposed framework, a rigid tube-based MPC (RTMPC) method is employed to predict a trajectory by solving the corresponding optimization problem. Then, the predicted trajectory is stored in a replay buffer as the form of data pairs. Further, the soft actor-critic (SAC) algorithm is applied to modify the loss function and update the policy online, based on the predicted data pairs. Numerical simulations validate the effectiveness of the proposed method. In addition, comparison results demonstrate the advantages of the proposed method including requirement of fewer real samples and providing better control performance with comparable computational complexity to RTMPC.
In Chapter 3, we investigate the application of three methods for manipulators. Firstly, we apply an MPC-based RL algorithm, a nonlinear MPC (NMPC) method, and two model-free RL algorithms to tackle the regulation problem for a 2-degree-of-freedom manipulator system, and compare their training control performance. Secondly, the training and control performance evaluation for the model-free RL algorithm and the MPC-based RL algorithm are provided. The MPC-based RL algorithm shows better training performance in terms of sample efficiency and total return but poorer control performance. Thirdly, simulation studies are provided to compare the training performance of the MPC-based RL algorithm and two model-free RL algorithms. From the simulation results, the MPC-based RL algorithm presents poorer training performance compared with model-free RL algorithms for the twelve-dimensional system.
In Chapter 4, conclusions and future work are summarized.Graduat
A comparative analysis of mould growth on exterior sheathing of a brick masonry wall in different Canadian climate zones
Get PDFThis study investigates the risk of mould growth on sheathing boards in brick masonry walls in four Canadian cities: Vancouver, Ottawa, Calgary, and Saskatoon. The hygrothermal simulation tool WUFI® Pro 6.8 (1D) and the VTT Mold Index were used to conduct this investigation. The impact of moisture penetration through brick veneer cladding on the potential for mould growth in Oriented Strand Board (OSB), Fiberboard (FB), and Plywood (Ply) sheathing was assessed.
For hygrothermal simulations, a severe weather year was selected based on a 31-year historical weather dataset (1986-2016) using the severity index (Isev) method prescribed in the ASHRAE Standard 160-2021. The calculation period was set for seven years, and two wall orientations were considered: (i) direction with the least solar radiation and (ii) maximum wind-driven rain direction. For each orientation, three rain penetration cases (1%, 2% and 3% of wind-driven rain) were considered, and two Air Change Rates (ACH 0 and ACH 15) were considered in the drainage cavity for each of the three rain penetration cases. As per ASHRAE 160-2021, the rain penetration was deposited on the outer layer of the water-resistive barrier (WRB).
The results showed that for the 1% rain penetration and no ventilation, the mould growth index (MGI) for all three sheathing boards remained at zero (i.e., No mould growth) for Vancouver’s north-oriented wall (least solar radiation); however, the southeast-facing wall (maximum wind-driven rain) experienced a higher MGI (up to 5.3). For the same case (i.e. 1% rain penetration), the remaining simulated cities experienced MGI>5 (i.e., 50% visually covered surface). In the case of increased rain penetration and no ventilation, each sheathing board’s mould growth performance significantly decreased (MGI>5) for all four cities in both orientations; however, an air change rate of 15/hour (ACH 15) in the drainage cavity reduced the mould growth (MGI<1, local growth microscopic level) in Calgary, Ottawa and Saskatoon. In contrast, ACH 15 was insufficient to reduce the MGI < 3 (i.e., visuals of mould <10% surface coverage) for the Vancouver location, except for the 1% rain penetration case.Graduat
Variability across subjects in free recall versus cued recall
Get PDFWe would like to thank Henry L. Roediger, Colleen M. Kelley, John Dunlosky, Larry Jacoby, Reed Hunt, and Roger Ratcliff for their helpful insights and suggestions.Memory scientists usually compare mean performance on some measure(s) (accuracy, confidence, latency) as a function of experimental condition. Some researchers have made within-subject variability in task performance a focal outcome measure (e.g., Yao et al., 2016). Here we explored between-subject variability in accuracy as a function of experimental conditions. This work was inspired by an incidental finding in a previous study in which we observed greater variability in accuracy of memory performance on cued recall (CR) versus free recall (FR) of English animal/object nouns (Mah et al., 2023). Here we report experiments designed to assess the reliability of that pattern and to explore its causes (e.g., differential interpretation of instructions, (un)relatedness of CR word pairs, encoding time). In Experiment 1 (N = 120 undergraduates), we replicated the CR:FR variability difference with a more representative set of English nouns. In Experiments 2A (N = 117 Prolific participants) and 2B (N = 127 undergraduates), we found that the CR:FR variability difference persisted in a forced-recall procedure. In Experiment 3 (N = 260 Prolific participants), we used meaningfully related word pairs and still found greater variability in CR than FR performance. In Experiment 4 (N = 360 Prolific participants), we equated CR and FR study phases by having all participants study pairs and again observed greater variability in CR than FR. The same was true in Experiment 5 (N = 120 undergraduates), in which study time was self-paced. Comparisons of variability across subjects can yield insights into the mechanisms underlying task performance.This work was supported by an NSERC Discovery grant (#RGPIN-2016-03944) awarded to DSL.FacultyUnreviewe
Test with name authorities and embargo
No full text1. ORCID in Solr field
2. Author ID in Solr field
3. ORCID iD in ORCID fiel
Quantitative Models for Accurate Reactivity Predictions and Mechanistic Elucidation
No full textAccurate prediction of reaction outcomes is among the most important goals in chemical and pharmaceutical synthesis. In recent years, the ultrafast growth in computing power and the advancement of high-throughput experimental (HTE) technology have paved new ways to apply data-rich approaches in chemistry research. In organic synthesis, data-driven methods have found many successful applications in accelerating reaction condition optimization and developing machine learning models for reaction prediction. Despite all the impressive progress made in this area, accurate prediction of chemical reactivity remains challenging.
This thesis describes development of quantitative reactivity models for accurate reaction prediction and mechanistic elucidation in organic synthesis. Starting with a minireview/perspective in Chapter 1, recent progress is discussed in data-rich approaches to reaction development and quantitative predictions for palladium-catalyzed reaction systems. In Chapter 2 and Chapter 3, quantitative predictive models are developed for two pharmaceutically important reaction systems: nucleophilic aromatic substitution (SNAr) and oxidative addition to palladium(0), a fundamental and usually the rate/selectivity determining step in palladium-catalyzed cross-coupling reactions. Both models focus on structure-reactivity relationships of the electrophiles. Diverse and reliable reaction rate data for training set was collected using high-throughput competition experimentation. These were used to construct multivariate linear regression models by quantitatively mapping a group of ground state molecular descriptors to the experimental reaction rates. Predictive accuracy is validated via a series of random train-test splits, as well as predicting outcomes for a wide variety of external reaction data. Following the procedures described above, generally applicable models for quantitative predictions on both the reaction rates and site-selectivity for both reaction systems have been realized.
In addition to making quantitative reaction predictions, a structure-reactivity model constructed using high-quality data and mechanistically meaningful descriptors is also very useful in gaining mechanistic insights. This is demonstrated by the solvent effect study in Chapter 4 and the reaction mechanistic study in Chapter 5. From the quantitative reactivity scales constructed for oxidative addition to palladium(0) in different solvents, specific electrophiles were identified that exhibit significant solvent effects; the role of solvent was investigated case by case. These include the importance of solvent hydrogen-bond basicity as well as solvent polarity. Finally, the underlying mechanistic causes behind a series of systematic prediction outliers from our oxidative addition model were investigated. These reveal that the frontier orbital symmetry also plays an important role in determining reaction outcomes. Insights into these mechanistic aspects, which have a significant impact on both the reaction rate and site-selectivity in oxidative addition to palladium(0), enabled a refined quantitative model that incorporates frontier orbital descriptors.Graduate2024-11-1
The choice of prediction curve method and its effect on the estimated amount of DNA
No full textWater samples from the field data that contained environmental DNA (eDNA) were taken from
multiple rivers where oolichan fish (Thaleichthys pacificus) are known to spawn. The samples
were split into 8 technical replicates and analyzed using quantitative real-time polymerase chain
reaction (qPCR). A qPCR experiment is the real time quantification of DNA amount at the
end of a full cycle of heating and cooling. CT values were determined from a qPCR experiment
or a replicate was given N/A (not available) if no DNA was detected. Four data sets from
two different labs were used, Bureau Vertitas Lab (BVL) and University of Victoria (UVic).
Both labs have a gblock and field data set with chemical assays named eTHPA2 and eTHPA6.
Gblock data is comprised of gblock samples which were synthetically constructed genes of known
concentration (copy number) and measured using qPCR. Field data is comprised of samples
taken from river sites in British Columbia where eDNA naturally occurs, and the copy numbers
were unknown for the field samples. The field samples were analyzed using qPCR technique to
determine the CT value for each technical replicate. Each data set was split into two subsets
named full and partial detect, resulting in eight working data sets. The full data sets were
comprised of samples whose technical replicates had (8/8) detects. The partial data set was
comprised of samples whose technical replicates had less than (8/8) detects. For the partial
detect data, a Binomial model for the proportion of detects in a sample was defined, where a
replicate with a CT value was an “event” and N/A was not an “event”. Assuming the number
of molecules in a sample followed a Poisson distribution with mean λ, we estimated the λ as
λˆ = −ln(1−pˆ), where pˆ is the estimated sample proportion of detect from the Binomial model.
Standard/calibration and prediction curves were built from the gblock data. Standard curves
were built using gblock data with known copy number values, and relate CT and λˆ values to
copy number values. Standard curves were used to estimate copy numbers given CT or λˆ values
for samples with unknown copy numbers. Prediction curves were built by fitting least squares
and orthogonal regression using an unweighted and weighted method for each, to the gblock
data. Prediction curves were used to estimate eTHPA6 CT or λˆ values given eTHPA2 CT or
λˆ values. Plots and model summaries for the four prediction curves for each data set were analyzed. Based off the analysis and recommendation of the literature, weighted orthogonal
regression was chosen as the best prediction model for each gblock data set. The prediction
curves were applied to the corresponding field data to investigate how well the models predict
the values of eTHPA6 given eTHPA2. All of the data sets saw majority well predicted final
eTHPA6 copy number values, which indicated that the weighted Deming model was a good
prediction method.
The purpose of this study was to determine the best statistical methods for eDNA assay
prediction for biologists and other researchers to use. From the methods validated in this study
researchers can go on to connect the population estimates of the oolichan species made from
the older and newer assays, make conclusions on the health of the species population, produce
plan(s) to safeguard the population against over harvesting, and more conservation work.Graduat
Integrating educational technology in primary classrooms: Purpose and pedagogy
Get PDFNumerous studies have provided information on teachers’ beliefs, attitudes, and perceptions of technology integration, the ubiquitous nature of technology and variation in its use, and the barriers experienced by teachers (e.g. Ertmer et al., 2012; Lu et al., 2017; Miller, 2018; Rowsell & Harwood, 2015). However, few studies have offered thorough examinations of Canadian primary teachers’ perspectives of and rationales for including specific technologies and digital experiences in their pedagogical practices with young learners. This doctoral study provides in-depth information about primary teachers’ technology-supported practices, including pedagogical approaches that have proven successful prior to pandemic teaching and documenting unique approaches to technology integration that have arisen as a response to COVID-19.
Grounded in sociocultural theory and multiliteracies theory, this mixed-methods multiple-case study investigated four Kindergarten-Grade 2 teachers’ practices and purposes for integrating technology. Participants were from one rural and one urban school division in Manitoba, and data were collected during pandemic teaching conditions. Findings revealed that although there was variation in teachers’ pedagogical beliefs, practices, and device access, there was significant similarity in their purposes for using technology. Common purposes for integrating technology included building community, enhancing teaching and learning, overcoming barriers, and maintaining educational continuity during COVID-19. All participants leveraged a range of platforms and tools to impart digital competencies while endeavouring to balance screentime, creation, and consumption. This study offers suggestions for how primary teachers may integrate technology in effective and innovative ways while documenting the barriers experienced by teachers in their efforts to do so.Graduat
Global chemical transport on hot Jupiters: Insights from the 2D VULCAN photochemical model
Get PDFThe atmospheric dynamics of tidally locked hot Jupiters is characterized by strong equatorial winds. Understanding the interaction between global circulation and chemistry is crucial in atmospheric studies and interpreting observations. Two-dimensional (2D) photochemical transport models shed light on how the atmospheric composition depends on circulation. In this paper, we introduce the 2D photochemical (horizontal and vertical) transport model, VULCAN 2D, which improves on the pseudo-2D approaches by allowing for nonuniform zonal winds. We extensively validate our VULCAN 2D with analytical solutions and benchmark comparisons. Applications to HD 189733 b and HD 209458 b reveal a transition in mixing regimes: horizontal transport predominates below ∼0.1 mbar, while vertical mixing is more important at higher altitudes above 0.1 mbar. Motivated by the previously inferred carbon-rich atmosphere, we find that HD 209458 b with supersolar carbon-to-oxygen ratio (C/O) exhibits pronounced C2H4 absorption on the morning limb but not on the evening limb, due to horizontal transport from the nightside. We discuss when a pseudo-2D approach is a valid assumption and its inherent limitations. Finally, we demonstrate the effect of horizontal transport in transmission observations and its impact on the morning−evening limb asymmetry with synthetic spectra, highlighting the need to consider global transport when interpreting exoplanet atmospheres.Part of this work is supported by the European community through the ERC advanced grant EXOCONDENSE (No. 740963; PI: R. T. Pierrehumbert). S.-M.T. acknowledges support from NASA Exobiology grant No. 80NSSC20K1437 and the University of California, Riverside. X.Z. acknowledges support from the NASA Exoplanet Research grant 80NSSC22K0236 and the NASA Interdisciplinary Consortia for Astrobiology Research (ICAR) grant 80NSSC21K0597. Financial support to R.D. was provided by a Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant to C. Goldblatt.FacultyReviewe
Inferring network topology for distributed machine learning model training
Get PDFWith the application of distributed machine learning in various industries, there is an increasing demand for model training using cloud computing resources. However, many cloud computing service providers refuse to provide end-users with information about the underlying network topology for commercial and security reasons. Due to this opaqueness, it is challenging to arrange the computation modules in different Virtual Machines (VMs) to achieve the best resource utilization efficiency. To address this problem, we propose an algorithm called Flow Tracking (FT), which uses external measurements to infer the internal structure of a general graph. Compared to the state-of-the-art topology inference algorithms, FT achieves the most accurate topology measured in four different metrics. Notably, FT achieves 100% reconstruction of the underlying topology under the shortest-path routing strategy of the underlying network. Experimentally, resource allocation using the inferred topology improves the model training efficiency significantly compared to random allocation.Graduat