Bayesian Knowledge Tracing (BKT) is a popular interpretable computational model in the educational mining community that can infer a student’s knowledge state and predict future performance based on practice history, enabling tutoring systems to adaptively select exercises to match the student’s compe- tency level. Existing BKT implementations do not scale to large datasets and are difficult to extend and improve in terms of prediction accuracy. On the other hand, uninterpretable neural network (NN) student models, such as Deep Knowledge Tracing, enjoy the speed and modeling flexibility of popular computational frameworks (e.g., PyTorch, Tensorflow, etc.), making them easy to develop and extend. To bridge this gap, we develop a collection of BKT recurrent neural network (RNN) cells that are much faster than brute-force implementations and are within an order of magnitude of a fast, fine-tuned but inflexible C++ implementation. We leverage our implementation’s modeling flexibility to create two novel extensions of BKT that significantly boost its performance. The first merges item response theory (IRT) and BKT by modeling multidimensional problem difficulties and student abilities without fitting student-specific parameters, allowing the model to easily generalize to new students in a principled way. The second extension discovers the discrete assignment matrix of problems to knowledge components (KCs) via stochastic neural network techniques and supports further guidance via problem input features and an auxiliary loss objective. Both extensions are learned in an end-to-end fashion; that is, problem difficulties, student abilities, and assignments to knowledge components are jointly learned with BKT parameters. In synthetic experiments, the skill discovery model can partially recover the true generating problem-KC assignment matrix while achieving high accuracy, even in some cases where the true KCs are structured unfavorably (interleaving sequences). On a real dataset where problem content is available, the skill discovery model matches BKT with expert-provided skills, despite using fewer KCs. On seven out of eight real-world datasets, our novel extensions achieve prediction performance that is within 0.04 AUC-ROC points of state-of-the-art models. We conclude by showing visualizations of the parameters and inferences to demonstrate the interpretability of our BKT RNN models on a real-life dataset.

Sand dunes in Kuwait naturally occur due to the arid climatic conditions and the geographic location of Kuwait in the northeastern corner of the Arabian Peninsula. The total area of Kuwait is 17,800 km2 out of which the sand dunes (including nebkhas) and sand sheets cover about 5093 km2 (29 % of the country). Sand dunes are located along the coastal zones and inland formulating different shapes based on their geomorphic characteristics, e.g. crescentic, star and falling dunes. Significant changes in aeolian conditions have been observed in Kuwait. These changes were mainly related to dramatic changes in land-use and infrastructure expansion. To alleviate sand dune hazards, a sustainable management plan needs to be adopted in Kuwait. In this chapter we identify the socioeconomic impact of sand dunes and propose a sustainable management plan for sand dune fixation in Kuwait.

The PhET Interactive Simulations project has begun an initiative to increase the accessibility of its suite of science simulations. In this work, we focus on use of the PhET sim Capacitor Lab: Basics by two visually impaired learners. Comparing responses to pre and posttest questions about capacitance and circuits, our results indicate that after using the simulation with new accessibility features, both learners better understood the relationship between plate separation, area, and capacitance. Additionally, while only one learner connected the charged capacitor to the light bulb in the simulation during use, both answered that the light bulb will be illuminated in posttest questions. These findings indicate that visually impaired students can master the learning goals of a PhET sim with well-designed accessibility features. Findings from this research contribute to understanding how to develop physics education resources capable of supporting diverse students, including students with disabilities.

Protected agriculture is a field in which the use of automatic systems is a key factor. In fact, the automatic harvesting of delicate fruit has not yet been perfected. This issue has received a great deal of attention over the last forty years, although no commercial harvesting robots are available at present, mainly due to the complexity and variability of the working environments. In this work we developed a computer vision system (CVS) to automate the detection and localization of fruit in a tomato crop in a typical Mediterranean greenhouse. The tasks to be performed by the system are: (1) the detection of the ripe tomatoes, (2) the location of the ripe tomatoes in the XY coordinates of the image, and (3) the location of the ripe tomatoes’ peduncles in the XY coordinates of the image. Tasks 1 and 2 were performed using a large set of digital image processing tools (enhancement, edge detection, segmentation, and the feature’s description of the tomatoes). Task 3 was carried out using basic trigonometry and numerical and geometrical descriptors. The results are very promising for beef and cluster tomatoes, with the system being able to classify 80.8% and 87.5%, respectively, of fruit with visible peduncles as “collectible”. The average processing time per image for visible ripe and harvested tomatoes was less than 30 ms.

Disjoint paths are defined as paths between the source and destination nodes where the intermediate nodes in any two paths are disjoint. They are helpful in fault-tolerance routing and securing message distribution in the network. Several research papers were proposed to solve the problem of finding disjoint paths for a variety of interconnection networks such as Hypercube, Generalized Hypercube, Mesh, Torus, Gaussian, Eisenstein–Jacobi, and many other topologies. In this research, we have developed a general algorithm that constructs maximal node-to-node disjoint paths for symmetric networks where all paths are shortest. The algorithm presented in this paper outperforms other algorithms in finding not only the disjoint paths but shortest and maximal disjoint paths with a complexity of $O(n^2)$. In addition, we have simulated the proposed algorithm on different networks. The solution of unsolved problem in Cube-Connected-Cycles is given in the simulation results.

A science camp for blind or visually impaired (BVI) students, covering mainly hands-on chemistry experiments, is described. The scientific experiments are detailed, and the feedback of the students is recorded, highlighting the benefits of the camp and its impact. BVI students actively participated in conducting the scientific experiments using adaptive methods and assistive technology. The camp increased the students’ awareness of scientific fields, namely, chemistry, and the students look forward to additional adaptable experiments in future camps.

There continues to be a persistent, widespread gender gap in multiple STEM disciplines at all educational and professional levels: from the self-reported interest of preschool aged students in scientific exploration to the percentages of tenured faculty in these disciplines, more men than women express an interest in science, a confidence in their scientific abilities, and ultimately decide to pursue scientific careers. Reported herein is an intensive outreach effort focused on addressing this gender gap: a full-time, week-long chemistry camp that was designed and implemented for middle school girls in the state of Rhode Island. The camp schedule included multiple hands-on experiments, field trips, and significant interactions with female scientists, all of which were designed to increase the participants’ interest in and enthusiasm for science. The success of the program in changing the participants’ attitudes toward science was measured through administration of a precamp and postcamp survey, and the survey results demonstrated a strong success in changing the participants’ attitudes toward the widespread applicability of science, their perceived level of support for scientific study, and their interest in pursuing STEM-related careers.

I will be presenting a lecture at KFAS on generative AI on December 12. This will be a dry run to get feedback.

This paper presents a comprehensive comparison of various Machine Learning (ML) classifiers for urban land use identification using social media data. Two analysis cycles were conducted, with the second cycle introducing the "popularity index" parameter. The results demonstrate that incorporating the popularity index significantly improved the accuracy rates of all classifiers. The Convolutional Neural Network (CNN) consistently outperformed other classifiers with a 0.9 accuracy, but the key highlight was the popularity index parameter, which optimized urban land use identification. By providing crucial contextual information and normalizing raw tweet counts, the popularity index proved vital for leveraging social media data in urban land use classification. Our findings support the value of social media data for analyzing urban land use dynamics and highlight ML classifiers with the popularity index as a promising approach to monitor and understand ever-changing patterns. This research makes a significant contribution to data-driven urban planning by highlighting the capacity of social media data to provide real-time and detailed insights into urban activities and land use patterns. These insights, in turn, can inform and enhance strategies for sustainable urban development and resource allocation, making them more informed and effective.

Optimizing oil refinery processes for fuel efficiency and product quality is becoming more important under increasingly tighter environmental regulations. In this paper, we consider the problems of offline and active optimization of the C5/C6 isomerization process using only process inputs and output key performance indicators (KPIs). For offline optimization, we simulate thousands of process configurations and study the impact of optimizing for one KPI (e.g., yield) on other KPIs (e.g., octane number). Surprisingly, for our choices of optimization variables, minimizing energy consumption is the least detrimental on other KPIs. Moreover, an artificial neural network (ANN) model significantly outperforms baseline models in predicting simulated data. For active optimization, we show that our easy-to-use and extensible method can find optimal feasible parameter configuration in as few as 30 experiments, enabling operators to optimize their processes without the need for a model of the refinery process.

An overview of supervised machine learning with neural networks. We will start from the very basics with linear models, then we will move into non-linear models, followed by multilayer perceptrons and classifiers, and conclude with an example.

Hands-on chemistry experiments often stimulate students’ curiosity about this subject, but blind or visually impaired (BVI) students usually do not get to participate in such hands-on activities. Thanks to adaptive methods and assisting technology, BVI students can actively participate in chemistry experiments. In this study, chemical laboratory experiments were modified and tested on BVI students. These adaptations use Sci-Voice Talking LabQuest, associated sensors, and other tactile tools. This article explains the procedures to conduct the experiments and presents practical ways to introduce assistive technology to enhance learning. We encourage accommodating Arabic language into the Sci-Voice Talking LabQuest.

KISR conducts periodic marine surveys in which vessels visit predefined locations in Kuwait's marine environment and log environmental indicators, such as depth and salinity, along with fish catch indicators, such as diversity, weight, and size. Currently, these results are logged in an electronic logging system lacking basic analytical and visualization capabilities, poor database organization, and outdated web technologies. Alternative pre-built systems are unsuitable because they either do not log the required data, require licensing fees, or are tailored toward commercial fisheries and marine regulations. This GRA aims to develop a new marine e-logging system for KISR that supports basic analytics and visualizations, is easy to use, and has robust data validation constraints. In this presentation, the design, implementation, and outcomes of the developed applications are discussed, and future recommendations are made.