Turkish Journal of Computer and Mathematics Education (TURCOMAT)

A COMPREHENSIVE SURVEY OF MEMORY UPDATE MECHANISMS FOR CONTINUAL LEARNING ON TEXT DATASETS

J. Ranjith — 2025-02-07

Over the last several years, there has been a growing focus on the CL field in the context of machine learning and its goal to create models capable of learning new tasks step by step without loss of prior knowledge. Among these, catastrophic forgetting is especially challenging in real-world settings where the data experience changes over time. To this effect, what has become pivotal for models is mechanisms for memory update to enable the models to learn information as well as update what has been previously learned easily. This survey specifically investigates the memory update strategy in the continual learning setup wherein new categories and domains are continuously added in the text datasets including sentiment analysis, named entity recognition, text classification tasks etc. Moving on, three primary memory update strategies of memory replay, memory consolidation, and parameter isolation are discussed; this paper further addresses certain adaptations of the proposed methods for text-based applications. Memory replay means that part of previous data is stored to be replayed when new tasks are learned while memory consolidation strengthens only significant memories. Parameter isolation helps avoid masking previous tasks or overwriting the parameters when the machine learning algorithm is trained to accomplish new tasks. In this paper, we discuss the latest in these techniques and offer a thorough insight into their use in text datasets such as Amazon Reviews and Yelp Reviews. Further, we outline the primary drawbacks of existing solutions for memory updates such as capacity limitations, domain variation, and continually learning without having access to new task information. In addition, a summary table of literature review identifying the most relevant works within the field is offered. Lastly, we discuss the remaining issues and potential research directions where more focus and development should be given in CL for text data by noting the importance of efficient and adaptive update policies towards the memory.

BUILDING A DYNAMIC CROPPING TREND VISUALIZATION SOFTWARE TO ASSESS FARMING PROGRESSIVENESS

MUHAMMAD ABDUL MANNAN — 2025-01-06

This thesis presents a comprehensive study that unfolds in two key dimensions. Firstly, a novel cropping trend visualization software was developed, aiming to provide a dynamic and insightful tool for assessing farming progressiveness. This software, a product of meticulous design and implementation, serves as an innovative instrument for visualizing and analyzing cropping trends in agriculture. In the second part of the study, the software was utilized to gather descriptive and inferential statistical data from a sample of 111 farmers. The data analysis reveals noteworthy correlations between specific agricultural practices and farming progressiveness. Key findings indicate significant positive associations between technology adoption, climate adaptation, crop variety diversity, and environmentally sustainable practices with higher levels of farming progressiveness. While certain trends exhibit statistical significance, caution is advised in drawing robust conclusions for factors with limited statistical support. These empirical insights provide a valuable foundation for policymakers, agricultural extension services, and stakeholders to strategically promote practices that enhance the overall progressiveness of farmers. The dual contribution of the developed software and the derived insights from its application underscores the significance of integrating technology in advancing sustainable and progressive agricultural practices.

Numerical method for simulating the model of using insecticide for the optimal control of mosquitoes for the eradication of malaria

Samuel Adamu — 2025-01-10

In order to simulate the mathematical model of eliminating malaria by controlling the population of
mosquitoes with insecticide and the insecticide's residual effects, the study developed a four points hybrid block
algorithm. The convergence and stability qualities of the block method are established. The block approach is
applied after the variable control problems are generated using Pontryagin's principle. The forward-backward sweep
methods of the block method are applied. The method is then implemented using a computer code using MATLAB
R2018a mathematical software. According to the findings of this study, the simulated result from this approach
displayed a significantly lower number of mosquitoes while lessening the negative effects of the insecticide, which
in turn will reduce the high rate of malaria spreading.

Mathematical Modelling and Argumentation: Designing a Task to Strengthen Variational Thinking by Integrating Data Science

Jonathan Alberto Cervantes-Barraza — 2025-01-21

This study explores the significance of task design in fostering variational thinking through the principles of PyLVar and task design in mathematics education, particularly in the incorporation of technological and scientific tools to enhance modelling and argumentation skills among secondary school students. To develop and implement interactive mathematical tasks based on the PyLVar principle and task design, incorporating data science tools and STEAM. A qualitative-descriptive approach was employed to design, implement, and analyse interactive tasks centred on the principles of variational thinking, argumentation, and modelling. The sample consisted of nine participants selected from a group of 11th-grade students at a state school, chosen for convenience. Data were gathered through experimental activities, processed using technological tools, and analysed with variational strategies. Students demonstrated a marked improvement in their abilities in mathematical modelling and argumentation. The tasks designed, structured around the PyLVar principles, enabled the identification of variation patterns and the development of critical and analytical skills. Contextualised interactive mathematical tasks not only bolster variational thinking but also equip students to address real-world problems by connecting abstract concepts to practical applications and encouraging the integration of technological tools and interdisciplinary approaches.

AI-Powered Encryption Revolutionizing Cybersecurity with Adaptive Cryptographic Algorithms

Zainab Rustum Mohsin — 2025-01-29

This paper proposes a state-of-the-art encryption technique that integrates artificial intelligence into a dynamic, content-aware security system. We introduce an AI-powered encryption framework that automatically adjusts cryptographic parameters based on message sensitivity, effectively balancing security requirements and computational efficiency. The system combines a DistilBERT-based neural network for real-time content sensitivity analysis with a flexible encryption mechanism that adapts key lengths, iteration counts, and entropy levels on the fly.

Our implementation demonstrates significant adaptability, with a correlation of 0.974 between content sensitivity and security parameters. The system distinguishes between different security requirements, using 32-byte keys with millions of iteration rounds for high-sensitivity content (sensitivity score of 8.64) and 16-byte keys with reduced iterations for low-sensitivity messages (sensitivity score of 3.10). The processing time scales linearly with security requirements, ranging from 300 ms for low-sensitivity content to 732 ms for high-security encryption.

Performance evaluation was highly effective, with the system achieving an overall score of 8.64/10, including 9.87/10 for adaptability and 9.12/10 for performance efficiency. The security level rated high at 7.37/10 while maintaining manageable computational overhead. The framework effectively handled different types of content without sacrificing encryption-decryption accuracy across all levels of sensitivity.

This work signifies a significant leap forward in the field of adaptive cryptography, demonstrating the capabilities of AI-driven security systems that can automate and optimize encryption parameters without compromising security standards. These results suggest that this approach may well represent the future of encrypted communications, providing scaled security appropriately without human intervention.