Self-Review Reinforcement Learning (SRRL) with Cross-Episode Memory and Policy Distillation

Authors: Muhammad Zain Amin, Kibele Sebnem Yildirim

Abstract: Reinforcement Learning is commonly used to train large language models using environmental feedback. In applied settings, the environment usually provides sparse or delayed feedback. This makes it difficult for the model to pinpoint which actions in its reasoning led to success or failure. So, learning effectively from these signals is hard because the model must determine how each failure should… ▽ More Reinforcement Learning is commonly used to train large language models using environmental feedback. In applied settings, the environment usually provides sparse or delayed feedback. This makes it difficult for the model to pinpoint which actions in its reasoning led to success or failure. So, learning effectively from these signals is hard because the model must determine how each failure should inform meaningful behavioral corrections in subsequent iterations. We introduce a training framework, Self-Review Reinforcement Learning, that embeds an explicit self-review step into each RL episode. When a first-pass response fails, the model generates a self-review to identify what went wrong, which conditions an improved second attempt. Unlike inference-time reflection approaches, such as Reflexion, the framework optimizes self-review with policy gradients and internalizes improvements into the base policy via selective distillation, ensuring they persist across future episodes. A cross-episode memory keeps successful self-reviews for reuse when encountering similar tasks in future episodes during training. We evaluate SRRL against a standard RLVR baseline using the GRPO optimizer across two language models, Qwen 3-4B and OLMo-3- 7B, on GSM8K benchmark. SRRL consistently outperforms the RLVR in final reward performance and achieves greater learning efficiency by successfully transforming feedback into behavioral improvement. △ Less

Submitted 6 July, 2026; originally announced July 2026.

Comparative Study of Probabilistic Atlas and Deep Learning Approaches for Automatic Brain Tissue Segmentation from MRI Using N4 Bias Field Correction and Anisotropic Diffusion Pre-processing Techniques

Authors: Mohammad Imran Hossain, Muhammad Zain Amin, Daniel Tweneboah Anyimadu, Taofik Ahmed Suleiman

Abstract: Automatic brain tissue segmentation from Magnetic Resonance Imaging (MRI) images is vital for accurate diagnosis and further analysis in medical imaging. Despite advancements in segmentation techniques, a comprehensive comparison between traditional statistical methods and modern deep learning approaches using pre-processing techniques like N4 Bias Field Correction and Anisotropic Diffusion remain… ▽ More Automatic brain tissue segmentation from Magnetic Resonance Imaging (MRI) images is vital for accurate diagnosis and further analysis in medical imaging. Despite advancements in segmentation techniques, a comprehensive comparison between traditional statistical methods and modern deep learning approaches using pre-processing techniques like N4 Bias Field Correction and Anisotropic Diffusion remains underexplored. This study provides a comparative analysis of various segmentation models, including Probabilistic ATLAS, U-Net, nnU-Net, and LinkNet, enhanced with these pre-processing techniques to segment brain tissues (white matter (WM), grey matter (GM) and cerebrospinal fluid (CSF)) on the Internet Brain Segmentation Repository (IBSR18) dataset. Our results demonstrate that the 3D nnU-Net model outperforms others, achieving the highest mean Dice Coefficient score (0.937 +- 0.012), while the 2D nnU-Net model recorded the lowest mean Hausdorff Distance (5.005 +- 0.343 mm) and the lowest mean Absolute Volumetric Difference (3.695 +- 2.931 mm) across five unseen test samples. The findings highlight the superiority of nnU-Net models in brain tissue segmentation, particularly when combined with N4 Bias Field Correction and Anisotropic Diffusion pre-processing techniques. Our implemented code can be accessed via GitHub. △ Less

Submitted 8 November, 2024; originally announced November 2024.

An Architectural Approach to Creating a Cloud Application for Developing Microservices

Authors: A. N. M. Sajedul Alam, Junaid Bin Kibria, Al Hasib Mahamud, Arnob Kumar Dey, Hasan Muhammed Zahidul Amin, Md Sabbir Hossain, Annajiat Alim Rasel

Abstract: The cloud is a new paradigm that is paving the way for new approaches and standards. The architectural styles are evolving in response to the cloud's requirements. In recent years, microservices have emerged as the preferred architectural style for scalable, rapidly evolving cloud applications. The adoption of microservices to the detriment of monolithic structures, which are increasingly being ph… ▽ More The cloud is a new paradigm that is paving the way for new approaches and standards. The architectural styles are evolving in response to the cloud's requirements. In recent years, microservices have emerged as the preferred architectural style for scalable, rapidly evolving cloud applications. The adoption of microservices to the detriment of monolithic structures, which are increasingly being phased out, is one of the most significant developments in business architecture. Cloud-native architectures make microservices system deployment more productive, adaptable, and cost-effective. Regardless, many firms have begun to transition from one type of architecture to another, though this is still in its early stages. The primary purpose of this article is to gain a better understanding of how to design microservices through developing cloud apps, as well as current microservices trends, the reason for microservices research, emerging standards, and prospective research gaps. Researchers and practitioners in software engineering can use the data to stay current on SOA and cloud computing developments. △ Less

Submitted 7 October, 2022; v1 submitted 5 October, 2022; originally announced October 2022.

A Survey: Implementations of Non-fungible Token System in Different Fields

Authors: A. N. M. Sajedul Alam, Junaid Bin Kibria, Al Hasib Mahamud, Arnob Kumar Dey, Hasan Muhammed Zahidul Amin, Md Sabbir Hossain, Annajiat Alim Rasel

Abstract: In the realm of digital art and collectibles, NFTs are sweeping the board. Because of the massive sales to a new crypto audience, the livelihoods of digital artists are being transformed. It is no surprise that celebs are jumping on the bandwagon. It is a fact that NFTs can be used in multiple ways, including digital artwork such as animation, character design, digital painting, collection of self… ▽ More In the realm of digital art and collectibles, NFTs are sweeping the board. Because of the massive sales to a new crypto audience, the livelihoods of digital artists are being transformed. It is no surprise that celebs are jumping on the bandwagon. It is a fact that NFTs can be used in multiple ways, including digital artwork such as animation, character design, digital painting, collection of selfies or vlogs, and many more digital entities. As a result, they may be used to signify the possession of any specific object, whether it be digital or physical. NFTs are digital tokens that may be used to indicate ownership of one of a-kind goods. For example, I can buy a shoe or T shirt from any store, and then if the store provides me the same 3D model of that T-Shirt or shoe of the exact same design and color, it would be more connected with my feelings. They enable us to tokenize items such as artwork, valuables, and even real estate. NFTs can only be owned by one person at a time, and they are protected by the Ethereum blockchain no one can alter the ownership record or create a new NFT. The word non-fungible can be used to describe items like your furniture, a song file, or your computer. It is impossible to substitute these goods with anything else because they each have their own distinct characteristics. The goal was to find all the existing implementations of Non-fungible Tokens in different fields of recent technology, so that an overall overview of future implementations of NFT can be found and how it can be used to enrich user experiences. △ Less

Submitted 30 September, 2022; originally announced September 2022.

Convolutional Neural Network: Text Classification Model for Open Domain Question Answering System

Authors: Muhammad Zain Amin, Noman Nadeem

Abstract: Recently machine learning is being applied to almost every data domain one of which is Question Answering Systems (QAS). A typical Question Answering System is fairly an information retrieval system, which matches documents or text and retrieve the most accurate one. The idea of open domain question answering system put forth, involves convolutional neural network text classifiers. The Classificat… ▽ More Recently machine learning is being applied to almost every data domain one of which is Question Answering Systems (QAS). A typical Question Answering System is fairly an information retrieval system, which matches documents or text and retrieve the most accurate one. The idea of open domain question answering system put forth, involves convolutional neural network text classifiers. The Classification model presented in this paper is multi-class text classifier. The neural network classifier can be trained on large dataset. We report series of experiments conducted on Convolution Neural Network (CNN) by training it on two different datasets. Neural network model is trained on top of word embedding. Softmax layer is applied to calculate loss and mapping of semantically related words. Gathered results can help justify the fact that proposed hypothetical QAS is feasible. We further propose a method to integrate Convolutional Neural Network Classifier to an open domain question answering system. The idea of Open domain will be further explained, but the generality of it indicates to the system of domain specific trainable models, thus making it an open domain. △ Less

Submitted 6 October, 2019; v1 submitted 7 September, 2018; originally announced September 2018.

Intelligent Paging Strategy for Multi-Carrier CDMA System

Authors: Sheikh Shanawaz Mostafa, Khondker Jahid Reza, Md. Ziaul Amin, Mohiuddin Ahmad

Abstract: Subscriber satisfaction and maximum radio resource utilization are the pivotal criteria in communication system design. In multi-Carrier CDMA system, different paging algorithms are used for locating user within the shortest possible time and best possible utilization of radio resources. Different paging algorithms underscored different techniques based on the different purposes. However, low serv… ▽ More Subscriber satisfaction and maximum radio resource utilization are the pivotal criteria in communication system design. In multi-Carrier CDMA system, different paging algorithms are used for locating user within the shortest possible time and best possible utilization of radio resources. Different paging algorithms underscored different techniques based on the different purposes. However, low servicing time of sequential search and better utilization of radio resources of concurrent search can be utilized simultaneously by swapping of the algorithms. In this paper, intelligent mechanism has been developed for dynamic algorithm assignment basing on time-varying traffic demand, which is predicted by radial basis neural network; and its performance has been analyzed are based on prediction efficiency of different types of data. High prediction efficiency is observed with a good correlation coefficient (0.99) and subsequently better performance is achieved by dynamic paging algorithm assignment. This claim is substantiated by the result of proposed intelligent paging strategy. △ Less

Submitted 7 December, 2011; originally announced December 2011.

An Efficient Paging Algorithm for Multi-Carrier CDMA System

Authors: Sheikh Shanawaz Mostafa, Khondker Jahid Reza, Gazi Maniur Rashid, Muhammad Moinuddin, Md. Ziaul Amin, Abdullah Al Nahid

Abstract: To cope with the increasing demand of wireless communication services multi-carrier systems are being used. Radio resources are very limited and efficient usages of these resources are inevitable to get optimum performance of the system. Paging channel is a low-bandwidth channel and one of the most important channels on which system performance depends significantly. Therefore it is vulnerable to… ▽ More To cope with the increasing demand of wireless communication services multi-carrier systems are being used. Radio resources are very limited and efficient usages of these resources are inevitable to get optimum performance of the system. Paging channel is a low-bandwidth channel and one of the most important channels on which system performance depends significantly. Therefore it is vulnerable to even moderate overloads. In this paper, an efficient paging algorithm, Concurrent Search, is proposed for efficient use of paging channel in Multi- carrier CDMA system instead of existing sequential searching algorithm. It is shown by the simulation that the paging performance in proposed algorithm is far better than the existing system. △ Less

Submitted 11 September, 2011; originally announced September 2011.