Mr. Abderrahim “Rahim” Amlou

Abstract:

Memory caching is a common practice to reduce application latencies by buffering relevant data in high-speed memory. When the volume of data to cache is too large or a DRAM-based solution too expensive, several technologies such as NVM or high-speed SSDs could complement DRAM to form a multi-tier cache. Additionally, most existing policies focus on categorizing the data based on factors like recency and frequency, setting aside the fact that applications/customers have varying Quality-of-Service requirements. This concept is well established in Cloud environment with Service Level Agreement (SLA). In this paper, by extending the Adaptive Replacement Cache (ARC), that uses recency and frequency lists, we propose a QoS-aware Multi-tier Adaptive Replacement Cache (QMARC) policy with the ability to consider data applications/customers priorities through the concept of penalty borrowed from the Cloud. QM-ARC is generic, as it can be applied whatever the number of tiers and can accommodate different penalty functions. Using synthetic and real traces, our solution improved QoS as compared to state-of-the-artwork.

Biography:

Mr. Abderrahim “Rahim” Amlou is a researcher at the National Institute of Standards and Technology (NIST) in Gaithersburg, Maryland. Mr. Amlou is specializing in networking and embedded software development. His work is characterized by a deep commitment to advancing technology, with a primary focus on exploring the intricacies of connectivity and communication systems.
Mr. Amlou is actively engaged in the evolving field of quantum networking, specifically focusing on simulation and software development for experiments. In addition, he is serving as an officer for both the Wash DC ComSoc as well as EMBS Chapters.

Address:United States

Mr. Nii-Emil Alexander-Reindorf

Abstract:

Artificial intelligence (AI) researchers are constantly looking for new and improved methods for making decisions. To improve the decision-making process of the AI systems, this PhD research investigates a novel combination of the Counterfactual Regret Minimization (CFR), Monte Carlo Tree Search (MCTS), and Multi-Agent Decision Planning (MADP) algorithms. As of today, there is no recognized method for the programmatic implementation and experimental assessment of such integrated algorithms. The difficulties in integrating these many decision-making processes, along with the lack of a standardized experimental protocol to evaluate the functionality and properties of an integrated system, is a challenging task. This integration process can be viewed as an optimization problem, with the goal being to find the best possible joint policy that maximizes the total expected benefit.

The potential synergy by integrating the above algorithms in the field of decision-making algorithms has not been investigated or validated so far, despite their separate strengths and uses. In addition, there is currently a need for a theoretical background required to incorporate these algorithms into a logical framework. By combining their own features and performances, such integration may produce a solid decision-making system that can handle a wider variety of AI challenging tasks. The new approach called the Integrated Multi-Agent Decision Planning (IMADP) presented in this PhD research is to provide a hybrid solution for a constrained scalable decision-making algorithm (centralized and distributed coordination) in multi-agent systems (MAS).

Our objective research was to investigate if this combined approach works better than using MADP alone, especially in complex situations where multiple decision-makers are involved. In this regard we incorporated CFR and MCTS. We set up various tests and did extensive simulations to compare these methods and algorithms, focusing on how quickly they find good strategies, how well they perform, and how efficiently they use computer resources like time and memory.

The results were quite clear: the integrated approach of IMADP outperformed the standard MADP in almost every way. IMADP was quicker at finding effective strategies, made better decisions, and even with its added complexity, IMADP did not require much more time or memory than MADP. Our findings show that combining CFR and MCTS with MADP is a promising way to improve decision-making in complex, multi-agent scenarios.

Biography:

Mr. Nii-Emil Alexander-Reindorf is a PhD candidate at the University of the District of Columbia. Mr. Alexander-Reindorf has both a bachelor’s and master’s degree in computer science from the institution. The focus of his research has been on the intersection of cybersecurity with game theory. Mr, Alexander-Reindorf has over two decades of experience with software development. Mr. Alexander-Reindorf has worked in the public sector for both the Federal Aviation Administration and Department of Veterans Affairs. Within the private sector, he worked at Price Waterhouse Coopers and is currently working for JPMorgan-Chase as a data architect.