Junkyu Lee

Modern deep neural networks heavily rely on a large number of multiply-accumulate operations, which constitute the predominant computational cost. To address this, Look-Up Table (LUT)-based matrix multiplications have emerged as a promising alternative for reducing the computational cost and time of the multiply-accumulate operations in a neural network. However, the LUT-based neural network still faces the scalability challenge due to the inherent limitations of LUT-based matrix multiplication. To mitigate these scalability limitations, this paper proposes a scalable and energy-efficient LUT-based approximate matrix multiplication unit (LUT-MU) constituting the basic component of the neural networks by integrating a pruning strategy on the MADDNESS algorithm, a LUT-based matrix multiplication methodology. With increasing problem size and precision demands in matrix multiplication, our proposed LUT-MU architecture effectively constrains resource expansion. The case study shows that deploying our LUT-MU in neural network architectures, including fully connected layers (MNIST) and ResNets (CIFAR-10, ImageNet)—on XCZU7EV and XCZU19EG FPGAs, produces up to 1.6× throughput improvement and 4.2× energy efficiency gains over mainstream CUDA-based network implementations, and 1.8× energy efficiency compared to leading quantised neural network implementations, with moderate impact on accuracy. Compared to original MADDNESS-based neural networks, our LUT-MU shows 1.3 to 2.6× resource savings based on various resolution configuration settings of MADDNESS.

This paper presents a first empirical study of agentic AI as autonomous decision-makers in decentralized governance. Using more than 3K proposals from major protocols, we build an agentic AI voter that interprets proposal contexts, retrieves historical deliberation data, and independently determines its voting position. The agent operates within a realistic financial simulation environment grounded in verifiable blockchain data, implemented through a modular composable program (MCP) workflow that defines data flow and tool usage via Agentics framework. We evaluate how closely the agent's decisions align with the human and token-weighted outcomes, uncovering strong alignments measured by carefully designed evaluation metrics. Our findings demonstrate that agentic AI can augment collective decision-making by producing interpretable, auditable, and empirically grounded signals in realistic DAO governance settings. The study contributes to the design of explainable and economically rigorous AI agents for decentralized financial systems.

AI-powered logistics struggles to meet GDPR compliance without degrading service quality. Drawing on insights from the PALLETS project, this article outlines a structured approach to embedding privacy and security by design with the Secure Tropos tool, ensuring compliance while minimizing impact on system performance.

In the post-transistor scaling era, defined in this paper as the era in which further reductions in operational voltage and capacitance by shrinking transistor size are no longer achievable, architectural optimizations alone become insufficient, necessitating the interdisciplinary optimizations including Algorithmic, Arithmetic, and Architectural optimizations (3A optimizations). To the best of our knowledge, no unified framework exists that captures the interactions of such cross-discipline optimizations and their collective impact on energy efficiency. Therefore, we propose such a unified framework that explains how interdisciplinary 3A optimizations interact and contribute to overall energy efficiency. To develop the framework, we introduce a new energy-efficiency metric, Transistor Arithmetic Density (TAD), defined as the arithmetic density per transistor toggling activity. We conclude that energy-efficient computing strategies in the post scaling era should focus on improving the two key components, operational efficiency and TAD, without affecting accuracy and stability of the algorithm.

Many practical reinforcement learning environments have a discrete factored action space that induces a large combinatorial set of actions, thereby posing significant challenges. Existing approaches leverage the regular structure of the action space and resort to a linear decomposition of Q-functions, which avoids enumerating all combinations of factored actions. In this paper, we consider Q-functions defined over a lower dimensional projected subspace of the original action space, and study the condition for the unbiasedness of decomposed Q-functions using causal effect estimation from the no unobserved confounder setting in causal statistics. This leads to a general scheme which we call action decomposed reinforcement learning that uses the projected Q-functions to approximate the Q-function in standard model-free reinforcement learning algorithms. The proposed approach is shown to improve sample complexity in a model-based reinforcement learning setting. We demonstrate improvements in sample efficiency compared to state-of-the-art baselines in online continuous control environments and a real-world offline sepsis treatment environment.

This paper presents the design and implementation of Software-Defined Floating-Point (SDF) number formats for high-speed implementation of the Belief Propagation (BP) algorithm. SDF formats are designed specifically to meet the numeric needs of the computation and are more compact representations of the data. They reduce memory footprint and memory bandwidth requirements without sacrificing accuracy, given that BP for loopy graphs inherently involves algorithmic errors. This paper designs several SDF formats for sum-product BP applications by careful analysis of the computation. Our theoretical analysis leads to the design of 16-bit (half-precision) and 8-bit (mini-precision) widths. We moreover present highly efficient software implementation of the proposed SDF formats which is centered around conversion to hardware-supported single-precision arithmetic hardware. Our solution demonstrates negligible conversion overhead on commercially available CPUs. For Ising grids with sizes from 100×100 to 500×500, the 16- and 8-bit SDF formats along with our conversion module produce equivalent accuracy to double-precision floating-point format but with 2.86× speedups on average on an Intel Xeon processor. Particularly, increasing the grid size results in higher speed-up. For example, the proposed half-precision format with 3-bit exponent and 13-bit mantissa achieved the minimum and maximum speedups of 1.30× and 1.39× over single-precision, and 2.55× and 3.40× over double-precision, by increasing grid size from 100×100 to 500×500.

The ability to generate multiple plans is central to using planning in real-life applications. Top-quality planners generate sets of such top-cost plans, allowing flexibility in determining equivalent ones. In terms of the order between actions in a plan, the literature only considers two extremes -- either all orders are important, making each plan unique, or all orders are unimportant, treating two plans differing only in the order of actions as equivalent. To allow flexibility in selecting important orders, we propose specifying a subset of actions the orders between which are important, interpolating between the top-quality and unordered top-quality planning problems. We explore the ways of adapting partial order reduction search pruning techniques to address this new computational problem and present experimental evaluations demonstrating the benefits of exploiting such techniques in this setting.