1: \begin{abstract}

2: Disentangling attributes of various sensory signals is central to human-like perception and reasoning and a critical task for higher-order cognitive and neuro-symbolic AI systems. An elegant approach to represent this intricate factorization is via high-dimensional holographic vectors drawing on brain-inspired vector symbolic architectures. However, holographic factorization involves iterative computation with high-dimensional matrix-vector multiplications and suffers from non-convergence problems.

3: 

4: In this paper, we present \textsc{H3DFact}, a heterogeneous 3D integrated in-memory compute engine capable of efficiently factorizing high-dimensional holographic representations. \textsc{H3DFact} exploits the computation-in-superposition capability of holographic vectors and the intrinsic stochasticity associated with memristive-based 3D compute-in-memory. Evaluated on large-scale factorization and perceptual problems, \textsc{H3DFact} demonstrates superior capability in factorization accuracy and operational capacity by up to five orders of magnitude,

5: % , as well as substantially lowering the computational time and space complexity. 

6: with 5.5$\times$ compute density, 1.2$\times$ energy efficiency improvements, and 5.9$\times$ less silicon footprint compared to iso-capacity 2D designs.

7: % Moreover, we experimentally demonstrate the ability to reliably and efficiently factorize visual perceptual representations.

8: 

9: 

10: % Factorizing the attributes of sensory inputs and high-order models is crucial in advanced neurosymbolic AI systems. Such a capability is offered through the brain-inspired Hyperdimensional computing, which allows 

11: 

12: % This paper presents H3DFact, an efficient compute engine capable of factorizing perceptual concepts represented as holographic hyperdimensional vectors. 

13: % Such a capability is made possible through hyperdimensional computing, which 

14: \end{abstract}

15: