Recent Submissions

  • Resistive Neural Hardware Accelerators

    Smagulova, Kamilya; Fouda, Mohammed E.; Kurdahi, Fadi; Salama, Khaled N.; Eltawil, Ahmed (arXiv, 2021-09-08) [Preprint]
    Deep Neural Networks (DNNs), as a subset of Machine Learning (ML) techniques, entail that real-world data can be learned and that decisions can be made in real-time. However, their wide adoption is hindered by a number of software and hardware limitations. The existing general-purpose hardware platforms used to accelerate DNNs are facing new challenges associated with the growing amount of data and are exponentially increasing the complexity of computations. An emerging non-volatile memory (NVM) devices and processing-in-memory (PIM) paradigm is creating a new hardware architecture generation with increased computing and storage capabilities. In particular, the shift towards ReRAM-based in-memory computing has great potential in the implementation of area and power efficient inference and in training large-scale neural network architectures. These can accelerate the process of the IoT-enabled AI technologies entering our daily life. In this survey, we review the state-of-the-art ReRAM-based DNN many-core accelerators, and their superiority compared to CMOS counterparts was shown. The review covers different aspects of hardware and software realization of DNN accelerators, their present limitations, and future prospectives. In particular, comparison of the accelerators shows the need for the introduction of new performance metrics and benchmarking standards. In addition, the major concerns regarding the efficient design of accelerators include a lack of accuracy in simulation tools for software and hardware co-design.

    Baumgartner, Stefan; Heitzinger, Clemens (COMMUNICATIONS IN MATHEMATICAL SCIENCES, 2012) [Article]
    We present existence and local uniqueness theorems for a system of partial differential equations modeling field-effect nano-sensors. The system consists of the Poisson(-Boltzmann) equation and the drift-diffusion equations coupled with a homogenized boundary layer. The existence proof is based on the Leray-Schauder fixed-point theorem and a maximum principle is used to obtain a-priori estimates for the electric potential, the electron density, and the hole density. Local uniqueness around the equilibrium state is obtained from the implicit-function theorem. Due to the multiscale problem inherent in field-effect biosensors, a homogenized equation for the potential with interface conditions at a surface is used. These interface conditions depend on the surfacecharge density and the dipole-moment density in the boundary layer and still admit existence and local uniqueness of the solution when certain conditions are satisfied. Due to the geometry and the boundary conditions of the physical system, the three-dimensional case must be considered in simulations. Therefore a finite-volume discretization of the 3d self-consistent model was implemented to allow comparison of simulation and measurement. Special considerations regarding the implementation of the interface conditions are discussed so that there is no computational penalty when compared to the problem without interface conditions. Numerical simulation results are presented and very good quantitative agreement with current-voltage characteristics from experimental data of biosensors is found.