-
CIMR-V: An End-to-End SRAM-based CIM Accelerator with RISC-V for AI Edge Device
Authors:
Yan-Cheng Guo and,
Tian-Sheuan Chang,
Chih-Sheng Lin,
Bo-Cheng Chiou,
Chih-Ming Lai,
Shyh-Shyuan Sheu,
Wei-Chung Lo,
Shih-Chieh Chang
Abstract:
Computing-in-memory (CIM) is renowned in deep learning due to its high energy efficiency resulting from highly parallel computing with minimal data movement. However, current SRAM-based CIM designs suffer from long latency for loading weight or feature maps from DRAM for large AI models. Moreover, previous SRAM-based CIM architectures lack end-to-end model inference. To address these issues, this…
▽ More
Computing-in-memory (CIM) is renowned in deep learning due to its high energy efficiency resulting from highly parallel computing with minimal data movement. However, current SRAM-based CIM designs suffer from long latency for loading weight or feature maps from DRAM for large AI models. Moreover, previous SRAM-based CIM architectures lack end-to-end model inference. To address these issues, this paper proposes CIMR-V, an end-to-end CIM accelerator with RISC-V that incorporates CIM layer fusion, convolution/max pooling pipeline, and weight fusion, resulting in an 85.14\% reduction in latency for the keyword spotting model. Furthermore, the proposed CIM-type instructions facilitate end-to-end AI model inference and full stack flow, effectively synergizing the high energy efficiency of CIM and the high programmability of RISC-V. Implemented using TSMC 28nm technology, the proposed design achieves an energy efficiency of 3707.84 TOPS/W and 26.21 TOPS at 50 MHz.
△ Less
Submitted 27 March, 2025;
originally announced March 2025.
-
Scaffold-Assisted Window Junctions for Superconducting Qubit Fabrication
Authors:
Chung-Ting Ke,
Jun-Yi Tsai,
Yen-Chun Chen,
Zhen-Wei Xu,
Elam Blackwell,
Matthew A. Snyder,
Spencer Weeden,
Peng-Sheng Chen,
Chih-Ming Lai,
Shyh-Shyuan Sheu,
Zihao Yang,
Cen-Shawn Wu,
Alan Ho,
R. McDermott,
John Martinis,
Chii-Dong Chen
Abstract:
The superconducting qubit is one of the promising directions in realizing fault-tolerant quantum computing (FTQC), which requires many high-quality qubits. To achieve this, it is desirable to leverage modern semiconductor industry technology to ensure quality, uniformity, and reproducibility. However, conventional Josephson junction fabrication relies mainly on resist-assistant double-angle evapor…
▽ More
The superconducting qubit is one of the promising directions in realizing fault-tolerant quantum computing (FTQC), which requires many high-quality qubits. To achieve this, it is desirable to leverage modern semiconductor industry technology to ensure quality, uniformity, and reproducibility. However, conventional Josephson junction fabrication relies mainly on resist-assistant double-angle evaporation, posing integration challenges. Here, we demonstrate a lift-off-free qubit fabrication that integrates seamlessly with existing industrial technologies. This method employs a silicon oxide (SiO$_2$) scaffold to define an etched window with a well-controlled size to form a Josephson junction. The SiO$_2$, which has a large dielectric loss, is etched away in the final step using vapor HF leaving little residue. This Window junction (WJ) process mitigates the degradation of qubit quality during fabrication and allows clean removal of the scaffold. The WJ process is validated by inspection and Josephson junction measurement. The scaffold removal process is verified by measuring the quality factor of the resonators. Furthermore, compared to scaffolds fabricated by plasma-enhanced chemical vapor deposition (PECVD), qubits made by WJ through physical vapor deposition (PVD) achieve relaxation time up to $57\,μ\text{s}$. Our results pave the way for a lift-off-free qubit fabrication process, designed to be compatible with modern foundry tools and capable of minimizing damage to the substrate and material surfaces.
△ Less
Submitted 13 March, 2025;
originally announced March 2025.
-
A generic approach to homogenization of a diffusion driven by growing incompressible drift
Authors:
Brice Franke,
Shuenn-Jyi Sheu
Abstract:
We study how the resolvent-family of a diffusion behaves, as thedrift grows to infinity. The limit turns out to be a selfadjoint pseudo-resolvent.After reduction of the underlying Hilbert-space, this pseudo-resolvent becomesa resolvent to a strongly continuous semi-group of contractions. We prove thatthis semi-group is associated to some Hunt-process on some suitable state-space which is construct…
▽ More
We study how the resolvent-family of a diffusion behaves, as thedrift grows to infinity. The limit turns out to be a selfadjoint pseudo-resolvent.After reduction of the underlying Hilbert-space, this pseudo-resolvent becomesa resolvent to a strongly continuous semi-group of contractions. We prove thatthis semi-group is associated to some Hunt-process on some suitable state-space which is constructed from equivalence classes of the drifts trajectories.Finally, we show a distributional limit theorem for the accelerated diffusiontoward the associated Hunt process.
△ Less
Submitted 21 August, 2024; v1 submitted 14 May, 2024;
originally announced May 2024.
-
Expected exponential utility maximization of insurers with a general diffusion factor model : The complete market case
Authors:
Hiroaki Hata,
Shuenn-Jyi Sheu,
Li-Hsien Sun
Abstract:
In this paper, we consider the problem of optimal investment by an insurer. The insurer invests in a market consisting of a bank account and $m$ risky assets. The mean returns and volatilities of the risky assets depend nonlinearly on economic factors that are formulated as the solutions of general stochastic differential equations. The wealth of the insurer is described by a Cramér--Lundberg proc…
▽ More
In this paper, we consider the problem of optimal investment by an insurer. The insurer invests in a market consisting of a bank account and $m$ risky assets. The mean returns and volatilities of the risky assets depend nonlinearly on economic factors that are formulated as the solutions of general stochastic differential equations. The wealth of the insurer is described by a Cramér--Lundberg process, and the insurer preferences are exponential. Adapting a dynamic programming approach, we derive Hamilton--Jacobi--Bellman (HJB) equation. And, we prove the unique solvability of HJB equation. In addition, the optimal strategy is also obtained using the coupled forward and backward stochastic differential equations (FBSDEs). Finally, proving the verification theorem, we construct the optimal strategy.
△ Less
Submitted 21 March, 2019;
originally announced March 2019.
-
Portfolio Optimization with Delay Factor Models
Authors:
Shuenn-Jyi Sheu,
Li-Hsien Sun,
Zheng Zhang
Abstract:
We propose an optimal portfolio problem in the incomplete market where the underlying assets depend on economic factors with delayed effects, such models can describe the short term forecasting and the interaction with time lag among different financial markets. The delay phenomenon can be recognized as the integral type and the pointwise type. The optimal strategy is identified through maximizing…
▽ More
We propose an optimal portfolio problem in the incomplete market where the underlying assets depend on economic factors with delayed effects, such models can describe the short term forecasting and the interaction with time lag among different financial markets. The delay phenomenon can be recognized as the integral type and the pointwise type. The optimal strategy is identified through maximizing the power utility. Due to the delay leading to the non-Markovian structure, the conventional Hamilton-Jacobi-Bellman (HJB) approach is no longer applicable. By using the stochastic maximum principle, we argue that the optimal strategy can be characterized by the solutions of a decoupled quadratic forward-backward stochastic differential equations(QFBSDEs). The optimality is verified via the super-martingale argument. The existence and uniqueness of the solution to the QFBSDEs are established. In addition, if the market is complete, we also provide a martingale based method to solve our portfolio optimization problem, and investigate its connection with the proposed FBSDE approach. Finally, two particular cases are analyzed where the corresponding FBSDEs can be solved explicitly.
△ Less
Submitted 3 May, 2018;
originally announced May 2018.
-
Asymptotics of the probability minimizing a "down-side" risk
Authors:
Hiroaki Hata,
Hideo Nagai,
Shuenn-Jyi Sheu
Abstract:
We consider a long-term optimal investment problem where an investor tries to minimize the probability of falling below a target growth rate. From a mathematical viewpoint, this is a large deviation control problem. This problem will be shown to relate to a risk-sensitive stochastic control problem for a sufficiently large time horizon. Indeed, in our theorem we state a duality in the relation b…
▽ More
We consider a long-term optimal investment problem where an investor tries to minimize the probability of falling below a target growth rate. From a mathematical viewpoint, this is a large deviation control problem. This problem will be shown to relate to a risk-sensitive stochastic control problem for a sufficiently large time horizon. Indeed, in our theorem we state a duality in the relation between the above two problems. Furthermore, under a multidimensional linear Gaussian model we obtain explicit solutions for the primal problem.
△ Less
Submitted 13 January, 2010;
originally announced January 2010.
-
Sodium ion ordering of Na0.77CoO2 under competing multi-vacancy cluster, superlattice and domain formation
Authors:
F. -T. Huang,
G. J. Shu,
M. -W. Chu,
Y. K. Kuo,
W. L. Lee,
H. S. Sheu,
F. C. Chou
Abstract:
Hexagonal superlattice formed by sodium multi-vacancy cluster ordering in Na$_{0.77}$CoO$_2$ has been proposed based on synchrotron X-ray Laue diffraction study on electrochemically fine-tuned single crystals. The title compound sits closely to the proposed lower end of the miscibility gap of x ~ 0.77-0.82 phase separated range. The average sodium vacancy cluster size is estimated to be 4.5 Na v…
▽ More
Hexagonal superlattice formed by sodium multi-vacancy cluster ordering in Na$_{0.77}$CoO$_2$ has been proposed based on synchrotron X-ray Laue diffraction study on electrochemically fine-tuned single crystals. The title compound sits closely to the proposed lower end of the miscibility gap of x ~ 0.77-0.82 phase separated range. The average sodium vacancy cluster size is estimated to be 4.5 Na vacancies per layer within a large superlattice size of sqrt{19}a*sqrt{19}a*3c. The exceptionally large Na vacancy cluster size favors large twinned simple hexagonal superlattice of sqrt{19}a, in competition with the smaller di-, tri- and quadri-vacancy clusters formed superlattices of sqrt{12}a and sqrt{13}a. Competing electronic correlations are revealed by the observed spin glass-like magnetic hysteresis below ~ 3K and the twin, triple and mono domain transformations during thermal cycling between 273-373K.
△ Less
Submitted 28 September, 2009;
originally announced September 2009.
-
Max-plus Stochastic Control and Risk-sensitivity
Authors:
Wendell H. Fleming,
Hidehiro Kaise,
Shuenn-Jyi Sheu
Abstract:
In the Maslov idempotent probability calculus, expectations of random variables are defined so as to be linear with respect to max-plus addition and scalar multiplication. This paper considers control problems in which the objective is to minimize the max-plus expectation of some max-plus additive running cost. Such problems arise naturally as limits of some types of risk sensitive stochastic co…
▽ More
In the Maslov idempotent probability calculus, expectations of random variables are defined so as to be linear with respect to max-plus addition and scalar multiplication. This paper considers control problems in which the objective is to minimize the max-plus expectation of some max-plus additive running cost. Such problems arise naturally as limits of some types of risk sensitive stochastic control problems. The value function is a viscosity solution to a quasivariational inequality (QVI) of dynamic programming. Equivalence of this QVI to a nonlinear parabolic PDE with discontinuous Hamiltonian is used to prove a comparison theorem for viscosity sub- and super-solutions. An example from math finance is given, and an application in nonlinear H-infinity control is sketched.
△ Less
Submitted 20 January, 2009;
originally announced January 2009.
-
Sodium vacancy ordering and the co-existence of localized spins and itinerant charges in NaxCoO2
Authors:
F. C. Chou,
M. -W. Chu,
G. J. Shu,
F. T. Huang,
Woei Wu Pai,
H. S. Sheu,
T. Imai,
F. L. Ning,
Patrick A. Lee
Abstract:
The sodium cobaltate family (NaxCoO2) is unique among transition metal oxides because the Co sits on a triangular lattice and its valence can be tuned over a wide range by varying the Na concentration x. Up to now detailed modeling of the rich phenomenology (which ranges from unconventional superconductivity to enhanced thermopower) has been hampered by the difficulty of controlling pure phases.…
▽ More
The sodium cobaltate family (NaxCoO2) is unique among transition metal oxides because the Co sits on a triangular lattice and its valence can be tuned over a wide range by varying the Na concentration x. Up to now detailed modeling of the rich phenomenology (which ranges from unconventional superconductivity to enhanced thermopower) has been hampered by the difficulty of controlling pure phases. We discovered that certain Na concentrations are specially stable and are associated with superlattice ordering of the Na clusters. This leads naturally to a picture of co-existence of localized spins and itinerant charge carriers. For x = 0.84 we found a remarkably small Fermi energy of 87 K. Our picture brings coherence to a variety of measurements ranging from NMR to optical to thermal transport. Our results also allow us to take the first step towards modeling the mysterious ``Curie-Weiss'' metal state at x = 0.71. We suggest the local moments may form a quantum spin liquid state and we propose experimental test of our hypothesis.
△ Less
Submitted 2 September, 2007;
originally announced September 2007.
-
Searching for Stable Na-ordered Phases in Single Crystal Samples of gamma-NaxCoO2
Authors:
G. J. Shu,
Andrea Prodi,
S. Y. Chu,
Y. S. Lee,
H. S. Sheu,
F. C. Chou
Abstract:
We report on the preparation and characterization of single crystal gamma phase NaxCoO2 with 0.25 < x < 0.84 using a non-aqueous electrochemical chronoamperemetry technique. By carefully mapping the overpotential versus x (for x < 0.84), we find six distinct stable phases with Na levels corresponding to x ~ 0.75, 0.71, 0.50, 0.43, 0.33 and 0.25. The composition with x ~0.55 appears to have a cri…
▽ More
We report on the preparation and characterization of single crystal gamma phase NaxCoO2 with 0.25 < x < 0.84 using a non-aqueous electrochemical chronoamperemetry technique. By carefully mapping the overpotential versus x (for x < 0.84), we find six distinct stable phases with Na levels corresponding to x ~ 0.75, 0.71, 0.50, 0.43, 0.33 and 0.25. The composition with x ~0.55 appears to have a critical Na concentration which separates samples with different magnetic behavior as well as different Na ion diffusion mechanisms. Chemical analysis of an aged crystal reveals different Na ion diffusion mechanisms above and below x_c ~ 0.53, where the diffusion process above x_c has a diffusion coefficient about five times larger than that below x_c. The series of crystals were studied with X-ray diffraction, susceptibility, and transport measurements. The crystal with x = 0.5 shows a weak ferromagnetic transition below T=27 K in addition to the usual transitions at T = 51 K and 88 K. The resistivity of the Curie-Weiss metallic Na0.71CoO2 composition has a very low residual resistivity, which attests to the high homogeneity of the crystals prepared by this improved electrochemical method. Our results on the various stable crystal compositions point to the importance of Na ion ordering across the phase diagram.
△ Less
Submitted 2 August, 2007;
originally announced August 2007.
-
Price systems for markets with transaction costs and control problems for some finance problems
Authors:
Tzuu-Shuh Chiang,
Shang-Yuan Shiu,
Shuenn-Jyi Sheu
Abstract:
In a market with transaction costs, the price of a derivative can be expressed in terms of (preconsistent) price systems (after Kusuoka (1995)). In this paper, we consider a market with binomial model for stock price and discuss how to generate the price systems. From this, the price formula of a derivative can be reformulated as a stochastic control problem. Then the dynamic programming approac…
▽ More
In a market with transaction costs, the price of a derivative can be expressed in terms of (preconsistent) price systems (after Kusuoka (1995)). In this paper, we consider a market with binomial model for stock price and discuss how to generate the price systems. From this, the price formula of a derivative can be reformulated as a stochastic control problem. Then the dynamic programming approach can be used to calculate the price. We also discuss optimization of expected utility using price systems.
△ Less
Submitted 27 February, 2007;
originally announced February 2007.
-
On the structure of solutions of ergodic type Bellman equation related to risk-sensitive control
Authors:
Hidehiro Kaise,
Shuenn-Jyi Sheu
Abstract:
Bellman equations of ergodic type related to risk-sensitive control are considered. We treat the case that the nonlinear term is positive quadratic form on first-order partial derivatives of solution, which includes linear exponential quadratic Gaussian control problem. In this paper we prove that the equation in general has multiple solutions. We shall specify the set of all the classical solut…
▽ More
Bellman equations of ergodic type related to risk-sensitive control are considered. We treat the case that the nonlinear term is positive quadratic form on first-order partial derivatives of solution, which includes linear exponential quadratic Gaussian control problem. In this paper we prove that the equation in general has multiple solutions. We shall specify the set of all the classical solutions and classify the solutions by a global behavior of the diffusion process associated with the given solution. The solution associated with ergodic diffusion process plays particular role. We shall also prove the uniqueness of such solution. Furthermore, the solution which gives us ergodicity is stable under perturbation of coefficients. Finally, we have a representation result for the solution corresponding to the ergodic diffusion.
△ Less
Submitted 27 February, 2006;
originally announced February 2006.
-
Accelerating diffusions
Authors:
Chii-Ruey Hwang,
Shu-Yin Hwang-Ma,
Shuenn-Jyi Sheu
Abstract:
Let U be a given function defined on R^d and π(x) be a density function proportional to \exp -U(x). The following diffusion X(t) is often used to sample from π(x), dX(t)=-\nabla U(X(t)) dt+\sqrt2 dW(t),\qquad X(0)=x_0. To accelerate the convergence, a family of diffusions with π(x) as their common equilibrium is considered, dX(t)=\bigl(-\nabla U(X(t))+C(X(t))\bigr) dt+\sqrt2 dW(t),\qquad X(0)=x_…
▽ More
Let U be a given function defined on R^d and π(x) be a density function proportional to \exp -U(x). The following diffusion X(t) is often used to sample from π(x), dX(t)=-\nabla U(X(t)) dt+\sqrt2 dW(t),\qquad X(0)=x_0. To accelerate the convergence, a family of diffusions with π(x) as their common equilibrium is considered, dX(t)=\bigl(-\nabla U(X(t))+C(X(t))\bigr) dt+\sqrt2 dW(t),\qquad X(0)=x_0. Let L_C be the corresponding infinitesimal generator. The spectral gap of L_C in L^2(π) (λ(C)), and the convergence exponent of X(t) to πin variational norm (ρ(C)), are used to describe the convergence rate, where λ(C)= Sup{real part of μ\dvtxμis in the spectrum of L_C, μis not zero}, {-2.8cm}ρ(C) = Inf\biggl{ρ\dvtx\int | p(t,x,y) -π(y)| dy \le g(x) e^{ρt}\biggr}.Roughly speaking, L_C is a perturbation of the self-adjoint L_0 by an antisymmetric operator C\cdot\nabla, where C is weighted divergence free. We prove that λ(C)\le λ(0) and equality holds only in some rare situations. Furthermore, ρ(C)\le λ(C) and equality holds for C=0. In other words, adding an extra drift, C(x), accelerates convergence. Related problems are also discussed.
△ Less
Submitted 12 May, 2005;
originally announced May 2005.
