56016 research outputs found
Sort by
Large-scale financial planning via a partially observable stochastic dual dynamic programming framework
No full textThe multi-stage stochastic programming (MSP) approach is widely used to solve financial planning problems owing to its flexibility. However, the size of an MSP problem grows exponentially with the number of stages, and such problem can easily become computationally intractable. Financial planning problems often consider planning horizons of several decades, and thus, the curse of dimensionality can become a critical issue. Stochastic dual dynamic programming (SDDP), a sampling-based decomposition algorithm, has emerged to resolve this issue. While SDDP has been successfully implemented in the energy domain, few applications of SDDP are found in the finance domain. In this study, we identify the major obstacle in using SDDP to solve financial planning problems to be the stagewise independence assumption and propose a partially observable SDDP (PO-SDDP) framework to overcome such limitations. We argue that the PO-SDDP framework, which models uncertainties using discrete-valued partially observable Markov states and introduces feasibility cuts, can properly address large-scale financial planning problems
Effect of Top-Gate Dielectric Deposition on the Performance of Indium Tin Oxide Transistors
No full textWe report ultrathin (similar to 4 nm) channel indium tin oxide (ITO) transistors, comparing different precursors for atomic layer deposition (ALD) of the Al2O3 top-gate dielectric, and analyze the role of dielectric deposition on transistor performance and gate bias stress stability. Water-based ALD leads to very negative threshold voltage (V-T), with devices remaining in the on-state. In contrast, both ozone and O-2-plasma precursors yield devices that can turn off, but ozone-based ALD devices have less negative V-T shift at short channel lengths, and relatively more positive V-T at all channel lengths. We achieve maximum drive current, I-max approximate to 260 (mu)A/(mu)m at V-DS = 1 V, on/off current ratio of >10(10) (limited by the instrument's noise floor) for L approximate to 700 nm ozone-Al2O3 top-gated transistors. Across multiple devices, the effective mobility is similar to 42 cm(2)V(-1)s(-1) and contact resistance is similar to 376 Omega center dot mu m. The transistors also show good gate bias stability with normalized VT shift of +0.12 V(MV/cm)(-1) at gate stress field >3 MV/cm, a similar to 3x improvement vs. our previous reports of uncapped ITO transistors
Experimental characterization of the flowline of a lithium film formed using an electromagnetic thruster for a RAON prototype charge stripper
No full textThis paper describes the variable optimization of a lithium charge-stripper device for the Rare isotope Accelerator complex for ON-line experiences (RAON) and the experimental characterization of a lithium film formed for the charge removal of uranium using a magnetohydrodynamic liquid metal circulation system. To this end, liquid lithium charge stripper was fabricated to form a thin lithium film flow through collisions between a liquid lithium jet and a planar deflector. The flow characteristics of the lithium film were analyzed through simulation and waterjet experiments in terms of the liquid lithium film thickness. To circulate the liquid lithium, a mechanically safe electromagnetic thruster with precise pressure control was designed by analyzing its electromagnetic and hydraulic parameters through finite element simulation. The analysis of the geometrical arrangement of the permanent magnet and input current of the thruster revealed that compared to existing electromagnetic thrusters, the nominal input current of the newly-constructed thruster was reduced to 57% and its weight reduced to 6%, while facilitating easy maintenance. The optimized nozzle diameter and angle of the lithium charge stripper was 0.7 mm and 34??, respectively. Further, the formation of a 22-??m thick liquid lithium thin film required to obtain the 79+ charge state of uranium was confirmed at an input current of 107 A
Kernel Code Integrity Protection at the Physical Address Level on RISC-V
No full textAn operating system kernel has the highest privilege in most computer systems, making its code integrity critical to the entire system's security. Failure to protect the kernel code integrity allows an attacker to modify the kernel code pages directly or trick the kernel into executing instructions stored outside the kernel code pages. Existing prevention mechanisms rely on the memory management unit in which certain memory pages are marked as not-executable in supervisor mode to prevent such attacks. However, an attacker can bypass these existing mechanisms by directly manipulating the page table contents to mark the memory pages with malicious code as supervisor-executable. This paper shows that a small architectural extension enables a physical address-level mechanism to stop this threat without relying on page table integrity. PRIV LOCK lets, at boot time, the kernel specifies the physical address ranges containing its code. At run time, PRIV LOCK ensures that the content within the range is not manipulated and that only the instructions from those pages are executed while the processor runs in supervisor mode. Despite this protection, the kernel can still create new code pages (e.g., for loadable kernel modules) and make them executable with the help of PRIV LOCK 's secure loader. The experimental results show that PRIV LOCK incurs low performance (<0.5%), area (0.14-0.3%), and energy/power (0.053-2%) overhead
Effective Slogan Generation with Noise Perturbation
No full textSlogans play a crucial role in building the brand???s identity of the firm. A slogan is expected to reflect firm???s mission/vision and the brand???s value propositions in memorable and likeable ways. Automating the generation of slogans with such characteristics is challenging. Previous studies developed and tested slogan generation with syntactic control and summarization models which lack distinctive generation. We introduce a novel approach of generating slogans utilizing pre-trained transformer T5 model and apply noise perturbation to the input embeddings, as a contributing factor in generating distinctive and cohered slogans. Furthermore, the proposed approach incorporates descriptions about the firm and target brand into the generation of slogans. We evaluate generated slogans based on ROUGE-1, ROUGE-L and Cosine Similarity metrics and also assess them with human subjects in terms of slogan???s distinctiveness, coherence, and fluency. The results demonstrate that our approach perturbing the embedding layer with Gaussian noise, yields better performance than baseline models and
other transformer-based models
Compute-in-Memory??? Time-Domain ????????? ????????? Spiking Neural Network ?????????
No full textSpiking Neural Network (SNN) draws attention for low energy consumption due to transmitting only binary information. Another way to implement low power hardware is Compute-in-Memory (CIM). It removes external memory access and minimizes data transaction energy. However, power-hungry Analog-to-Digital Converter (ADC) is essential in analog CIM. In time-domain computation, ADC is replaced with time-to-digital converter. Nevertheless, time-domain CIM supporting multi-bit computation has low energy efficiency by digital logic. To resolve this issue, this paper proposes time-domain SNN CIM with no digital circuit and it improves 83% throughput by pipelined architecture. It achieves 701.7 TOPS/W of energy efficiency