RESUMEN
As a fundamental problem in algorithmic trading, portfolio optimization aims to maximize the cumulative return by continuously investing in various financial derivatives within a given time period. Recent years have witnessed the transformation from traditional machine learning trading algorithms to reinforcement learning algorithms due to their superior nature of sequential decision making. However, the exponential growth of the imperfect and noisy financial data that is supposedly leveraged by the deterministic strategy in reinforcement learning, makes it increasingly challenging for one to continuously obtain a profitable portfolio. Thus, in this work, we first reconstruct several deterministic and stochastic reinforcement algorithms as benchmarks. On this basis, we introduce a risk-aware reward function to balance the risk and return. Importantly, we propose a novel interpretable stochastic reinforcement learning framework which tailors a stochastic policy parameterized by Gaussian Mixtures and a distributional critic realized by quantiles for the problem of portfolio optimization. In our experiment, the proposed algorithm demonstrates its superior performance on U.S. market stocks with a 63.1% annual rate of return while at the same time reducing the market value max drawdown by 10% when back-testing during the stock market crash around March 2020.
RESUMEN
Pd-SnO2 composite nanoceramics have been prepared from SnO2 and Pd nanoparticles through traditional pressing and sintering. Their responses to CO at room temperature are found to depend greatly on the content of Pd. For those samples with 1.0 and 5.0 mol% Pd, their resistance increases dramatically upon being exposed to CO in air; while for samples of 0.2 mol% Pd, their resistance decreases greatly upon being exposed to CO in air, and extraordinary room-temperature CO sensing capabilities, including high sensitivities around 15, short response time of 20 s and recovery time of 60 s for 100 ppm CO in air, a high selectivity against H2, have been observed for them. X-ray photoelectron spectroscopy analyses showed that Pd2+ was formed in samples of 1 mol% Pd, while both Pd2+ and Pd4+ were formed in samples of 0.2 mol% Pd. It is proposed that for Pd-SnO2 composite nanoceramics, Pd2+ is responsible for CO-induced increase while Pd4+ is responsible for CO-induced decrease in resistance.
RESUMEN
Cooling methods for multiple heat sources with high heat flux have rarely been reported, but such situations threaten the stable operation of electronic devices. Therefore, in this paper, the use of two microchannel heat sinks is proposed, with and without grooves, labeled Type A and Type B, respectively. Experimental investigations on the flow boiling of two microchannel heat sinks connected in parallel and in series are carried out under different mass fluxes. In addition, a high-speed camera is used to observe flow patterns in the microchannels. The cold plate wall temperature (Tw), heat transfer coefficient (HTC), and pressure drop (PD) are obtained with the use of two microchannel heat sinks. The flow patterns of the bubbly flow and elongated bubbles in the microchannels are observed. The results of the analysis indicated that the Tw, HTC, and PD of the two microchannel heat sinks connected in parallel were degraded, especially when using the Type A-B parallel connection. Compared to the use of a single heat sink, the maximum decrease in HTC was 9.44 kW/(m2K) for Type A heat sinks connected in parallel, which represents a decrease of 45.95%. The influence of the series connection on the Tw, HTC, and PD of the two heat sinks is obvious. The Type A-A series connection exerted the greatest positive effect on the performance of the two heat sinks, especially in the case of the postposition heat sink. The maximum increase in HTC was 12.77 kW/(m2K) for the postposition Type A heat sink, representing an increase of 72.88%. These results could provide a reference for a two-phase flow-cooling complex for multiple heat sources with high heat flux.