arXiv subjects

Zhishun Wang

Publications and source records attributed to Zhishun Wang.

3 recordsLinked to original sources

arXiv2023

Connectivity based Real-Time fMRI Neurofeedback Training in Youth with a History of Major Depressive Disorder

Background: Real-time functional magnetic resonance imaging neurofeedback (rtfMRI-nf) has proven to be a powerful technique to help subjects to gauge and enhance emotional control. Traditionally, rtfMRI-nf has focused on emotional regulation through self-regulation of amygdala. Recently, rtfMRI studies have observed that regulation of a target brain region is accompanied by connectivity changes beyond the target region. Therefore, the aim of present study is to investigate the use of connectivity between amygdala and prefrontal regions as the target of neurofeedback training in healthy individuals and subjects with a life-time history of major depressive disorder (MDD) performing an emotion regulation task. Method: Ten remitted MDD subjects and twelve healthy controls (HC) performed an emotion regulation task in 4 runs of rtfMRI-nf training followed by one transfer run without neurofeedback conducted in a single session. The functional connectivity between amygdala and prefrontal cortex was presented as a feedback bar concurrent with the emotion regulation task. Participants' emotional state was measured by the Positive and Negative Affect Schedule (PANAS) prior to and following the rtfMRI-nf. Psychological assessments were used to determine subjects' history of depression. Results: Participants with a history of MDD showed a trend of decreasing functional connectivity across the four rtfMRI-nf runs, and there was a marginally significant interaction between the MDD history and number of training runs. The HC group showed a significant increase of frontal cortex activation between the second and third neurofeedback runs. Comparing PANAS scores before and after connectivity-based rtfMRI-nf, we observed a significant decrease in negative PANAS score in the whole group overall, and a significant decrease in positive PANAS score in the MDD group alone.

q-bio.NC↗

arXiv2022

TENSILE: A Tensor granularity dynamic GPU memory scheduling method toward multiple dynamic workloads system

Recently, deep learning has been an area of intense research. However, as a kind of computing-intensive task, deep learning highly relies on the scale of GPU memory, which is usually prohibitive and scarce. Although some extensive works have been proposed for dynamic GPU memory management, they are hard to apply to systems with multiple dynamic workloads, such as in-database machine learning systems. In this paper, we demonstrated TENSILE, a method of managing GPU memory in tensor granularity to reduce the GPU memory peak, considering the multiple dynamic workloads. TENSILE tackled the cold-starting and across-iteration scheduling problem existing in previous works. We implemented TENSILE on a deep learning framework built by ourselves and evaluated its performance. The experiment results show that TENSILE can save more GPU memory with less extra overhead than prior works in single and multiple dynamic workloads scenarios.

cs.DC↗

arXiv2021

A parallel-network continuous quantitative trading model with GARCH and PPO

It is a difficult task for both professional investors and individual traders continuously making profit in stock market. With the development of computer science and deep reinforcement learning, Buy\&Hold (B\&H) has been oversteped by many artificial intelligence trading algorithms. However, the information and process are not enough, which limit the performance of reinforcement learning algorithms. Thus, we propose a parallel-network continuous quantitative trading model with GARCH and PPO to enrich the basical deep reinforcement learning model, where the deep learning parallel network layers deal with 3 different frequencies data (including GARCH information) and proximal policy optimization (PPO) algorithm interacts actions and rewards with stock trading environment. Experiments in 5 stocks from Chinese stock market show our method achieves more extra profit comparing with basical reinforcement learning methods and bench models.

q-fin.TR↗