Energy and Performance Trade-Off Optimization in Heterogeneous Computing via Reinforcement Learning

Yu, Zheqi; Machado, Pedro; Zahid, Adnan; Abdulghani, Amir M.; Dashtipour, Kia; Heidari, Hadi

doi:10.3390/electronics9111812

Cited by 24 publications

(12 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Unfortunately, it is difficult to detect a silent AF with such a short duration even using a 12-lead electrocardiogram because it requires longer monitoring intervals of the patient wearing devices, which can be troublesome (Jan and Koo, 2018). The diagnosis of patients with silent AF can be accelerated by improving the screening process in the detection of such unnoticed AF (Preethi and Sathiyakumari, 2016;Yu et al, 2020). Here, the machine learning approaches can be used to speed up the decision-making process while improving the reliability, efficiency, and accuracy of AF diagnosis.…”

Section: Machine Learning Methods For Atrial Fibrillationmentioning

confidence: 99%

A Review and Comparison of the State-of-the-Art Techniques for Atrial Fibrillation Detection and Skin Hydration

et al. 2021

Self Cite

View full text Add to dashboard Cite

Atrial fibrillation (AF) is one of the most common types of cardiac arrhythmia, with a prevalence of 1–2% in the community, increasing the risk of stroke and myocardial infarction. Early detection of AF, typically causing an irregular and abnormally fast heart rate, can help reduce the risk of strokes that are more common among older people. Intelligent models capable of automatic detection of AF in its earliest possible stages can improve the early diagnosis and treatment. Luckily, this can be made possible with the information about the heart's rhythm and electrical activity provided through electrocardiogram (ECG) and the decision-making machine learning-based autonomous models. In addition, AF has a direct impact on the skin hydration level and, hence, can be used as a measure for detection. In this paper, we present an independent review along with a comparative analysis of the state-of-the-art techniques proposed for AF detection using ECG and skin hydration levels. This paper also highlights the effects of AF on skin hydration level that is missing in most of the previous studies.

show abstract

Section: Machine Learning Methods For Atrial Fibrillationmentioning

confidence: 99%

A Review and Comparison of the State-of-the-Art Techniques for Atrial Fibrillation Detection and Skin Hydration

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…To find falsifying inputs for CPS, states, actions and rewards were defined based on inputs, outputs and function of past-dependent output signal, respectively, using Double DQN and A3C algorithms [21]. Authors proposed PMU-RL method [22] to balance energy power consumption and efficiency which showed promising results for heterogeneous computing platforms. We summarized papers [23][24][25][26] as that states were defined according to the systems status, agents took certain actions and rewards were fed back to the agents.…”

Section: Related Researchesmentioning

confidence: 99%

Application of Improved Asynchronous Advantage Actor Critic Reinforcement Learning Model on Anomaly Detection

Zhou

Wang

et al. 2021

Entropy

View full text Add to dashboard Cite

Anomaly detection research was conducted traditionally using mathematical and statistical methods. This topic has been widely applied in many fields. Recently reinforcement learning has achieved exceptional successes in many areas such as the AlphaGo chess playing and video gaming etc. However, there were scarce researches applying reinforcement learning to the field of anomaly detection. This paper therefore aimed at proposing an adaptable asynchronous advantage actor-critic model of reinforcement learning to this field. The performances were evaluated and compared among classical machine learning and the generative adversarial model with variants. Basic principles of the related models were introduced firstly. Then problem definitions, modelling processes and testing were detailed. The proposed model differentiated the sequence and image from other anomalies by proposing appropriate neural networks of attention mechanism and convolutional network for the two kinds of anomalies, respectively. Finally, performances with classical models using public benchmark datasets (NSL-KDD, AWID and CICIDS-2017, DoHBrw-2020) were evaluated and compared. Experiments confirmed the effectiveness of the proposed model with the results indicating higher rewards and lower loss rates on the datasets during training and testing. The metrics of precision, recall rate and F1 score were higher than or at least comparable to the state-of-the-art models. We concluded the proposed model could outperform or at least achieve comparable results with the existing anomaly detection models.

show abstract

“…To record the new information into the memory device, a spin-polarized current is applied to selected tracks (yellow arrows in Figure 7c). From the experimental point-of-view, this could be performed following a nano-identation procedure described elsewhere [49,50] optimized through different automatic compilation frameworks [51] or through reinforced learning [52]. Although existing characterization techniques do not yet allow a fast access of individual tracks in a vertical array, this type of equipment is constantly being updated with innovative designs and functionalities, so these measurements may become feasible in the near future.…”

Section: Arrays Of Nanowires As 3d Racetrack Memoriesmentioning

confidence: 99%

A Novel Design of a 3D Racetrack Memory Based on Functional Segments in Cylindrical Nanowire Arrays

Rial

Proença

2020

Nanomaterials

View full text Add to dashboard Cite

A racetrack memory is a device where the information is stored as magnetic domains (bits) along a nanowire (track). To read and record the information, the bits are moved along the track by current pulses until they reach the reading/writing heads. In particular, 3D racetrack memory devices use arrays of vertically aligned wires (tracks), thus enhancing storage density. In this work, we propose a novel 3D racetrack memory configuration based on functional segments inside cylindrical nanowire arrays. The innovative idea is the integration of the writing element inside the racetrack itself, avoiding the need to implement external writing heads next to the track. The use of selective magnetic segments inside one nanowire allows the creation of writing and storage sections inside the same track, separated by chemical constraints identical to those separating the bits. Using micromagnetic simulations, our study reveals that if the writing section is composed of two segments with different coercivities, one can reverse its magnetization independently from the rest of the memory device by applying an external magnetic field. Spin-polarized current pulses then move the information bits along selected tracks, completing the writing process by pushing the new bit into the storage section of the wire. Finally, we have proven the efficacy of this system inside an array of 7 nanowires, opening the possibility to use this configuration in a 3D racetrack memory device composed of an array of thousands of nanowires produced by low-cost and high-yield template-electrodeposition methods.

show abstract

Energy and Performance Trade-Off Optimization in Heterogeneous Computing via Reinforcement Learning

Cited by 24 publications

References 52 publications

A Review and Comparison of the State-of-the-Art Techniques for Atrial Fibrillation Detection and Skin Hydration

A Review and Comparison of the State-of-the-Art Techniques for Atrial Fibrillation Detection and Skin Hydration

Application of Improved Asynchronous Advantage Actor Critic Reinforcement Learning Model on Anomaly Detection

A Novel Design of a 3D Racetrack Memory Based on Functional Segments in Cylindrical Nanowire Arrays

Contact Info

Product

Resources

About