Adaptive Compression for Online Computer Vision: An Edge Reinforcement Learning Approach

2021 ◽  
Vol 17 (4) ◽  
pp. 1-23
Author(s):  
Zhaoliang He ◽  
Hongshan Li ◽  
Zhi Wang ◽  
Shutao Xia ◽  
Wenwu Zhu
Keyword(s):  
Computer Vision ◽  
Deep Learning ◽  
Reinforcement Learning ◽  
Quality Level ◽  
Image Size ◽  
Standard Quality ◽  
Encapsulation Method ◽  
Cloud Servers ◽  
Model Structures ◽  
Deep Learning Model

With the growth of computer vision-based applications, an explosive amount of images have been uploaded to cloud servers that host such online computer vision algorithms, usually in the form of deep learning models. JPEG has been used as the de facto compression and encapsulation method for images. However, standard JPEG configuration does not always perform well for compressing images that are to be processed by a deep learning model—for example, the standard quality level of JPEG leads to 50% of size overhead (compared with the best quality level selection) on ImageNet under the same inference accuracy in popular computer vision models (e.g., InceptionNet and ResNet). Knowing this, designing a better JPEG configuration for online computer vision-based services is still extremely challenging. First, cloud-based computer vision models are usually a black box to end-users; thus, it is challenging to design JPEG configuration without knowing their model structures. Second, the “optimal” JPEG configuration is not fixed; instead, it is determined by confounding factors, including the characteristics of the input images and the model, the expected accuracy and image size, and so forth. In this article, we propose a reinforcement learning (RL)-based adaptive JPEG configuration framework, AdaCompress. In particular, we design an edge (i.e., user-side) RL agent that learns the optimal compression quality level to achieve an expected inference accuracy and upload image size, only from the online inference results, without knowing details of the model structures. Furthermore, we design an explore-exploit mechanism to let the framework fast switch an agent when it detects a performance degradation, mainly due to the input change (e.g., images captured across daytime and night). Our evaluation experiments using real-world online computer vision-based APIs from Amazon Rekognition, Face++, and Baidu Vision show that our approach outperforms existing baselines by reducing the size of images by one-half to one-third while the overall classification accuracy only decreases slightly. Meanwhile, AdaCompress adaptively re-trains or re-loads the RL agent promptly to maintain the performance.

scholarly journals DrugEx v2: De Novo Design of Drug Molecule by Pareto-based Multi-Objective Reinforcement Learning in Polypharmacology

2021 ◽  
Author(s):  
Xuhan Liu ◽  
Kai Ye ◽  
Herman Van Vlijmen ◽  
Michael T. M. Emmerich ◽  
Adriaan P. IJzerman ◽  
...  
Keyword(s):  
Deep Learning ◽  
Reinforcement Learning ◽  
Drug Target ◽  
De Novo ◽  
Multi Objective ◽  
Learning Framework ◽  
Drug Molecules ◽  
Pareto Ranking ◽  
Speed Up ◽  
Deep Learning Model

<p>In polypharmacology, ideal drugs are required to bind to multiple specific targets to enhance efficacy or to reduce resistance formation. Although deep learning has achieved breakthrough in drug discovery, most of its applications only focus on a single drug target to generate drug-like active molecules in spite of the reality that drug molecules often interact with more than one target which can have desired (polypharmacology) or undesired (toxicity) effects. In a previous study we proposed a new method named <i>DrugEx</i> that integrates an exploration strategy into RNN-based reinforcement learning to improve the diversity of the generated molecules. Here, we extended our <i>DrugEx</i> algorithm with multi-objective optimization to generate drug molecules towards more than one specific target (two adenosine receptors, A<sub>1</sub>AR and A<sub>2A</sub>AR, and the potassium ion channel hERG in this study). In our model, we applied an RNN as the <i>agent</i> and machine learning predictors as the <i>environment</i>, both of which were pre-trained in advance and then interplayed under the reinforcement learning framework. The concept of evolutionary algorithms was merged into our method such that <i>crossover</i> and <i>mutation</i> operations were implemented by the same deep learning model as the <i>agent</i>. During the training loop, the agent generates a batch of SMILES-based molecules. Subsequently scores for all objectives provided by the <i>environment</i> are used for constructing Pareto ranks of the generated molecules with non-dominated sorting and Tanimoto-based crowding distance algorithms. Here, we adopted GPU acceleration to speed up the process of Pareto optimization. The final reward of each molecule is calculated based on the Pareto ranking with the ranking selection algorithm. The agent is trained under the guidance of the reward to make sure it can generate more desired molecules after convergence of the training process. All in all we demonstrate generation of compounds with a diverse predicted selectivity profile toward multiple targets, offering the potential of high efficacy and lower toxicity.</p>

scholarly journals Masked Face Analysis via Multi-Task Deep Learning

2021 ◽  
Vol 7 (10) ◽  
pp. 204
Author(s):  
Vatsa S. Patel ◽  
Zhongliang Nie ◽  
Trung-Nghia Le ◽  
Tam V. Nguyen
Keyword(s):  
Computer Vision ◽  
Deep Learning ◽  
Face Recognition ◽  
Face Mask ◽  
Learning Model ◽  
Unified Framework ◽  
Improve Performance ◽  
Face Analysis ◽  
And Gender ◽  
Deep Learning Model

Face recognition with wearable items has been a challenging task in computer vision and involves the problem of identifying humans wearing a face mask. Masked face analysis via multi-task learning could effectively improve performance in many fields of face analysis. In this paper, we propose a unified framework for predicting the age, gender, and emotions of people wearing face masks. We first construct FGNET-MASK, a masked face dataset for the problem. Then, we propose a multi-task deep learning model to tackle the problem. In particular, the multi-task deep learning model takes the data as inputs and shares their weight to yield predictions of age, expression, and gender for the masked face. Through extensive experiments, the proposed framework has been found to provide a better performance than other existing methods.

scholarly journals Multimodal Multi-tasking for Skin Lesion Classification Using Deep Neural Networks

2021 ◽  
pp. 27-38
Author(s):  
Rafaela Carvalho ◽  
João Pedrosa ◽  
Tudor Nedelcu
Keyword(s):  
Neural Networks ◽  
Computer Vision ◽  
Deep Learning ◽  
Skin Lesion ◽  
Deep Neural Networks ◽  
Classification Performance ◽  
Additional Information ◽  
Deep Learning Model ◽  
Lesion Classification

AbstractSkin cancer is one of the most common types of cancer and, with its increasing incidence, accurate early diagnosis is crucial to improve prognosis of patients. In the process of visual inspection, dermatologists follow specific dermoscopic algorithms and identify important features to provide a diagnosis. This process can be automated as such characteristics can be extracted by computer vision techniques. Although deep neural networks can extract useful features from digital images for skin lesion classification, performance can be improved by providing additional information. The extracted pseudo-features can be used as input (multimodal) or output (multi-tasking) to train a robust deep learning model. This work investigates the multimodal and multi-tasking techniques for more efficient training, given the single optimization of several related tasks in the latter, and generation of better diagnosis predictions. Additionally, the role of lesion segmentation is also studied. Results show that multi-tasking improves learning of beneficial features which lead to better predictions, and pseudo-features inspired by the ABCD rule provide readily available helpful information about the skin lesion.

scholarly journals How to Generate Image Dataset based on 3D Model and Deep Learning Method

2018 ◽  
Vol 7 (3.34) ◽  
pp. 221
Author(s):  
Sooyoung Cho ◽  
Sang Geun Choi ◽  
Daeyeol Kim ◽  
Gyunghak Lee ◽  
Chae BongSohn
Keyword(s):  
Computer Vision ◽  
Deep Learning ◽  
Object Recognition ◽  
Object Tracking ◽  
3D Model ◽  
Object Segmentation ◽  
Learning Method ◽  
Data Set ◽  
Image Dataset ◽  
Deep Learning Model

Performances of computer vision tasks have been drastically improved after applying deep learning. Such object recognition, object segmentation, object tracking, and others have been approached to the super-human level. Most of the algorithms were trained by using supervised learning. In general, the performance of computer vision is improved by increasing the size of the data. The collected data was labeled and used as a data set of the YOLO algorithm. In this paper, we propose a data set generation method using Unity which is one of the 3D engines. The proposed method makes it easy to obtain the data necessary for learning. We classify 2D polymorphic objects and test them against various data using a deep learning model. In the classification using CNN and VGG-16, 90% accuracy was achieved. And we used Tiny-YOLO of YOLO algorithm for object recognition and we achieved 78% accuracy. Finally, we compared in terms of virtual and real environments it showed a result of 97 to 99 percent for each accuracy.

scholarly journals DrugEx v2: De Novo Design of Drug Molecule by Pareto-based Multi-Objective Reinforcement Learning in Polypharmacology

2021 ◽  
Author(s):  
Xuhan Liu ◽  
Kai Ye ◽  
Herman Van Vlijmen ◽  
Michael T. M. Emmerich ◽  
Adriaan P. IJzerman ◽  
...  
Keyword(s):  
Deep Learning ◽  
Reinforcement Learning ◽  
Drug Target ◽  
De Novo ◽  
Multi Objective ◽  
Learning Framework ◽  
Drug Molecules ◽  
Pareto Ranking ◽  
Speed Up ◽  
Deep Learning Model

<p>In polypharmacology, ideal drugs are required to bind to multiple specific targets to enhance efficacy or to reduce resistance formation. Although deep learning has achieved breakthrough in drug discovery, most of its applications only focus on a single drug target to generate drug-like active molecules in spite of the reality that drug molecules often interact with more than one target which can have desired (polypharmacology) or undesired (toxicity) effects. In a previous study we proposed a new method named <i>DrugEx</i> that integrates an exploration strategy into RNN-based reinforcement learning to improve the diversity of the generated molecules. Here, we extended our <i>DrugEx</i> algorithm with multi-objective optimization to generate drug molecules towards more than one specific target (two adenosine receptors, A<sub>1</sub>AR and A<sub>2A</sub>AR, and the potassium ion channel hERG in this study). In our model, we applied an RNN as the <i>agent</i> and machine learning predictors as the <i>environment</i>, both of which were pre-trained in advance and then interplayed under the reinforcement learning framework. The concept of evolutionary algorithms was merged into our method such that <i>crossover</i> and <i>mutation</i> operations were implemented by the same deep learning model as the <i>agent</i>. During the training loop, the agent generates a batch of SMILES-based molecules. Subsequently scores for all objectives provided by the <i>environment</i> are used for constructing Pareto ranks of the generated molecules with non-dominated sorting and Tanimoto-based crowding distance algorithms. Here, we adopted GPU acceleration to speed up the process of Pareto optimization. The final reward of each molecule is calculated based on the Pareto ranking with the ranking selection algorithm. The agent is trained under the guidance of the reward to make sure it can generate more desired molecules after convergence of the training process. All in all we demonstrate generation of compounds with a diverse predicted selectivity profile toward multiple targets, offering the potential of high efficacy and lower toxicity.</p>

scholarly journals DrugEx v2: De Novo Design of Drug Molecule by Pareto-based Multi-Objective Reinforcement Learning in Polypharmacology

2021 ◽  
Author(s):  
Xuhan Liu ◽  
Kai Ye ◽  
Herman Van Vlijmen ◽  
Michael T. M. Emmerich ◽  
Adriaan P. IJzerman ◽  
...  
Keyword(s):  
Deep Learning ◽  
Reinforcement Learning ◽  
Drug Target ◽  
De Novo ◽  
Multi Objective ◽  
Learning Framework ◽  
Drug Molecules ◽  
Pareto Ranking ◽  
Speed Up ◽  
Deep Learning Model

<p>In polypharmacology, ideal drugs are required to bind to multiple specific targets to enhance efficacy or to reduce resistance formation. Although deep learning has achieved breakthrough in drug discovery, most of its applications only focus on a single drug target to generate drug-like active molecules in spite of the reality that drug molecules often interact with more than one target which can have desired (polypharmacology) or undesired (toxicity) effects. In a previous study we proposed a new method named <i>DrugEx</i> that integrates an exploration strategy into RNN-based reinforcement learning to improve the diversity of the generated molecules. Here, we extended our <i>DrugEx</i> algorithm with multi-objective optimization to generate drug molecules towards more than one specific target (two adenosine receptors, A<sub>1</sub>AR and A<sub>2A</sub>AR, and the potassium ion channel hERG in this study). In our model, we applied an RNN as the <i>agent</i> and machine learning predictors as the <i>environment</i>, both of which were pre-trained in advance and then interplayed under the reinforcement learning framework. The concept of evolutionary algorithms was merged into our method such that <i>crossover</i> and <i>mutation</i> operations were implemented by the same deep learning model as the <i>agent</i>. During the training loop, the agent generates a batch of SMILES-based molecules. Subsequently scores for all objectives provided by the <i>environment</i> are used for constructing Pareto ranks of the generated molecules with non-dominated sorting and Tanimoto-based crowding distance algorithms. Here, we adopted GPU acceleration to speed up the process of Pareto optimization. The final reward of each molecule is calculated based on the Pareto ranking with the ranking selection algorithm. The agent is trained under the guidance of the reward to make sure it can generate more desired molecules after convergence of the training process. All in all we demonstrate generation of compounds with a diverse predicted selectivity profile toward multiple targets, offering the potential of high efficacy and lower toxicity.</p>

scholarly journals Offloading and Transmission Strategies for IoT Edge Devices and Networks

Sensors ◽  
2019 ◽  
Vol 19 (4) ◽  
pp. 835 ◽  
Author(s):  
Jiheon Kang ◽  
Doo-Seop Eom
Keyword(s):  
Deep Learning ◽  
Reinforcement Learning ◽  
Power Consumption ◽  
Mac Protocol ◽  
Learning Model ◽  
Channel Access ◽  
Learning Method ◽  
Resource Constrained ◽  
Raw Data ◽  
Deep Learning Model

We present a machine and deep learning method to offload trained deep learning model and transmit packets efficiently on resource-constrained internet of things (IoT) edge devices and networks. Recently, the types of IoT devices have become diverse and the volume of data has been increasing, such as images, voice, and time-series sensory signals generated by various devices. However, transmitting large amounts of data to a server or cloud becomes expensive owing to limited bandwidth, and leads to latency for time-sensitive operations. Therefore, we propose a novel offloading and transmission policy considering energy-efficiency, execution time, and the number of generated packets for resource-constrained IoT edge devices that run a deep learning model and a reinforcement learning method to find an optimal contention window size for effective channel access using a contention-based medium access control (MAC) protocol. A Reinforcement learning is used to improve the performance of the applied MAC protocol. Our proposed method determines the offload and transmission strategies that are better to directly send fragmented packets of raw data or to send the extracted feature vector or the final output of deep learning networks, considering the operation performance and power consumption of the resource-constrained microprocessor, as well as the power consumption of the radio transceiver and latency for transmitting the all the generated packets. In the performance evaluation, we measured the performance parameters of ARM Cortex-M4 and Cortex-M7 processors for the network simulation. The evaluation results show that our proposed adaptive channel access and learning-based offload and transmission methods outperform conventional role-based channel access schemes. They transmit packets of raw data and are effective for IoT edge devices and network protocols.

scholarly journals A deep learning model for gas storage optimization

2021 ◽  
Author(s):  
Nicolas Curin ◽  
Michael Kettler ◽  
Xi Kleisinger-Yu ◽  
Vlatka Komaric ◽  
Thomas Krabichler ◽  
...  
Keyword(s):  
Risk Management ◽  
Monte Carlo ◽  
Deep Learning ◽  
Reinforcement Learning ◽  
State Of The Art ◽  
Gas Storage ◽  
High Dimensional ◽  
Forward Market ◽  
Storage Optimization ◽  
Deep Learning Model

AbstractTo the best of our knowledge, the application of deep learning in the field of quantitative risk management is still a relatively recent phenomenon. In this article, we utilize techniques inspired by reinforcement learning in order to optimize the operation plans of underground natural gas storage facilities. We provide a theoretical framework and assess the performance of the proposed method numerically in comparison to a state-of-the-art least-squares Monte-Carlo approach. Due to the inherent intricacy originating from the high-dimensional forward market as well as the numerous constraints and frictions, the optimization exercise can hardly be tackled by means of traditional techniques.

Sign in / Sign up

Export Citation Format

Share Document