An adaptive deep Q-learning strategy for handwritten digit recognition

Qiao, Junfei; Wang, Gongming; Li, Wenjing; Chen, Min

doi:10.1016/j.neunet.2018.02.010

Cited by 78 publications

(36 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This has been mentioned in the literature. Qiao et al ( 2018 ) proposed an adaptive DQN strategy and applied it to text recognition. These results showed that the DQN algorithm is significantly better than other algorithms, which also indicated the advantages of the DQN algorithm in image recognition (Qiao et al, 2018 ).…”

Section: Discussionmentioning

confidence: 99%

“…Qiao et al ( 2018 ) proposed an adaptive DQN strategy and applied it to text recognition. These results showed that the DQN algorithm is significantly better than other algorithms, which also indicated the advantages of the DQN algorithm in image recognition (Qiao et al, 2018 ). Compared with the deep learning algorithm DQN, the DDQN algorithm is better than DQN in terms of value accuracy and strategy, which is also consistent with previous reports (Qu et al, 2020 ).…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

The Path Planning of Mobile Robot by Neural Networks and Hierarchical Reinforcement Learning

Liao

2020

Front. Neurorobot.

View full text Add to dashboard Cite

Existing mobile robots cannot complete some functions. To solve these problems, which include autonomous learning in path planning, the slow convergence of path planning, and planned paths that are not smooth, it is possible to utilize neural networks to enable to the robot to perceive the environment and perform feature extraction, which enables them to have a fitness of environment to state action function. By mapping the current state of these actions through Hierarchical Reinforcement Learning (HRL), the needs of mobile robots are met. It is possible to construct a path planning model for mobile robots based on neural networks and HRL. In this article, the proposed algorithm is compared with different algorithms in path planning. It underwent a performance evaluation to obtain an optimal learning algorithm system. The optimal algorithm system was tested in different environments and scenarios to obtain optimal learning conditions, thereby verifying the effectiveness of the proposed algorithm. Deep Deterministic Policy Gradient (DDPG), a path planning algorithm for mobile robots based on neural networks and hierarchical reinforcement learning, performed better in all aspects than other algorithms. Specifically, when compared with Double Deep Q-Learning (DDQN), DDPG has a shorter path planning time and a reduced number of path steps. When introducing an influence value, this algorithm shortens the convergence time by 91% compared with the Q-learning algorithm and improves the smoothness of the planned path by 79%. The algorithm has a good generalization effect in different scenarios. These results have significance for research on guiding, the precise positioning, and path planning of mobile robots.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

The Path Planning of Mobile Robot by Neural Networks and Hierarchical Reinforcement Learning

Liao

2020

Front. Neurorobot.

View full text Add to dashboard Cite

show abstract

“…Neglecting friction such as air resistance, all small bodies accelerate in a gravitational field at the same rate relative to the center of mass. [7][8][9] This is true regardless of the masses or compositions of the bodies. At different points on Earth, objects fall with an acceleration between 9.764 m/s 2 and 9.834 m/s 2 depending on altitude and latitude, with a conventional standard value of 9.80 m/s 2 (approximately 32.174 ft/s 2 ) [10].…”

Section: Introductionmentioning

confidence: 96%

Realistic Simulations of Non-Linear Acceleration of the Rocket in the Air

Xiong¹

2018

AJPA

View full text Add to dashboard Cite

The motion of the object in the medium has always been a hot research topic, and it is closely connected with many applications in our life. The acceleration of the object with multiple forces becomes very complicated, especially when these forces depend on the motion of the object. The exact formula for the object motion is a differential-integral equation and is very difficult to be solved analytically. One example of this kind of motions is the rocket launch. With sufficient thrust, the rocket can obtain an acceleration large enough to escape from the gravity of the earth. With the increasing height, the gravity from the earth becomes smaller, which affects the net acceleration of the rocket. Meanwhile, the air resistance becomes more and more important when the velocity of the rocket increases. It even plays the main role in the middle stage of the launch. Also, as the air resistance depends on both the velocity of the rocket and the air density (there is no air resistance in vacuum), the air resistance will decrease when the air density becomes small enough at the large height. In this article, a model that includes all of the factors mentioned above is established, and how these forces change the velocity of the rocket is analyzed. Two scenarios, one with air resistance and one without, are described. The velocity of the rocket in each scenario is represented by graphs, which are compared. With justification, the Taylor series is used to solve the differential-integral equation, and it is found that the fuel thrust and the gravity become important in the rocket launch at the beginning stage. In the middle stage, the air resistance begins to have a significant effect and reduces the acceleration of the rocket. In the final stage, there is virtually no gravity or air resistance, and only the fuel thrust contributes to the acceleration of the rocket.

show abstract

“…Recent literature on the subject of digit recognition proposes deep learning, reinforcement learning and graphbased learning approaches [12]- [16]. Using such techniques very high accuracies are obtained, but these techniques suffer from the issue that it is not possible to explain the reasoning behind such high accuracies or which extracted features are the best for a given problem.…”

Section: Introductionmentioning

confidence: 99%

High-Quality Wavelets Features Extraction for Handwritten Arabic Numerals Recognition

Akhtar

Qureshi

Alquhayz

2019

Int. J. Adv. Sci. Eng. Inf. Technol.

View full text Add to dashboard Cite

Arabic handwritten digit recognition is the science of recognition and classification of handwritten Arabic digits. It has been a subject of research for many years with rich literature available on the subject. Handwritten digits written by different people are not of the same size, thickness, style, position or orientation. Hence, many different challenges have to overcome for resolving the problem of handwritten digit recognition. The variation in the digits is due to the writing styles of different people which can differ significantly. Automatic handwritten digit recognition has wide application such as automatic processing of bank cheques, postal addresses, and tax forms. A typical handwritten digit recognition application consists of three main stages namely features extraction, features selection, and classification. One of the most important problems is feature extraction. In this paper, a novel feature extraction approach for off-line handwritten digit recognition is presented. Wavelets-based analysis of image data is carried out for feature extraction, and then classification is performed using various classifiers. To further reduce the size of training data-set, high entropy subbands are selected. To increase the recognition rate, individual subbands providing high classification accuracies are selected from the over-complete tree. The features extracted are also normalized to standardize the range of independent variables before providing them to the classifier. Classification is carried out using k-NN and SVMs. The results show that the quality of extracted features is high as almost equivalently high classification accuracies are acquired for both classifiers, i.e. k-NNs and SVMs.

show abstract

An adaptive deep Q-learning strategy for handwritten digit recognition

Cited by 78 publications

References 37 publications

The Path Planning of Mobile Robot by Neural Networks and Hierarchical Reinforcement Learning

The Path Planning of Mobile Robot by Neural Networks and Hierarchical Reinforcement Learning

Realistic Simulations of Non-Linear Acceleration of the Rocket in the Air

High-Quality Wavelets Features Extraction for Handwritten Arabic Numerals Recognition

Contact Info

Product

Resources

About