Convergence of Stochastic Gradient Descent in Deep Neural Network

Zhou, Baicun; Han, Congying; Guo, Tong

doi:10.1007/s10255-021-0991-2

Cited by 17 publications

(3 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…where: f : ℝ m → ℝ is the loss function, fi, for i ∈ {1, …N}, denotes the contribution to the loss function from data point i, N denotes the total number of data points [8]. The optimization algorithm implemented for the purpose of ANN training was stochastic gradient descent [42].…”

Section: Scenario Faults Sc1mentioning

confidence: 99%

Feasibility study of a rail vehicle damper fault detection by artificial neural Indexed by: networks

Melnik¹,

Koziak²,

Dižo³

et al. 2023

Eksploatacja i Niezawodność – Maintenance and Reliability

View full text Add to dashboard Cite

The aim of the study was to investigate rail vehicle dynamics under primary suspension dampers faults and explore possibility of its detection by means of artificial neural networks. For these purposes two types of analysis were carried out: preliminary analysis of 1 DOF rail vehicle model and a second one - a passenger coach benchmark model was tested in multibody simulation software - MSC.Adams with use of VI-Rail package. Acceleration signals obtained from the latter analysis served as an input data into the artificial neural network (ANN). ANNs of different number of hidden layers were capable of detecting faults for the trained suspension fault cases, however, achieved accuracy was below 63% at the best. These results can be considered satisfactory considering the complexity of dynamic phenomena occurring in the vibration system of a rail vehicle.

show abstract

Section: Scenario Faults Sc1mentioning

confidence: 99%

Feasibility study of a rail vehicle damper fault detection by artificial neural Indexed by: networks

Melnik¹,

Koziak²,

Dižo³

et al. 2023

Eksploatacja i Niezawodność – Maintenance and Reliability

View full text Add to dashboard Cite

show abstract

“…In Algorithm 1, we describe the pseudocode of our proposed DRL algorithm. We use the Stochastic Gradient Descent (SGD) algorithm [51] to perform Deep-Q-Network (DQN) agent training. Then, we tune the main hyperparameter to decide about optimal DNN configurations such as epoch/iteration numbers, optimizer parameters, and action selection strategies.…”

Section: B Dveap: the Proposed Dqn-drlmentioning

confidence: 99%

Next-Generation Edge Computing Assisted Autonomous Driving Based Artificial Intelligence Algorithms

et al. 2022

View full text Add to dashboard Cite

Edge Computing and Network Function Virtualization (NFV) concepts can improve network processing and multi-resources allocation when intelligent optimization algorithms are deployed. Multiservice offloading and allocation approaches pose interesting challenges in the current and nextgeneration vehicle networks. The state-of-the-art optimization approaches still formulate exact algorithms, and tune approximation methods to get sufficient solutions. These approaches are data-centric that aim to use heterogeneous data inputs to find the near optimal solutions. In the context of connected and autonomous vehicles (CAVs), these techniques show an exponential computational time and deal only with small and medium scale networks. Therefore, we are motivated by using recent Deep Reinforcement Learning (DRL) techniques to learn the behavior of exact optimization algorithms while enhancing the Quality of Service (QoS) of network operators and satisfying the requirements of the next-generation Autonomous Vehicles (AVs). DRL algorithms can improve AVs service offloading and optimize edge resources. An Optimal Virtual Edge Autopilot Placement (OVEAP) algorithm is proposed using Integer Linear Programming (ILP). Moreover, an autopilot placement protocol is presented to support the algorithm. Optimal allocation and Virtual Network Function (VNF) placement and chaining of the autopilot, based on several new constraints such as computing and networking loads, network edge infrastructure, and placement cost, are designed. Further, a DRL approach is formulated to deal with dense Internet of Autonomous Vehicle (IoAV) networks. Extensive simulations and evaluations are carried out. Results show that the proposed allocation strategies outperform the state-of-the-art solutions and give better performance in terms of Total Edge Servers Utilization, Total Edge Servers Allocation Time, and Successfully Allocated autopilots.

show abstract

“…Ilustrasi penurunan gradien dapat dilihat pada Gambar 4. Stochastic gradient descent (SGD) akan digunakan untuk melatih CNN, dimana kelebihan dari SGD dibanding penurunan gradien total yaitu konvergen lebih cepat dan mampu untuk keluar dari local minimum [24].…”

Section: B Arsitektur Cnnunclassified

Multi-Scale Convolutional Networks untuk Pengenalan Rambu Lalu Lintas di Indonesia

Akbar¹,

Purnomo²,

Supatman³

2022

SISFOKOM

View full text Add to dashboard Cite

Teknologi pengenalan rambu lalu lintas yang sering disebut dengan traffic sign recognition (TSR), digunakan untuk mengenali rambu lalu lintas melalui pemanfaatan pengolahan citra. TSR sendiri dapat diaplikasikan pada sistem pembantu pengemudi, sistem pembantu pengemudi tingkat lanjut, sistem mengemudi otonom, keamanan jalan raya, pemahaman suasana perkotaan, dan pemantauan rambu untuk kepentingan perawatan. Perbaruan dari pengenalan rambu lalu lintas di Indonesia menggunakan multi-scale convolutional neural network (CNN) telah disajikan pada artikel ini. Dataset yang digunakan pada penelitian ini berjumlah 2050 data citra rambu lalu lintas yang dikelompokkan kedalam 10 kelas. Model CNN terdiri dari tiga lapisan konvolusi berukuran 3x3, tiga lapisan penggabungan (Maxpool) berukuran 2x2 dan satu lapisan fully-connected yang memanfaatkan fungsi aktivasi Softmax. Jumlah filter yang digunakan pada setiap lapisan konvolusi adalah 32. Algoritma pelatihan yang digunakan yaitu Stochastic gradient descent (SGD). Dengan menggunakan 1750 data citra latih, nilai epoch 20, dan laju pelatihan 0,005, nilai galat dan nilai akurasi yang didapatkan pada tahap pelatihan adalah masing-masing 0,0026 dan 100%. Sedangkan pada tahap pengujian, dengan 300 data citra uji, model CNN mampu memperoleh nilai galat 0,017 dan nilai akurasi mencapai 99,67%.

show abstract

Convergence of Stochastic Gradient Descent in Deep Neural Network

Cited by 17 publications

References 10 publications

Feasibility study of a rail vehicle damper fault detection by artificial neural Indexed by: networks

Feasibility study of a rail vehicle damper fault detection by artificial neural Indexed by: networks

Next-Generation Edge Computing Assisted Autonomous Driving Based Artificial Intelligence Algorithms

Multi-Scale Convolutional Networks untuk Pengenalan Rambu Lalu Lintas di Indonesia

Contact Info

Product

Resources

About