0% Complete
صفحه اصلی
/
سی و یکمین کنفرانس بین المللی مهندسی برق
Connective Reconstruction-based Novelty Detection
نویسندگان :
Seyyed Morteza Hashemi
1
Parvaneh Aliniya
2
Parvin Razzaghi
3
1- دانشگاه تحصیلات تکمیلی علوم پایه زنجان
2- University of Nevada, Reno
3- دانشگاه تحصیلات تکمیلی علوم پایه زنجان
کلمات کلیدی :
Novelty class detection،Out-of-distribution detection،Computer vision،Deep learning
چکیده :
Detection of out-of-distribution samples is one of the critical tasks for real-world applications of computer vision. The advancement of deep learning has enabled us to analyze real-world data which contain unexplained samples, accentuating the need to detect out-of-distribution instances more than before. GAN-based approaches have been widely used to address this problem due to their ability to perform distribution fitting; however, they are accompanied by training instability and mode collapse. We propose a simple yet efficient reconstruction-based method that avoids adding complexities to compensate for the limitations of GAN models while outperforming them. Unlike previous reconstruction-based works that only utilize reconstruction error or generated samples, our proposed method simultaneously incorporates both of them in the detection task. Our model, which we call "Connective Novelty Detection" has two subnetworks, an autoencoder, and a binary classifier. The autoencoder learns the representation of the positive class by reconstructing them. Then, the model creates negative and connected positive examples using real and generated samples. Negative instances are generated via manipulating the real data, so their distribution is close to the positive class to achieve a more accurate boundary for the classifier. To boost the robustness of the detection to reconstruction error, connected positive samples are created by combining the real and generated samples. Finally, the binary classifier is trained using connected positive and negative examples. We demonstrate a considerable improvement in novelty detection over state-of-the-art methods on MNIST and Caltech-256 datasets.
لیست مقالات
لیست مقالات بایگانی شده
Joint Energy and Throughput Optimization in Energy Harvesting Cognitive Sensor Networks
Morteza Sharifi - Mahmood Mohassel Feghhi
Efficiency Enhancement of Heterojunction IBC Solar Cell: Surface Passivation
Amirmohammad Shahryari - Zohreh Golshan bafghi - Negin Manavizadeh
Mapping Human Grasping to 3-Finger Grippers: A Deep Learning Perspective
Fatemeh Naeinian - Elnaz Balazadeh - Mehdi Tale Masouleh
Defense Against Spectrum Sensing Data Falsification Attack in Cognitive Radio Networks Using Machine Learning
Nazanin Parhizgar - Ali Jamshidi - Peyman Setoodeh
A Brief Review on DC-Link Control Strategies in Microgrids
Mehran Seydi - Hassan Moradi CheshmehBeigi - Mohammad Hossein Mousavi
Power Transformer Vibration Study and its Application in Winding Deformation Detection
Amir Esmaeili Nezhad - Mohammad Hamed Samimi
Safe Cooperative Control of Non-Holonomic Mobile Manipulators
Zahra Kashi - Nargess Sadeghzadeh-Nokhodberiz
Performance Analysis of the Modified Flux-Coupling-Type SFCL in VSC-HVDC System
Mohammad Khakroei - Ashkan Mirzaei Rajeooni - Mahdi Rahimi Pirbasti - Hossein Heydari
A novel approach for recommender systems based on query likelihood and sentiment analysis
Mohammadreza Soltaninezhad - Alireza Basiri
Single-Channel Recursive Speech Separation with Unknown Speaker Count by Mask Estimation
Hadi Alizadeh - Rahil Mahdian Toroghi - Hassan Zareian
بیشتر
ثمین همایش، سامانه مدیریت کنفرانس ها و جشنواره ها - نگارش 41.7.4