Zilong Ji
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Zilong Ji has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Si Wu
Peking University
4 shared publications - Tiejun Huang
Peking University
4 shared publications Xiaolong Zou
2 shared publications- Xiaolong Zou
School of Psychological and Cognitive Sciences, IDG/McGovern Institute for Brain Research, Peking-Tsinghua Center for Life Sciences, Center of Quantitative Biology, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China; Institute of Artificial Intelligence, Hefei Comprehensive National Science Center, Hefei, China; Beijing Academy of Artificial Intelligence, Beijing, China
2 shared publications - Yuanyuan Mi
Department of Psychology, Tsinghua University, Beijing, China
2 shared publications - Aobo Shi
Key Lab for Cognitve Neurosience & Learning, Beijing Normal University
1 shared publication
Latest Publications
- Open Access
- 4 April 2022
- Published by Elsevier in Neural Networks
- Vol. 151, 349-364
Neural feedback facilitates rough-to-fine information retrieval
- 4 April 2022
- Published by Elsevier in Neural Networks
- Vol. 151, 349-364
Open AccessNeural feedback facilitates rough-to-fine information retrieval
- Open Access
- 1 November 2021
- Published by Elsevier in Neural Networks
- Vol. 143, 74-87
A brain-inspired computational model for spatio-temporal information processing
- 1 November 2021
- Published by Elsevier in Neural Networks
- Vol. 143, 74-87
Open AccessA brain-inspired computational model for spatio-temporal information processing
- Open Access
Vision at A Glance: Interplay between Fine and Coarse Information Processing Pathways
Object recognition is often viewed as a feedforward, bottom-up process in machine learning, but in real neural systems, object recognition is a complicated process which involves the interplay between two signal pathways. One is the parvocellular pathway (P-pathway), which is slow and extracts fine features of objects; the other is the magnocellular pathway (M-pathway), which is fast and extracts coarse features of objects. It has been suggested that the interplay between the two pathways endows the neural system with the capacity of processing visual information rapidly, adaptively, and robustly. However, the underlying computational mechanisms remain largely unknown. In this study, we build a computational model to elucidate the computational advantages associated with the interactions between two pathways. Our model consists of two convolution neural networks: one mimics the P-pathway, referred to as FineNet, which is deep, has small-size kernels, and receives detailed visual inputs; the other mimics the M-pathway, referred to as CoarseNet, which is shallow, has large-size kernels, and receives low-pass filtered or binarized visual inputs. The two pathways interact with each other via a Restricted Boltzmann Machine. We find that: 1) FineNet can teach CoarseNet through imitation and improve its performance considerably; 2) CoarseNet can improve the noise robustness of FineNet through association; 3) the output of CoarseNet can serve as a cognitive bias to improve the performance of FineNet. We hope that this study will provide insight into understanding visual information processing and inspire the development of new object recognition architectures.
Open AccessVision at A Glance: Interplay between Fine and Coarse Information Processing Pathways
Object recognition is often viewed as a feedforward, bottom-up process in machine learning, but in real neural systems, object recognition is a complicated process which involves the interplay between two signal pathways. One is the parvocellular pathway (P-pathway), which is slow and extracts fine features of objects; the other is the magnocellular pathway (M-pathway), which is fast and extracts coarse features of objects. It has been suggested that the interplay between the two pathways endows the neural system with the capacity of processing visual information rapidly, adaptively, and robustly. However, the underlying computational mechanisms remain largely unknown. In this study, we build a computational model to elucidate the computational advantages associated with the interactions between two pathways. Our model consists of two convolution neural networks: one mimics the P-pathway, referred to as FineNet, which is deep, has small-size kernels, and receives detailed visual inputs; the other mimics the M-pathway, referred to as CoarseNet, which is shallow, has large-size kernels, and receives low-pass filtered or binarized visual inputs. The two pathways interact with each other via a Restricted Boltzmann Machine. We find that: 1) FineNet can teach CoarseNet through imitation and improve its performance considerably; 2) CoarseNet can improve the noise robustness of FineNet through association; 3) the output of CoarseNet can serve as a cognitive bias to improve the performance of FineNet. We hope that this study will provide insight into understanding visual information processing and inspire the development of new object recognition architectures.
- Open Access
Unsupervised Few-shot Learning via Self-supervised Training
Learning from limited exemplars (few-shot learning) is a fundamental, unsolved problem that has been laboriously explored in the machine learning community. However, current few-shot learners are mostly supervised and rely heavily on a large amount of labeled examples. Unsupervised learning is a more natural procedure for cognitive mammals and has produced promising results in many machine learning tasks. In the current study, we develop a method to learn an unsupervised few-shot learner via self-supervised training (UFLST), which can effectively generalize to novel but related classes. The proposed model consists of two alternate processes, progressive clustering and episodic training. The former generates pseudo-labeled training examples for constructing episodic tasks; and the later trains the few-shot learner using the generated episodic tasks which further optimizes the feature representations of data. The two processes facilitate with each other, and eventually produce a high quality few-shot learner. Using the benchmark dataset Omniglot and Mini-ImageNet, we show that our model outperforms other unsupervised few-shot learning methods. Using the benchmark dataset Market1501, we further demonstrate the feasibility of our model to a real-world application on person re-identification.
Open AccessUnsupervised Few-shot Learning via Self-supervised Training
Learning from limited exemplars (few-shot learning) is a fundamental, unsolved problem that has been laboriously explored in the machine learning community. However, current few-shot learners are mostly supervised and rely heavily on a large amount of labeled examples. Unsupervised learning is a more natural procedure for cognitive mammals and has produced promising results in many machine learning tasks. In the current study, we develop a method to learn an unsupervised few-shot learner via self-supervised training (UFLST), which can effectively generalize to novel but related classes. The proposed model consists of two alternate processes, progressive clustering and episodic training. The former generates pseudo-labeled training examples for constructing episodic tasks; and the later trains the few-shot learner using the generated episodic tasks which further optimizes the feature representations of data. The two processes facilitate with each other, and eventually produce a high quality few-shot learner. Using the benchmark dataset Omniglot and Mini-ImageNet, we show that our model outperforms other unsupervised few-shot learning methods. Using the benchmark dataset Market1501, we further demonstrate the feasibility of our model to a real-world application on person re-identification.
- Open Access
Closing the loop: tracking and perturbing behaviour of individuals in a group in real-time
Quantitative description and selective perturbation of individual animals in a social group is prerequisite for understanding complex social behaviors. Tracking behavioral patterns of individuals in groups is an active research field, however, reliable software tools for long-term or real-time tracking are still scarce. We developed a new open-source platform, called xyTracker , for online tracking and recognition of individual animals in groups. Featuring a convenient M atlab -based interface and a fast multi-threading C++ core, we achieved an > 30× speed-up over a popular existing tracking method without loss in accuracy. Moreover, since memory usage is low, many hours of high-resolution video files can be tracked in reasonable time, making long-term observation of behavior possible. In a number of exemplary experiments on zebra fish, we show the feasibility of long-term observations and how to use the software to perform closed-loop experiments, where the tracked position of individuals is fed-back in real-time to a stimulus presentation screen installed below the fish-tank. Visual stimulation capabilities is incorporated into xyTracker and can be based on any behavioral features of all members of the group, such as, collective location, speed, or direction of movement, making interesting closed-loop experiments for investigating group behavior in a virtual reality setting possible.
Open AccessClosing the loop: tracking and perturbing behaviour of individuals in a group in real-time
Quantitative description and selective perturbation of individual animals in a social group is prerequisite for understanding complex social behaviors. Tracking behavioral patterns of individuals in groups is an active research field, however, reliable software tools for long-term or real-time tracking are still scarce. We developed a new open-source platform, called xyTracker , for online tracking and recognition of individual animals in groups. Featuring a convenient M atlab -based interface and a fast multi-threading C++ core, we achieved an > 30× speed-up over a popular existing tracking method without loss in accuracy. Moreover, since memory usage is low, many hours of high-resolution video files can be tracked in reasonable time, making long-term observation of behavior possible. In a number of exemplary experiments on zebra fish, we show the feasibility of long-term observations and how to use the software to perform closed-loop experiments, where the tracked position of individuals is fed-back in real-time to a stimulus presentation screen installed below the fish-tank. Visual stimulation capabilities is incorporated into xyTracker and can be based on any behavioral features of all members of the group, such as, collective location, speed, or direction of movement, making interesting closed-loop experiments for investigating group behavior in a virtual reality setting possible.