Attended the Exciting International Joint Conference on Neural Networks (IJCNN) 2021
Introduction:
In this post, I will share my experience attending the International Joint Conference on Neural Networks (IJCNN) 2021. Organized by the International Neural Network Society, this conference is well-known and popular in the field of neural networks. It covers a wide range of research areas including bioinformatics, data mining, deep learning, sensor networks, machine learning, and neuroengineering.
Due to the pandemic, the conference was held online on a virtual platform. Participants were provided with a downloadable link to access the complete proceedings. Our paper, titled “Fruit classification using deep feature maps in the presence of deceptive similar classes,” was accepted for publication and I had the opportunity to present it.
During the conference, I attended various plenary talks and paper presentations. I found the plenary talk by Prof. Marios M. Polycarpou on Smart Interactive Buildings particularly interesting, as it covered distributed and smart HVAC systems and other monitoring systems. Another notable talk was by Prof. Karl J. Friston on active interfaces and decision-making processes.
The second day of the conference featured paper presentations in different tracks, including Bayesian Neural Networks, AutoML applications, and Machine Learning in Computer Vision. I also had the chance to present our paper again and interact with other participants and session chairs.
On the third day, I attended a fascinating plenary talk by Prof. Peter Tino on Recurrent Networks, where he delved into the learning process from dynamic data. Prof. Zongben Xu’s talk on the presuppositions of machine learning provided insights into different approaches in the field.
The fourth day included more paper presentations and an engaging keynote talk by Prof. Dietmar Plenz on Self-Organized Criticality in the Brain. This talk explored the evolution capability of complex systems and the benefits of scale-invariant events.
The fifth day started with a keynote talk by Prof. Nikola Kasabov on Transfer Learning and Knowledge Transfer Between Humans and Machines. He discussed brain-inspired spiking neural network architectures and their applications in adaptable and explainable AI.
The final day of the conference featured tutorial sessions on various topics such as Accelerating Deep Learning Computation and Deep Learning for Graphs. These sessions provided valuable insights and were followed by paper presentations, workshops, and competitions.
Overall, the conference was well-managed, and I enjoyed the opportunity to learn from renowned speakers, present our paper, and interact with fellow researchers. I look forward to participating in future editions of this conference.
Full Article: Attended the Exciting International Joint Conference on Neural Networks (IJCNN) 2021
International Joint Conference on Neural Networks (IJCNN) 2021: A Comprehensive Overview
The International Joint Conference on Neural Networks (IJCNN) 2021, organized by the International Neural Network Society, is a prestigious conference that focuses on neural networks theory, analysis, and applications across various research areas. This year, the conference was held online on a virtual platform, allowing participants from all over the world to attend and present their papers.
Accepted Paper: “Fruit classification using deep feature maps in the presence of deceptive similar classes”
One of the highlights of the conference was the acceptance of a paper titled “Fruit classification using deep feature maps in the presence of deceptive similar classes” for publication. The authors, Mohit Dandekar, Narinder Singh Punn, Sanjay Kumar Sonbharda, and Sonali Agarwal, presented their findings during the conference. The paper explores the use of deep feature maps to accurately classify fruits, even when faced with deceptive similar classes. It provides valuable insights into the application of deep learning in the field of fruit classification.
Day 1: Welcome Talks and Paper Presentations
The conference kicked off with welcome talks and plenary sessions on various tracks, including Adversarial Machine Learning and Cyber Security, Artificial Intelligence and Security, and Healthcare Analytics. Participants had the opportunity to attend paper presentations in parallel sessions, where pre-recorded videos followed by Q&A sessions were shown. Notable plenary talks included Prof. Marios M. Polycarpou’s presentation on Smart Interactive Buildings, which focused on distributed and smart HVAC systems and other monitoring systems. Another interesting talk by Prof. Karl J. Friston discussed active interfaces and decision-making processes. The only area for improvement noted was the limited participation of attendees in asking questions.
Day 2: Keynote Talks and Paper Presentations
Day 2 began with paper presentations across various tracks, covering topics such as Bayesian Neural Networks, AutoML applications, Computer Vision, and Learning from Imbalanced and Difficult Data. Prof. Riitta Salmelin delivered a keynote talk on the advancements in neuroimaging and its potential for individual-level predictions. Later, the authors of the accepted paper on fruit classification had the opportunity to present their findings. The day was filled with insightful interactions between participants and session chairs.
Day 3: Plenary Talks on Recurrent Networks and Machine Learning
Day 3 started with a plenary talk by Prof. Peter Tino on the world of Recurrent Networks. He discussed the impact of temporal structures in dynamic data and provided deep insights into the working mechanism of recurrent neural networks. Later, Prof. Zongben Xu delivered a talk on presuppositions of Machine Learning, presenting different hypothesis approaches. The day also featured paper and poster presentations on machine learning and deep learning, data analytics, and computation intelligence.
Day 4: Paper Presentations and Keynote Talk on Self-Organized Criticality
Day 4 focused on paper presentations across various tracks, including bio-inspired systems, deep learning, ensemble modeling, and more. Notable presentations included “PANDA: Perceptually Aware Neural Detection of Anomalies” and “A Meta-Learning Approach for Automated Hyperparameter Tuning in Evolving Data Streams.” Prof. Dietmar Plenz delivered a keynote talk on Self-Organized Criticality in the Brain, highlighting its role in system performance. The day concluded with presentations on neural networks applications, neuroengineering, and perception.
Day 5: Keynote Talk on Transfer Learning and Knowledge Transfer
The day started with an intriguing keynote talk by Prof. Nikola Kasabov on Transfer Learning and Knowledge Transfer between humans and machines using brain-inspired spiking neural networks. Prof. Kasabov discussed the use of transfer learning and self-organizing learning principles for adaptable and explainable AI. The day continued with paper presentations on various tracks, showcasing the latest advancements in different research areas.
Final Day: Tutorial Sessions and Closing Remarks
On the final day of the conference, participants had the opportunity to attend tutorial sessions on topics such as accelerating deep learning computation, deep learning for graphs, and machine learning for brain-computer interfaces. These sessions were followed by paper presentations, workshops, and competitions. The conference concluded with closing remarks, leaving participants excited for future editions of the IJCNN.
In conclusion, the International Joint Conference on Neural Networks (IJCNN) 2021 provided an excellent platform for researchers and practitioners to showcase their work in the field of neural networks. The conference featured a range of plenary talks, paper presentations, and tutorial sessions, covering various tracks and topics. Despite being held online, the conference offered valuable interactions and opportunities for knowledge sharing among participants.
Summary: Attended the Exciting International Joint Conference on Neural Networks (IJCNN) 2021
The International Joint Conference on Neural Networks (IJCNN) 2021 was a prestigious event organized by the International Neural Network Society. The conference covered a wide range of topics in neural networks theory, analysis, and applications. It was held online on a virtual platform, with participants able to access the complete proceedings through a downloadable link. I had the opportunity to present our paper on fruit classification using deep feature maps. The conference featured plenary talks by experts in the field, covering topics like smart interactive buildings and active interfaces. The schedule was well-managed, although attendee participation in asking questions could be improved.
Frequently Asked Questions:
1. What is deep learning and how does it work?
Deep learning is a subset of artificial intelligence that enables computers to learn, analyze, and make decisions similar to the way humans do. It uses artificial neural networks, which are inspired by the structure of the human brain, to process complex data and extract meaningful patterns or features. Through a process known as training, deep learning models can improve their performance over time by adjusting their weights and biases based on the feedback received from the provided data.
2. What are the applications of deep learning?
Deep learning has found applications in various fields due to its ability to handle large amounts of complex data. Some common applications include image and speech recognition, natural language processing, autonomous vehicles, recommendation systems, fraud detection, and medical diagnosis. Deep learning algorithms have also been used in areas like finance, manufacturing, and even creative fields like art and music.
3. How does deep learning differ from traditional machine learning?
Traditional machine learning models often require manual feature extraction, where domain experts convert raw data into specific features that the model can understand. On the other hand, deep learning models can automatically learn and extract relevant features from the raw data without human intervention. This makes deep learning particularly suitable for complex and unstructured data with high dimensionality. Deep learning models also tend to excel at tasks that involve large-scale data and can learn hierarchical representations.
4. What are the limitations of deep learning?
While deep learning has demonstrated remarkable achievements, it does have some limitations. One limitation is the requirement for a large amount of labeled training data, which can be costly and time-consuming to acquire. Deep learning models are also known to be computationally resource-intensive, often requiring powerful hardware and GPUs to train effectively. Interpreting the decisions made by deep learning models can be challenging, as they can be considered as black boxes due to their complexity. Finally, deep learning may not perform optimally for tasks with limited training data or in scenarios where interpretability and explainability are crucial.
5. What are some popular deep learning frameworks?
There are several popular deep learning frameworks that provide the necessary tools and libraries to build and train deep learning models. Some examples include TensorFlow, developed by Google Brain, which is widely used and offers a flexible ecosystem for different applications. PyTorch, developed by Facebook’s AI Research lab, is known for its dynamic computational graphs and intuitive programming interface. Other frameworks like Keras, Caffe, and MXNet are also commonly used and offer their unique features and advantages. The choice of framework often depends on the specific requirements of the project and personal preferences.
