Sedjro Salomon Hotegni

Sedjro Salomon Hotegni

PhD student in Data Science for Safe Autonomous Systems, TU Dortmund

Efficient Feature Selection using Random Adaptive Sub-space Learning (RASSEL): Python implementation

less than 1 minute read

Published:

An appropriate mix of good core learners leads to a reduction in prediction error and improves the precision of a regression or a classification task. This technique is known as ensemble learning (aggregation).

Direct Link

Reproducibility and Training State Management in PyTorch: Best Practices and Tips Permalink

less than 1 minute read

Published:

In the ever-evolving field of machine learning, reproducibility, and efficient training state management are crucial for research and practical applications. Whether you’re developing new models, experimenting with hyperparameters, or deploying solutions to production, ensuring that your results are consistent and that you can seamlessly resume training after interruptions can save you valuable time and resources. Read more

Can Artificial Neural Networks Truly Map Our Natural Neural Networks (the human brain)? Permalink

less than 1 minute read

Published:

Artificial Neural Networks (ANNs) have gained significant attention in the field of Deep Learning, driven by the belief that they can replicate the functioning of our brains. Geoffrey Hinton, a prominent figure in the development of ANNs, argues that the brain’s ability to learn through connection strengths serves as the foundation for creating intelligent machines. In a 2019 interview with Nicholas Thompson, Hinton emphasized the importance of mimicking this process in ANNs. This article delves into the perspectives of Geoffrey Hinton and Yann LeCun on the potential of ANNs to map our natural neural networks and examines their predictions for the future of artificial intelligence. Read more