Students drop-out rates is one of the challenges faced by many schools globally.Addressing
this problem requires a better planning for interventions and a more thorough understanding
of the fundamental issues that cause the problem. Machine learning provides a convenient
way to accurately identify and prioritize students and schools at risk and thus deliver
good guarantees for the solutions.
Notebook
Money has become an inseparable aspect of human life. It can be used for making payments of various services such as buying
food, transportation, health care and many more. Despite the enormous use of cash in
our daily life, recognition of the banknotes is still a big challenge for visually impaired
people. Deep learning techniques offers both cost effective and scalable ways for image
recognition. Using a dataset of banknote images, I have applied transfer learning to
retrain a deep convolution neural network to identify five types of Tanzanian banknotes
(500, 1000, 2000, 5000, 10000). The robustness and generalizability of the proposed model
are evaluated on a dataset with Tanzanian banknotes collected under a variety of conditions.
Having a solid foundation in machine learning (ML) by implementing core concepts from scratch such as backpropagation algorithm
(for NNs, CNNs and RNNs) is important. Take the time to understand its derivations, and
try to derive it yourself from scratch and also implement it from scratch in code and
see if you can make it work. The knowledge you gain will stick and will be independent
of any framework you decide to learn later. In this article I present to you my simple
implementation of a two layer NN in pure java.
View on Medium
I like spectacular graphics provided by the JavaFX API, so I thought it could be nice to
find a way I can invoke a trained Deep Learning model from Java.
View on GitHub
Simple steps to understanding the redux architecture by implementing a todo app.
View on GitHub
More on GitHub