ReLU, Sigmoid and Tanh: today’s most used activation functions
Today’s deep neural networks can handle highly complex data sets. For example, object detectors have grown capable of predicting the...
Vanishing and exploding gradients
Neural networks must be initialized before one can start training them. As with any aspect of deep learning, however, there...
Weight initialization in neural networks: what is it?
An important predictor for deep learning success is how you initialize the weights of your model, or weight initialization in...
MLP for regression with TensorFlow 2 and Keras
Machine learning is a wide field and machine learning problems come in many flavors. If, say, you wish to group...
How to create an MLP classifier with TensorFlow 2 and Keras
In one of my previous blogs, I showed why you can’t truly create a Rosenblatt’s Perceptron with Keras. Fortunately for...
Why you can’t truly create Rosenblatt’s Perceptron with Keras
…and what to do about it! It was January 1957 when a report was released by Cornell Aeronautical Laboratory. It...
Linking maths and intuition: Rosenblatt’s Perceptron in Python
According to Wikipedia, Frank Rosenblatt is an “American psychologist notable in the field of artificial intelligence”. And notable, he is....
Commoditizing AI? The state of automated machine learning.
It cannot go unnoticed that machine learning has been quite the hype these past few years. AI programs at universities...
CNNs and feature extraction? The curse of data sparsity.
I recently finished work on my master’s thesis in which I investigated how the process of mapping underground utilities such...
Can neural networks approximate mathematical functions?
In the paper Multilayer feedforward networks are universal approximators written by Kurt Hornik, Maxwell Stinchcombe and Halbert White in 1989,...
