Avatar
Marvin Fuchs
Student and aspiring Data Scientist
Enthusiastic about Machine Learning - the intersection of Math, CS and Programming

Welcome!

My name is Marvin Fuchs and I am a 19-year old Computer Science student from Darmstadt, Germany. At the moment I am interested in Convolutional Neural Networks (CNNs) and their magical application in Neural Style Transfer. While I am exploring the landscape of Machine Learning autodidacticially, especially Deep Learning, I want to document my journey of understanding along these algorithms via BlogPosts. I hope that I will be able to unveil the underlying mathematical intuition behind those Deep Learning Archtitectures and share interesting insights. Feel free to contact me :)

More Information

Papers recently read:

  • An Introduction to Convolutional Neural Networks (link)
  • A Neural Algorithm of Artistic Style (link)

Current Interests:

  • How to implement CNN Backpropagation effieciently in plain NumPy
  • How to define the Costfunction for Artistic Style Transfer
  • Achieving Neural Sytle Transfer from scratch / in plain NumPy
  • Using RNNs to make Stock Price Predictions
  • Are both ideas connected ?
    • idea 1: Maximizing the activation of a specific neuron in a specific convolutional layer in a pretrained CNN model by adapting the input image
    • idea 2: Haluszination Phenomen stimulated by f.i. psychedelical drugs that may be activating a specific neuron in a specific layer in the visual apparatus in the brain
  • Tensor Calculus & Graph Theory
  • Tensorflow Library for ML
In this article I want to derive mathematically the back propagation algorithm used by Artificial Fully Connected Neural Networks to optimise a cost function commonly used in Deep Learning.
2021-02-07
4 min read