Written By: Chris Nosowsky
This repository is dedicated for the most common ML algorithms in the data science world.
Without further do, I will present my work on the machine learning algorithms I have designed for specific use-cases.
2) Keras
Haha funny you should ask, here is a guide posted by the very well known scikit-learn:
Machine Learning is taking data, training that data initially, then sending in some test data to see if it predicts the right outcome. Data in machine learning is represented in a vectorized fashion. Each piece of data has a point in space. This is what we know. We know the data and it's location, but can the machine know how to predict what class that data belongs too? That all depends on your optimizer and how good it truly is.
Deep Learning is built off of Machine Learning. It actually hasn't even come into existance since the 2010's. Deep Learning has another element..layers. It created these layered representations of data using a neural network. A neural network is loosely, and I mean LOOSELY modeled after the brain. By all means, this is not a system that will eventually lead to a chip being implanted into our brains to create singularity. It will never reach that point, in fact. Neural Networks is just a mathematical framework for being able to create representations from data. Deep Learning adds the layers to create better representations. For example, digits. Deep Learnig can backpropogate digits (backpropogate for now is just breaking down an image into smaller pixels for training). The digits get split up into segments where the machine learns certain features and geometric differences between digits. It can go through a few layers, or hundreds of layers to find the right answer. It is all up to you to decide whether more would be better for your deep neural network or not.