Introduction

This post is based off a dataset I was working with recently as a side project. I wanted to put some of Python’s libraries to use, like matplotlib, seaborn, keras and pandas. My goal was to identify some correlations in the data through visualization techniques, and design a multi-layer perceptron that could leverage these correlations in order to make accurate predictions on previously unseen parts of the dataset.

Introduction

When trading the financial markets, traders will want to use any and all tools they can to pick out important information from the vast amounts of noisy data. Many technical indicators are used in the financial markets as a way to detect profitable trade signals, as well as entries and exits.

I thought that I would switch up the mood from technical to theoretical. The question is, what’s impeding the emergence of Super AIs - in other words, why don’t we have extremely intelligent AIs like HAL 9000 from 2001: A Space Odyssey, or Cortana from Halo - ones that can assist us in our daily lives and have the potential to learn almost anything. This is my take on it.

In the last tutorial, we discussed what neural networks are, as well as the underlying math and theory behind them. Specifically, we created a one-layer neural network that tries to learn the trend of an XOR logic gate. In this tutorial, we will discuss hidden layers, and why the XOR problem cannot be solved using a simple one-layer neural network.

The basics of neural networks

Traditionally, programs need to be hard coded with whatever you want it to do. If they are programmed using extensive techniques and painstakingly adjusted, they may be able to cover for a majority of situations, or at least enough to complete the necessary tasks. However, neural networks are a type of algorithm that’s capable of learning. Well, kinda. Let me explain..

Blog incoming my guy…