Neural Networks and Trading
There are of course many kinds of networks that exist in our world, but one particular system is proving to be revolutionary in that it will continue to change engineering and business forever. The concept of neural networks is such that there is a set of algorithms that are “trainable”. These self-serving algorithms are able to take inputs and “re-feed” themselves these outputs that result in the machines working self-sufficiently. This process of constantly reevaluating its own outputs as inputs is very similar to the concepts of machine learning and artificial intelligence at their most basic forms. The algorithms are able to function very similarly to the human brain as neural networks are made to implement the neurological processes that occur in our brains. The self-training ability they have allows them to formalize information and gives them the ability to make predictions based on the information that they were able to store and organize. This allows individual algorithms to simulate our decision making process so that we would be able to make accurate predictions, especially in areas such as trading. The article focuses on how neural networks can be incorporated into the strategy of a trader and how the important aspect of these structures is the underlying network rather than the algorithms themselves.
Neural networks can be represented with the node and edge graphs that we have covered in class. A basic graph would look something like this:
The input units would be the carefully organized information that will be fed to the network. The algorithms within the network’s hidden units will evaluate the data so that they are able to produce output units. However, the output units are fed back as input units and are evaluated again for several times before giving traders the results needed. These seemingly magical algorithms however, are not how traders are able to make huge returns. By having several different algorithms store information in a certain way in this fully connected system, traders are able to plan exactly how to go about their strategy. The weights for the hidden units are changed and morphed so that they would be able to present a data set where traders can find opportunities to make huge returns. Yet the neural network itself never truly strays away from the model of a general network. It is created as a baseline, as many social networks are for social media sites, except for ranges of business applications and engineering applications. We can see how they are similar to social networks in that their model is similar except for friends as our nodes we have data/information and our edges are determined by what our algorithms are capable of doing while social network edges represent friendship/friend grouping. Regardless, we can expect to see more trading uses of neural networks to show in the near future and how traders will utilize these networks to expose the holes necessary for furthering business and the economy all over the world.