To understand how A.I. Artificial Intelligence functions are developed, we first need to understand the technology that makes this possible. The term deep comes from deep learning, a branch of Machine Learning that focuses on deep neural networks.
Neural networks are computational systems that are inspired by the ways a human brain processes certain information. Special cells called neurons are connected to each other in a complex network allowing information to be processed and communicated.
In Computer Science, artificial neural networks are made out of thousands of nodes, connected in a specific manner. Nodes are arranged in layers; the way in which they are connected determines the type of the network and, ultimately, its ability to perform a certain computational task over another one. A traditional neural network might look like this:
Each node from the input layer contains a numerical value that encodes the input we want to feed the network. If we are trying to predict the Dow Jones Industrial Average for tomorrow, the input nodes might contain the future and past price changes, stock splits, divisors, contracts, fear and greed index encoded as numbers in the range.
These values are broadcasted to the next layer; each result-curve dampens or amplifies the values it transmits. Each node sums all the values it receives, and outputs a new value based on its own function. The result of the computation can be retrieved from the output layer; in this case, only one value is produced, the probability of the Dow Jones Industrial Average.
When using Adept Enterprise we use multiple input nodes that are triggered by events, either by humans or by machine code. The Adept Enterprise software outputs the desired result based on the sum of all result-curves via the input layer values it receives. The output layer result is the most desired result. This allows for the software to give and assign unlimited desired results, without a human manually doing it. This allows one human with the software to accomplish the same work load as hundreds of humans without the software, and in some cases based on the job thousands of humans.
Training a neural network means finding a set of weights for all result-curves, so that the output layer produces the desired result. One of the most used technique to achieve this is called back-propagation, and it works by re-adjusting the weights every time the network makes a mistake. The mistake is determined by not accomplishing the desired result, or the desired result is not desired anymore.
The basic idea behind training a neural network is that each layer will represent progressively core complex features. In the case of a workflow task, for instance, the first layer might detect violations, the second layer detects date and time restrictions, which the third layer is able to use to approve or deny an application or task.
In experience, what each layer responds to is far from being that simple. This is based on humans and software creating a library of input layers, and the software producing the desired results for the output layers.
The desired result is a perfect result, eliminating imperfection. This saves money, time and frustration.
Contact us today toll free 1-888-392-9623 to find out more on how Adept Technologies can save you money by utilizing our technology.
