Deep Learning – From Stanley Kubrick’s 1968 movie“2001: A Space Odyssey” and the infamous HAL 9000, the concept and the idea of an AI that can learn as humans learn has been one of the go-to ideas for sci-fi movie makers and video game developers alike.
But up until now, true AI wasn’t really something you could see in real life and outside of your TV or computer screen. Now, however, things are changing rapidly when it comes to machine learning, especially one of its most interesting branches – deep learning. So let’s take a look at the 10 most important things you need to know about deep learning.
1. Deep Learning is a Branch of Machine Learning
Deep learning is not an isolated concept when it comes to artificial intelligence or AI, but is actually a branch or a subclass of machine learning that contains networks capable of learning from unstructured or unlabeled data.
2. First Introduced by the World School Council
The first algorithm for deep, supervised, multilayer perceptrons was published in 1965 by Lapa and Ivakhnenko, whose ideas were later used by the World School Council in London for their computer identification system named “Alpha”.
3. Real World Applications
So far, deep learning has found its way in many fields, including automatic speech recognition, audio recognition, bioinformatics and computer vision. In many cases, deep learning has produced results that are equal, if not even better than when the task is performed by humans.
4. Deep Learning Aims to Replace Hand-Crafted Feature Engineering
In many ways, deep learning has made feature engineering less complicated, but it hasn’t completely removed it yet. According to experts in deep learning, the architectures of the machine learning models themselves have in turn become more intricate.
5. Neural Networks
Neural networks are formed by inter-connected neurons and their goal is to calculate an unknown function. A neural network is defined as “made of numerous interconnected conceptualized artificial neurons, which pass data between themselves, and which have associated weights which are tuned based upon the network’s ‘experience’”.
6. Weight and Bias of Neurons
Neural networks consist of neurons, which in turn have two linear components – weight (usually labeled as W1) and bias. Every input entered into the neuron has a weight assigned to it. This represents how correct or incorrect the input is relative to the task performed. On the other hand, bias is added to the result of the weight multiplication.
A neural network has three layers – input, output and a hidden layer. An input layer is the first layer of the network and it receives the input. The output layer is the final layer and it receives the output. Between them is the hidden layer or the processing layer, which performs certain tasks on the incoming data and passes them on to next layers.
8. Two Steps of Deep Learning
The process of deep learning has two steps. The first step is the analysis of data and generating rules or algorithms that can describe the characteristics of an object. The second step is using those rules and algorithms to identify objects based on real time data.
9. A Machine Learning System (Yes, This Includes Deep Learning) Consists of:
A target function – what is being learned; performance element – the component that takes an action in the world; training data – data points used to calculate the target function; learning algorithm – an algorithm that uses training data to calculate the approximation of the target function and the hypothesis space – the space learning algorithm can consider for possible functions.
10. It Still Requires a Human Touch
Although we’ve seen some amazing advances in deep learning, machine learning in general and in artificial intelligence, it will be some time (decades, if ever) before we see a machine that can compare to a human brain. At this point, deep learning still requires a lot of help from humans in effectively solving a problem that was set to it.
There you go. 10 things you should know about deep learning. If you were worried about a Terminator scenario coming soon, don’t be. Machines have a long way to go before they catch up with us. For now, at least, they can beat us at Go.