The previous article briefly talks about ML and DL. Now let’s talk more about them.
Machine Learning (ML)
Machine Learning is a part/subset of AI, but not the same as AI. In general, ML is a discipline focused on two inter-related questions:
- “How can one construct computer systems that automatically improve through experience?”
- “What are the fundamental theoretical laws that govern every learning system, regardless of whether it is implemented in computers, humans or organizations?”
In other viewpoint, ML is to develop algorithms which can produce some output values based on some input values with the help of statistical analysis. This procedure is dynamic and does not require human intervention to make certain changes.
The analysis in ML usually attempts to optimize along a certain dimension; i.e. try to minimize error or maximize the likelihood of their predictions being true. Although there is really a lot of math under it, we can just leave it to the scientists and only consider how ML can help us build the analytical model automatically. We can simply treat ML as the ability of machines automatically learn without being explicitly programmed. A much detail ML pipeline is illustrated in the following figure.
A simple machine learning pipeline
There are several groups of machine learning: supervised, unsupervised, reinforcement learning, etc. More detailed explanations regarding the division will come in future articles. At this point, we will focus on machine learning itself and deep learning, which is a subset of ML.
Deep Learning (DL)
When it comes to DL, it does not have a really precise definition, but it often refers to the deep artificial neural networks, which is inspired by how neurons work in human brain. “Deep” is a technical term here. It refers to the number of layers in a neural network. The artificial neural networks are capable of learning and taking decisions intelligently with the help of algorithm. Since DL usually composes of algorithms like multilayered neural networks to deal with vast amounts of data, it is only a subset of machine learning.
The layered structure of algorithms in DL, called the artificial neural network, is inspired by the human brain. It might be something looks like Figure 5. Neurons, axons and dendrites are imitated in the neural network structures. The structure and relations between the more complex data structures need to be computed during the model construction. When everything is in place, the experiments can begin in order to test the desired output.
In a nutshell…
We could apply the same definition to deep learning as it to machine learning. However, deep learning tends to result in higher accuracy and performs exceptionally well on machine perception tasks that involved unstructured data. It also requires more hardware, training time, and computing resources. This is why computing hardware such as GPUs are in large demand with in recent years.
Here we roughly introduced through AI, ML, and DL. In these years, AI field is changing faster than its history can be written. It shows that AI is capable of solving harder and harder problems better than humans can. In the following articles, we will move on to some machine learning applications and examples.