Deep Learning is a subfield of MachineLearning that deals with algorithms whose structure and function are inspired by the human brain and are referred to as artificial neural networks. Artificial neural networks resemble the structure of our nervous system: Individual neurons are interconnected and pass on information.
A typical artificial neural network has at least three layers: an input layer, a hidden layer and an output layer. Depending on the number of hidden layers, the artificial neural network can have any complexity. One speaks of Deep Learning, if an artificial neural network has a particularly deep network structure (thus many hidden layers).
Deep Learning is used in a wide variety of areas: For example, in speech recognition (Alexa, Google Assistant , Siri, etc.), autonomous driving or cancer cell detection in medicine.
Deep Learning is partly responsible for the rapid progress of artificial intelligence in recent years. Neural networks are particularly well suited for processing unstructured data. This is especially crucial in the age of Big Data and rapidly growing data volumes. One example of processing unstructured data is unstructured image recognition. Frequently used algorithms for image recognition are Convolutional Neural Networks.
Basically, Deep Learning is a special form of Machine Learning. Accordingly, every form of DeepLearning is also Machine Learning. Nevertheless, there are some differences.
The most important difference between Deep Learning and Machine Learning is that humans play a greater role in Machine Learning than in Deep Learning. While humans intervene in the analysis of the data in Machine Learning, they only provide the information in Deep Learning. The actual decision making lies with the Deep Learning model. This makes it possible to make discoveries in data even though the developers don't know exactly what they are looking for.
Another important difference between Deep Learning and Machine Learning is the amount of data required. Although both require a large amount of data, Deep Learning often requires a larger amount of data to perform well. If Deep Learning models do not have a correspondingly large amount of data available, they often perform worse than classic Machine Learning models. However, if the amount of data is large enough, the performance of deep learning models improves as the amount of data increases, while the performance of classical machine learning models shows no improvement after a saturation point.
The two major advantages of Deep Learning are the ability to process large amounts of unstructured data and the reduced need for human intervention in the algorithm. Some disadvantages of Deep Learning are the large amount of data required and the accompanying computational power needed. In addition, Deep Learning models (mostly) cannot provide explanations for their predictions because they function as a "black box." This complicates both the interpretation of the predictions and the evaluation of the model's performance.
Sources (translated): Betterprogramming and Medium
Damage good. All good.
