Deep Learning

Created by: Felix Weber

Deep learning can be referred to as a type of machine learning which utilizes either unsupervised or supervised algorithm or at times both. It is not a new technology but has gained popularity in the recent past as a good approach for accelerating the solution of different types of complex computer issues in the field of Natural Language Processing (NLP) and computer vision fields.

Table of Content [Hide]

Deep learning can be referred to as a type of machine learning which utilizes either unsupervised or supervised algorithm or at times both. It is not a new technology but has gained popularity in the recent past as a good approach for accelerating the solution of different types of complex computer issues in the field of Natural Language Processing (NLP) and computer vision fields.

There are a number of hidden layers in neural networks and deep learning is a form of representation learning associated with the theory of machine learning [1]. Data learning techniques extract complex and high-level abstractions in the form of data representations via a hierarchical learning process which helps them in yielding quick results. In simple terms, deep learning automatically acquires the important features rather than wait for a data scientist to select the important features manually.

The term ‘deep’ in the phrase deep learning has been derived from the multiple layers that make up the deep learning models generally referred to as neural networks. Two key factors are considered by deep learning and they are unsupervised or supervised learning and non-linear processing in different stages or layers. The latter indicates an algorithm in which the existing layer acquires the output from the previous layer in the form of an input. The layers are hierarchically set for identifying the significance of the data [2]. Unsupervised and supervised learning is associated with a class target label where availability refers to a supervised systems and its absence refers to an unsupervised system.

The user in supervised learning trains a program to provide an answer which is based on a labelled and known data set. Regression and classification algorithms that include random forests, support vector machines and decision trees are used generally for administered learning tasks. The algorithms in unmonitored machine learning provide answers for unlabelled and unknown data. Unsupervised methods are used by data scientists for identifying the patterns of advanced data sets. One such unsupervised ML is clustering algorithms like K-means.

ML algorithms can be programmed by data scientists with the help of different languages and technologies like Python, Java etc. pre-built ML frameworks can also be used for accelerating the process. For example, Mahout was quite popular on the Apache Hadoop whereas the ML library of Apache Spark is standardized now.

References

  1. Pariwat Ongsulee; Artificial intelligence, machine learning and deep learning. 2017 15th International Conference on ICT and Knowledge Engineering (ICT&KE) 2017, 1, xx, 10.1109/ictke.2017.8259629.
  2. Michalski, R. S., Carbonell, J. G., & Mitchell, T. M.. Machine learning: An artificial intelligence approach; Springer: Berlin, 2013; pp. xx.

Cite this article

Felix, Weber. Deep Learning, Encyclopedia, 2019, v1, Available online: https://encyclopedia.pub/119