View source for Unsupervised learning

{{template: machine learning}}
'''Unsupervised learning''' is a type of [[machine learning]] where algorithms find patterns in data without any labels. This is different from supervised learning, where data comes with labels to guide the learning process. Other types of learning, like weak or semi-supervised learning, use some labeled data, while self-supervised learning is sometimes considered a form of unsupervised learning.

In unsupervised learning, the data is often collected from sources like the [[internet]] with little processing, making it cheaper than supervised learning, where labeled datasets are created manually.

There are special algorithms for unsupervised learning, such as k-means for grouping similar data, PCA for reducing data complexity, and autoencoders for learning useful representations. With the rise of deep learning, large-scale unsupervised learning is mostly done using neural networks trained with gradient descent and special training techniques.

Sometimes, models trained with unsupervised learning are used directly, but often they are adjusted for other tasks. For example, a model trained to generate text can later be fine-tuned for tasks like text classification. Similarly, autoencoders can learn useful features that help improve other models, such as those used in image generation.<ref>{{Cite journal|last=Wu|first=Wu|date=2022|title=Unsupervised Learnin|url=https://na.uni-tuebingen.de/ex/ml_seminar_ss2022/Unsupervised_Learning%20Final.pdf|journal=Mathematisches Institut Universität Tübingen}}</ref><ref>
== Unsupervised Learning Media ==
<gallery widths='160px' heights='100%' mode='traditional' caption=''>
File:Task-guidance.png|Tendency for a task to employ supervised vs. unsupervised methods. Task names straddling circle boundaries is intentional. It shows that the classical division of imaginative tasks (left) employing unsupervised methods is blurred in today's learning schemes.

File:Hopfield-net-vector.svg          |A network based on magnetic domains in iron with a single self-connected layer. It can be used as a content addressable memory.

File:Boltzmannexamplev1.png          |Network is separated into 2 layers (hidden vs. visible), but still using symmetric 2-way weights. Following Boltzmann's thermodynamics, individual probabilities give rise to macroscopic energies.

File:Restricted Boltzmann machine.svg|Restricted Boltzmann Machine. This is a Boltzmann machine where lateral connections within a layer are prohibited to make analysis tractable.

File:Stacked-boltzmann.png|This network has multiple RBM's to encode a hierarchy of hidden features. After a single RBM is trained, another blue hidden layer (see left RBM) is added, and the top 2 layers are trained as a red & blue RBM. Thus the middle layers of an RBM acts as hidden or visible, depending on the training phase it is in.

File:Helmholtz Machine.png           |Instead of the bidirectional symmetric connection of the stacked Boltzmann machines, we have separate one-way connections to form a loop. It does both generation and discrimination.

File:Autoencoder_schema.png          |A feed forward network that aims to find a good middle layer representation of its input world. This network is deterministic, so it is not as robust as its successor the VAE.

File:VAE blocks.png                  |Applies Variational Inference to the Autoencoder. The middle layer is a set of means & variances for Gaussian distributions. The stochastic nature allows for more robust imagination than the deterministic autoencoder.
</gallery>
{{Cite journal|last=Liu|first=Xiao|last2=Zhang|first2=Fanjin|last3=Hou|first3=Zhenyu|last4=Mian|first4=Li|last5=Wang|first5=Zhaoyu|last6=Zhang|first6=Jing|last7=Tang|first7=Jie|date=2023-01|title=Self-Supervised Learning: Generative or Contrastive|url=https://ieeexplore.ieee.org/document/9462394|journal=[[IEEE]] Transactions on Knowledge and Data Engineering|volume=35|issue=1|pages=857–876|doi=10.1109/TKDE.2021.3090866|issn=1558-2191}}</ref>
== References ==
<references />
[[Category:Machine learning]]