In this study, unsupervised and supervised classification methods were compared for comprehensive analysis of the fingerprints of 26 Phyllanthus samples from different geographical regions and species. A total of 63 compounds were identified and tentatively assigned structures for the establishment of fingerprints using high-performance liquid chromatography time-of-flight mass spectrometry (HPLC/TOFMS). Unsupervised and supervised pattern recognition technologies including principal component analysis (PCA), nearest neighbors algorithm (NN), partial least squares discriminant analysis (PLS-DA), and artificial neural network (ANN) were employed. Results showed that Phyllanthus could be correctly classified according to their geographical locations and species through ANN and PLS-DA. Important variables for clusters discrimination were also identified by PCA. Although unsupervised and supervised pattern recognitions have their own disadvantage and application scope, they are effective and reliable for studying fingerprints of traditional Chinese medicines (TCM). These two technologies are complementary and can be superimposed. Our study is the first holistic comparison of supervised and unsupervised pattern recognition technologies in the TCM chemical fingerprinting. They showed advantages in sample classification and data mining, respectively.
There are two classification methods in pattern recognition: supervised and unsupervised classification. To apply supervised pattern recognition, you need a large set of labelled data; otherwise you can try to apply an unsupervised approach.
Extreme learning machines (ELMs) have proven to be efficient and effective learning mechanisms for pattern classification and regression. However, ELMs are primarily applied to supervised learning problems. Only a few existing research papers have used ELMs to explore unlabeled data. In this paper, we extend ELMs for both semi-supervised and unsupervised tasks based on the manifold regularization, thus greatly expanding the applicability of ELMs. The key advantages of the proposed algorithms are as follows: 1) both the semi-supervised ELM (SS-ELM) and the unsupervised ELM (US-ELM) exhibit learning capability and computational efficiency of ELMs; 2) both algorithms naturally handle multiclass classification or multicluster clustering; and 3) both algorithms are inductive and can handle unseen data at test time directly. Moreover, it is shown in this paper that all the supervised, semi-supervised, and unsupervised ELMs can actually be put into a unified framework. This provides new perspectives for understanding the mechanism of random feature mapping, which is the key concept in ELM theory. Empirical study on a wide range of data sets demonstrates that the proposed algorithms are competitive with the state-of-the-art semi-supervised or unsupervised learning algorithms in terms of accuracy and efficiency.
With unsupervised learning, ML algorithms are used to examine and group unlabeled datasets. Such algorithms can uncover unknown patterns in data without human supervision. There are three main categories of algorithms:
In contrast, unsupervised learning models are constantly working without human interference. They find and arrive at a structure of sorts using unlabeled data. The only human help needed here is for the validation of output variables. For example, when someone shops for a new laptop online, an unsupervised learning model will figure out that the person belongs to a group of buyers who buy a set of related products together. However, it is the job of a data analyst to validate that a recommendation engine offers options for a laptop bag, a screen guard, and a car charger.
The goals with supervised and unsupervised learning are different. While the former is about the prediction of outcomes for new data that is introduced, the latter is about getting new insights from massive amounts of new data. In supervised learning, a user will know what results they can expect, whereas in unsupervised learning, they hope to discover something new and unknown.
Unsupervised learning sometimes produces completely erroneous results unless there is some form of human intervention to validate the results. Quite in contrast to supervised learning, unsupervised learning can work on any amount of data in real-time but, since the machine teaches itself, transparency on classification is lower. This increases the chances of poor results.
The decision of whether or not to opt for supervised or unsupervised ML approaches is subject to context, the basic assumptions that can be arrived at on the data on hand, and its final application. The use of either form can change over time as the needs of the organization change.
While an organization may begin training with unlabeled data and therefore use the unsupervised approach, with time, the correct labels will be identified and the machine can switch to supervised learning. This can happen over various stages of labeling. On the other hand, the supervised learning data approach may not be providing the insights required, and unsupervised learning may discover unknown patterns and give deeper insight into business mechanisms.
The use of image intensity based segmentation techniques are proposed to improve MRI contrast and provide greater confidence levels in 3-D visualization of pathology. Pattern recognition methods are proposed using both supervised and unsupervised methods. This paper emphasizes the practical problems in the selection of training data sets for supervised methods that result in instability in segmentation. An unsupervised method, namely fuzzy c- means, that does not require training data sets and produces comparable results is proposed.
Unsupervised learning algorithms allow you to perform more complex processing tasks compared to supervised learning. Although, unsupervised learning can be more unpredictable compared with other natural learning deep learning and reinforcement learning methods.
Clustering is an important concept when it comes to unsupervised learning. It mainly deals with finding a structure or pattern in a collection of uncategorized data. Clustering algorithms will process your data and find natural clusters(groups) if they exist in the data. You can also modify how many clusters your algorithms should identify. It allows you to adjust the granularity of these groups.
Association rules allow you to establish associations amongst data objects inside large databases. This unsupervised technique is about discovering exciting relationships between variables in large databases. For example, people that buy a new home most likely to buy new furniture.
In this direction, there are a few unsupervised systems for skeleton-based action recognition at this moment, since obtaining relevant feature representation of the skeleton can be challenging. The method Predict and Cluster  proposes a method of clustering similar activities into the same cluster and different activities into separate clusters in a fully unsupervised manner. They successfully implemented a method based on an encoder-decoder recurrent neural network, which can produce a feature space in its hidden states that can effectively learn distinct actions.
Unsupervised systems for skeleton-based activity recognition are limited [17,18], since obtaining relevant feature representations from spatial representation of the skeleton is challenging. The main objective of the approach presented in this paper is to obtain an intrinsic representation of the activities in such a manner that similar activities are clustered together, while different activities are placed apart from each other, when inspecting their hidden representation. The authors proposed two decoder strategies called Fixed Weights and Fixed States in order to penalize the decoder. Predict and Cluster unsupervised system achieves high accuracy performance and outperforms the supervised methods. The hidden representation comes from the hidden states of the last layer of the encoder module, which is then fed to the Decoder submodule. The important thing is to notice how, during the training of the encoder-decoder module, the labels of the activities are not used at all, as the encoder-decoder module does not need any information about the type of activity that is performed. Keeping this in mind, in order to validate that the encoder-decoder module works well, the authors proposed encoding each activity and extracting the features from the latent space, and then they proposed useing a KNN algorithm to assign labels based on the distances between those features in latent space and the ground-truth labels, resulting in over 90% accuracy on the NTU-RGB+D dataset. On the large scale NTU-RGB+D  dataset, Predict and Cluster performs extremely well on the cross-view test (see Table 1). These results prove that the encoder-decoder module is a good method for representing the activities in a latent space.
Combining supervised and unsupervised methods has been proposed in several works. Paper  proposed the enhancement of an unsupervised approach in image classification. The authors state that by jointly training supervised and unsupervised cost functions using backpropagation in a neural network, one can reduce the error rate of a task when compared to individual approaches. In essence, the supervised technique is depicted as a multi-layer perceptron (MLP) or convolutional neural network (CNN), while the unsupervised one is established on a decoder submodule. A more general approach is presented in . It aggregates between supervised and unsupervised techniques via unconstrained probabilistic embeddings over their ensemble. The proposal consists of conditioning the outputs of the supervised models on the constraints implied by the unsupervised ones.
The first part, the Discriminant, employs an unsupervised method of clustering data where the labels of the dataset are not used. The advantage of using unsupervised techniques is that it allows the addition of completely new, unseen samples without the need of retraining from scratch. The choice for the unsupervised algorithm is not fixed, as one can use any unsupervised clustering approach. 59ce067264