Automated Object Recognition System based on Convolutional Autoencoder

2020

Preprint

Self Cite

An analysis of a combined dataset of Wave 1 and 2 cases, aligned at approximately Local Time Zero + 2 months with unsupervised machine learning methods such as PCA and deep autoencoder dimensionality reduction allows to clearly separate milder background cases from those with more rapid and aggressive onset of the epidemics. The analysis and findings of the study can be used in evaluation of possible epidemiological scenarios and as an effective modeling tool to design corrective and preventative measures to avoid developments with potentially heavy impact.

Section: Resultsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Unsupervised Autoencoder Modelmentioning

confidence: 99%

See 1 more Smart Citation

Identifying Explosive Epidemiological Cases with Unsupervised Machine Learning

2020

Preprint

Self Cite

“…Disentangled representations were produced and discussed in [13] with deep variational autoencoder architecture and different types of visual data, pointing at the possibility of a general nature of the effect. Concept-associated structure in artificial neural networks was observed in the representations of generative models with different types of real-world data as Internet traffic [14], anomaly detection [15], medical and aerial surveillance imaging [16], [17], linguistics [18] and a range of generative architectures [19], [20], pointing to a possibility of a deep connection between unsupervised generative learning and characteristic structures in the latent representations of learning models that can be interpreted as general concepts in the observable data.…”

Section: A Related Workmentioning

confidence: 99%

Can Jellyfish Dream? Conceptual Representations in Unsupervised Generative Learning

2021

Preprint

Self Cite

Representations play essential role in learning of artificial and biologic systems by identifying characteristic patterns in the observed environment. In this work we examine unsupervised latent representations of image data with a collective of generative neural network models. A convolutional autoencoder with strong redundancy reduction was used to create low-dimensional latent representations of a dataset of geometrical shapes. The structure of the resulting latent representations was studied comprehensively with several methods, including density clustering, direct visualization, generative probing and scanning. It was demonstrated that conceptual representations with good decoupling of principal concepts can be produced with generative models of limited complexity; that latent representations have high level of consistency between individual learners; and that the resulting latent representations have a well-defined geometrical and topological structure correlated with principal patterns in the observable data. The results of this work support the hypothesis that conceptual latent representations can emerge naturally in unsupervised generative learning under certain essential constraints and can be a natural platform for emergence of some intelligent functions and behaviors.

“…On the other hand, methods of unsupervised machine learning [9,10] have shown an effective ability to achieve significant reduction of dimensionality, or redundancy of the observable parameter space that in a number of cases were instrumental in the analysis and determination of characteristic patterns and trends in complex data [11][12][13] including constrained data [14]. Importantly, application of these methods does not require data labeled with confidently known outcome and generally can be performed with smaller samples of data.…”

Section: Introductionmentioning

confidence: 99%

Analysis and Augmentation of Small Datasets with Unsupervised Machine Learning

2021

Preprint

Self Cite

Analysis of small datasets presents a number of essential challenges not in the least due to insufficient sampling of characteristic patterns in the data making confident conclusions about the unknown distribution elusive and resulting in lower statistical confidence and higher error. In this work, a novel approach to augmentation of small datasets is proposed based on an ensemble of neural network models of unsupervised generative self-learning. Applying generative learning with an ensemble of individual models allowed to identify stable clusters of data points in the latent representations of the observable data. Several techniques of augmentation based on identified latent cluster structure were applied to produce new data points and enhance the dataset. The proposed method can be used with small and extremely small datasets to identify characteristics patterns, augment data and in some cases, improve accuracy of classification in the scenarios with strong deficit of labels.