A comprehensive survey and analysis of generative models in machine learning

Harshvardhan, GM; Gourisaria, Mahendra Kumar; Pandey, Manjusha; Rautaray, Siddharth Swarup

doi:10.1016/j.cosrev.2020.100285

Cited by 126 publications

(28 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The hardware with processor specification Intel® Core ™ i5-8250U CPU with 64-bit operating system, x64-based processor was used. The programming language that was used is Python which is a dynamic language extensively used in the field of machine learning [38,39], neural networks [33], internet of things and many other such newer technologies. Scientific Python Development EnviRonment (Spyder) 3.3.6 version and Jupyter Notebook 6.0.0 version were used as the software for the experimental implementation.…”

Section: E Software and Hardwarementioning

confidence: 99%

Distinguishing Pneumonia and COVID-19: Utilizing Computer Vision to Mimic Clinician Efficacy

Saha

Gourisaria

Harshvardhan

2021

2021 International Conference on Artificial Intelligence and Smart Systems (ICAIS)

Self Cite

View full text Add to dashboard Cite

Corona Virus has caused a disruption to the normalcy in the world and require thorough patient management. Looking at the present scenario and the kind of pandemic that it has turned out to be, this paper aims at providing an enhancement to the RT-PCR way of testing and uses the chest X-Rays to detect the presence and severity of the corona virus in a body to successfully differentiate between pneumonia and the Covid-19. The main motive is to help doctors and medical experts with an advance aid to the nursing of critically affected patients. This is feasible because the X-Ray machines are widely available throughout the country and they can assist in the advanced detection of the disease. Around 6000 images of the three kinds of chest X-Rays of patients with pneumonia, Covid, as well as completely normal patients, were used in the process. The paper concludes with the explicit comparison of all the models and their results. Primarily, a simple CNN model was opted for the scrutiny and then later on VGG-16, VGG-19, ResNet50, MobileNet and MobileNetV2 pre-trained models were utilised for anatomizing Covid-19 with respect to pneumonia and normal cases. In case of CNN, the maximum accuracy that was attained was 95.30% whereas, for the VGG-16, VGG-19, ResNet50, MobileNet and MobileNetV2, the maximum accuracies in correctly predicting the diseases were 95.63% , 96.02% , 94.82% , 95.23% and 93.39% respectively.

show abstract

Section: E Software and Hardwarementioning

confidence: 99%

Distinguishing Pneumonia and COVID-19: Utilizing Computer Vision to Mimic Clinician Efficacy

Saha

Gourisaria

Harshvardhan

2021

2021 International Conference on Artificial Intelligence and Smart Systems (ICAIS)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Como se indicó anteriormente, una de las principales características de los modelos generativos es la capacidad de crear nuevos datos en base a cierta información proporcionada (Gm et al, 2020), por ejemplo, una distribución probabilística latente. De esta manera, se identificará como x al dato ingresado al modelo, también entendido como una instancia del conjunto de datos.…”

Section: Características De Los Modelos Generativosunclassified

Modelos Generativos en el Aprendizaje Automático y su aplicación a la generación de Imágenes Digitales

Carrasco

2021

Jour. Bol. Cien.

View full text Add to dashboard Cite

En el ámbito del Aprendizaje Automático se identifican dos tipos de algoritmos desde el punto de vista de la naturaleza de las salidas que estos proporcionan. Los modelos discriminativos asocian un dato proporcionado a una respuesta, mientras que los modelos generativos generan nuevos datos en base a una distribución probabilística de respuestas o variables latentes. Por otra parte, en años recientes, se han dado importantes avances en el Deep Learning, que es el estudio de las redes neuronales profundas. En este contexto, las redes convolucionales han ganado considerable terreno en diversas tareas relacionadas al análisis de imágenes. Entre las varias aplicaciones de las redes convolucionales se mencionan: clasificación de imágenes, detección de objetos, segmentación de instancias, reconocimiento facial, entre otros varios. Sin embargo, en el ámbito del Deep Learning, no solamente se han tenido importantes avances con relación a los puntos aquí mencionados, sino también en cuanto a la habilidad de los modelos en generar nuevas imágenes. De ese modo, hoy en día se tiene a disposición una variedad de modelos generativos para diversos propósitos, con aplicaciones tales como la creación de imágenes faciales de personas que no existen en la realidad. Es así que el propósito del presente artículo es realizar un análisis de diversos modelos generativos para imágenes digitales, así como de las bases teóricas que sustentan a los modelos generativos en el Deep Learning. El enfoque de análisis se centra en dos modelos esenciales: Las redes adversarias generativas (GAN) y los autoencoders variacionales (VAE).

show abstract

“…These waveforms can convey complex and varied information. Deep generative networks [1] have demonstrated great potential for such tasks having been used for the synthesis of a range of sounds, from pleasant pieces of music to natural speech [2]. These models discover latent representations based on the distribution of the initial data and then sample from this distribution to generate new acoustic signals with the same properties as the original ones.…”

Section: Introductionmentioning

confidence: 99%

Audio representations for deep learning in sound synthesis: A review

Anastasia¹,

O’Leary²

2022

Preprint

View full text Add to dashboard Cite

The rise of deep learning algorithms has led many researchers to withdraw from using classic signal processing methods for sound generation. Deep learning models have achieved expressive voice synthesis, realistic sound textures, and musical notes from virtual instruments. However, the most suitable deep learning architecture is still under investigation. The choice of architecture is tightly coupled to the audio representations. A sound's original waveform can be too dense and rich for deep learning models to deal with efficientlyand complexity increases training time and computational cost. Also, it does not represent sound in the manner in which it is perceived. Therefore, in many cases, the raw audio has been transformed into a compressed and more meaningful form using upsampling, feature-extraction, or even by adopting a higher level illustration of the waveform. Furthermore, conditional on the form chosen, additional conditioning representations, different model architectures, and numerous metrics for evaluating the reconstructed sound have been investigated. This paper provides an overview of audio representations applied to sound synthesis using deep learning. Additionally, it presents the most significant methods for developing and evaluating a sound synthesis architecture using deep learning models, always depending on the audio representation.

show abstract

A comprehensive survey and analysis of generative models in machine learning

Cited by 126 publications

References 31 publications

Distinguishing Pneumonia and COVID-19: Utilizing Computer Vision to Mimic Clinician Efficacy

Distinguishing Pneumonia and COVID-19: Utilizing Computer Vision to Mimic Clinician Efficacy

Modelos Generativos en el Aprendizaje Automático y su aplicación a la generación de Imágenes Digitales

Audio representations for deep learning in sound synthesis: A review

Contact Info

Product

Resources

About