In their paper, Kreiman et al. examined what it takes for an artificial neural network to be able to perform well on a new task without forgetting its previous knowledge. By comparing methods that stop task forgetting, they found that longer training times and maintenance of the most important connections in a particular task while training on a new one helped the neural network maintain its performance on both tasks. The authors hope that this proof-of-principle research will someday contribute to artificial intelligence that better mimics natural human intelligence.
The mountain chain of the Western Ghats on the Indian peninsula, a UNESCO World Heritage site, is home to about 200 frog species, 89 of which are endemic. Distinctive to each frog species, their vocalizations can be used for species recognition. Manually surveying frogs at night during the rain in elephant and big cat forests is difficult, so being able to autonomously record ambient soundscapes and identify species is essential. An effective machine learning (ML) species classifier requires substantial training data from this area. The goal of this study was to assess data augmentation techniques on a dataset of frog vocalizations from this region, which has a minimal number of audio recordings per species. Consequently, enhancing an ML model’s performance with limited data is necessary. We analyzed the effects of four data augmentation techniques (Time Shifting, Noise Injection, Spectral Augmentation, and Test-Time Augmentation) individually and their combined effect on the frog vocalization data and the public environmental sounds dataset (ESC-50). The effect of combined data augmentation techniques improved the model's relative accuracy as the size of the dataset decreased. The combination of all four techniques improved the ML model’s classification accuracy on the frog calls dataset by 94%. This study established a data augmentation approach to maximize the classification accuracy with sparse data of frog call recordings, thereby creating a possibility to build a real-world automated field frog species identifier system. Such a system can significantly help in the conservation of frog species in this vital biodiversity hotspot.
Droughts have a wide range of effects, from ecosystems failing and crops dying, to increased illness and decreased water quality. Drought prediction is important because it can help communities, businesses, and governments plan and prepare for these detrimental effects. This study predicts drought conditions by using predictable weather patterns in machine learning models.
Here seeking to develop a method to diagnose, hypertrophic cardiomyopathy which can cause sudden cardiac death, the authors investigated the use of a convolutional neural network (CNN) and long short-term memory (LSTM) models to classify cardiac magnetic resonance and heart electrocardiogram scans. They found that the CNN model had a higher accuracy and precision and better other qualities, suggesting that machine learning models could be valuable tools to assist physicians in the diagnosis of hypertrophic cardiomyopathy.
There are believed to be ~20,000 nebulae in the Milky Way Galaxy. However, humans have only cataloged ~1,800 of them even though we have gathered 1.3 million nebula images. Classification of nebulae is important as it helps scientists understand the chemical composition of a nebula which in turn helps them understand the material of the original star. Our research on nebulae classification aims to make the process of classifying new nebulae faster and more accurate using a hybrid of deep learning and machine learning techniques.
Many common respiratory illnesses like bronchitis, asthma, and chronic obstructive pulmonary disease (COPD) lead to bronchial inflammation and, subsequently, a blockage. However, there are many difficulties in measuring the severity of the blockage. A numeric metric to determine the degree of the blockage severity is necessary. To tackle this demand, we aimed to develop a novel human respiratory model and design a deep-learning program that can constantly monitor and report bronchial blockage by recording breath sounds in a non-intrusive way.
This study hypothesized that a machine learning model could accurately predict the severity of California wildfires and determine the most influential meteorological factors. It utilized a custom dataset with information from the World Weather Online API and a Kaggle dataset of wildfires in California from 2013-2020. The developed algorithms classified fires into seven categories with promising accuracy (around 55 percent). They found that higher temperatures, lower humidity, lower dew point, higher wind gusts, and higher wind speeds are the most significant contributors to the spread of a wildfire. This tool could vastly improve the efficiency and preparedness of firefighters as they deal with wildfires.
Here, recognizing the difficulty associated with tracking the progression of dementia, the authors used machine learning models to predict between the presence of cognitive normalcy, mild cognitive impairment, and Alzheimer's Disease, based on blood DNA methylation levels, sex, and age. With four machine learning models and two dataset dimensionality reduction methods they achieved an accuracy of 53.33%.
Emissions from oil and natural gas (O&G) wells such as nitrogen dioxide (NO2), volatile organic compounds (VOCs), and ozone (O3) can severely impact the health of communities located near wells. In this study, we used O&G activity and wind-carried emissions to quantify the extent to which O&G wells affect the air quality of nearby communities, revealing that NO2, NOx, and NO are correlated to O&G activity. We then developed a novel land use regression (LUR) model using machine learning based on O&G prevalence to predict emissions.
