Once again, science fiction turns real. Scenarios that lived in the minds of fiction writers are becoming a reality. In the form of coronavirus. As of this writing, infections are spreading worldwide and death rate percentages keep rising. Lockdowns are in effect across multiple countries and grocery shelves across the globe are empty of essential items as panicked consumers grab whatever essential goods remain.
Many are comparing coronavirus to the Spanish flu pandemic of 1918. During that time, at least 17 million people died from influenza. Unlike the early 20th century, the early 21st century has access to advanced tools to help battle contagious organisms. One of those tools is machine learning (also called deep learning).
Enter Machine Learning
Definition and Usages
Machine learning is a subset of artificial intelligence. It’s a collection of techniques used to spot patterns and works by being trained using what’s called training data. Machine learning can be thought of as the basis for AI and automation. One advantage of machine learning is that it can discover patterns across multiple data sets faster than humans and sometimes more accurate—depending on how well it’s “trained”.
Healthcare is starting to adopt machine learning to make breakthroughs in research. Here are a few examples:
- Google was able to train computers to detect cancer.
- MIT found an antibiotic compound that could be useful against microscopic “superbugs”.
- Researchers found that machine learning could be used to detect diabetic retinopathy based on photographs.
Most of the breakthroughs mentioned previously are the results of a machine learning technique called neural networks. To put it simply, a neural network model is an attempt to emulate the human mind. Codementor.io’s James Le provides a slightly more thorough explanation:
Neural Networks are a class of models within the general machine learning literature. Neural networks are a specific set of algorithms that have revolutionized machine learning. They are inspired by biological neural networks and the current so-called deep neural networks have proven to work quite well. Neural Networks are themselves general function approximations, which is why they can be applied to almost any machine learning problem about learning a complex mapping from the input to the output space.—James Le
Top Neural Network Models
Various neural network models already exist and accessible for other developers to utilize. The top 10 “public” models are listed below. Companies using neural networks might develop their own proprietary algorithms or variations of the ones listed below.
- Convolutional Neural Networks
- Recurrent Neural Networks
- Long / Short Term Memory
- Gated Recurrent Unit
- Hopfield Network
- Boltzmann Machine
- Deep Belief Networks
- Generative Adversarial Network
(For a more detailed description of each model, visit the page on Codementor.io).
Machine Learning vs. Coronavirus
Ever since the outbreak of the COVID-19, various companies have been adopting machine learning techniques to spot outbreak patterns, diagnose new infections, and find weaknesses in the structure of the virus—all to stop and destroy active infections.
Early Warnings of Outbreaks
Who’s the first person to warn about COVID-19? Dr. Wen Liang. What’s the first system to warn about it? That would be BlueDot. Their health monitoring platform analyzes news reports to detect patterns of possible outbreaks. One week before the World Health Organization and the CDC alerts the world to COVID-19, this firm’s platform sends an email to clients warning about the virus. BlueDot’s platform even analyzes airline ticketing data and successfully predicts the spread of coronavirus from Wuhan to the rest of China, Bangkok, Seoul, Taipei, and Tokyo.
Another company is perfecting technology to detect outbreaks as well. Called Nanox, their technology is being used to diagnose unsuspecting medical patients who may unknowingly show signs of coronavirus. This allows medical professionals to act quickly to isolate the patient and start treatment to slow the effects of the virus.
China is utilizing machine learning to diagnose possible coronavirus cases. Wuhan’s Zhongnan Hospital and 34 other hospitals in the country have employed software by Infervision to detect for early signs of COVID-19. So far, their AI platform has reviewed 32,000 cases and it’s being evaluated for use in Europe and the United States. Wired.com has more details about how the technology is being implemented.
Machine learning is beginning to be used for estimating the chances of patient survival. A recent study produced by multiple scientists surveyed an AI system that was able to predict the severity of the virus progression within individuals. Training data for the machine learning algorithm is based on clinical information from 300 patients in the Wuhan region.
Destroying COVID-19 and Curing Infections
Efforts are underway to find ways to stop the transmission of COVID-19 and cure individuals who carry the infection.
BenevolentAI and Imperial College London used AI to discover that an existing medicine may already provide relief from coronavirus, called baricitinib. It was theorized that the medicine could disrupt the microscopic receptor used by the virus. However, it has never been utilized as a vaccine.
Another company, Insilico Medicine, made an announcement about their AI algorithms designing 100 molecules capable of stopping the virus from replicating in the body. According to the article, CEO Alex Zhavoronkov described using separate AI networks (aka neural networks) based on the known structure of COVID-19.
The most promising development comes from California. Inovio Pharmaceuticals claimed to have used its algorithms to find a cure—three hours after China shared public research details about the virus. The company is pushing the FDA to approve a DNA-based vaccine that doesn’t rely on the virus at all (unlike traditional immunizations, which depend on using inactive or weakened pathogens from a microorganism). Optimism is high since human trials have been successful so far.
Open Data Collaboration
Tim Churches, a Senior Research Fellow at the UNSW Medicine South Western Sydney Clinical School, published a tutorial about using the R programming language to conduct COVID-19 epidemiology analysis. He covers the best sources to use as data sets, how to clean the data, ways to scrape data from the coronavirus WikiPedia page, and how to use different R packages to data analyze and create data visualizations.
Alphabet’s DeepMind has also released data about AI insights about COVID-19’s structure. Its goal is to aid researchers in finding a cure for the virus and other treatment options.
What’s Winning: Machine Learning or Coronavirus?
As of this writing, coronavirus is winning. Quarantines have been announced in Italy and sustained in China; Washington and New York in the United States. Cases grow in California and other locales on the U.S. east coast.
Machine learning is emerging as a tool in the battle against COVID-19. Companies and governments are deploying technologies using ML techniques to predict, detect, and find a cure for the virus. Hopefully AI will help make coronavirus extinct within the next several months.