Latest in Machine Learning

Predicting Earthquakes, Teaching Robots, and more!

Team Research Nest
Jun 2, 2018 · 5 min read

In the modern era, machines undoubtedly play a crucial role in everyday life. Machine learning is no ordinary branch of science. It dominates and influences every aspect of our everyday life. Here is an assortment of the top 5 research and discoveries in machine learning in recent times, presented crisp and concise.

Can we predict human behavior?

Deep learning is a subset of machine learning. In a deep learning system, algorithms are adept at concluding on their own, if the prediction is accurate or not. In an interesting experiment, A deep learning based Predictive Vision system was built at Massachusetts Institute of Technology, which learned to predict actions, by analyzing behavior of humans on YouTube videos from TV shows like “Desperate Housewives” and “The Office”. It was thus used to conclude whether two persons will hug, kiss or slap a five. The model was trained on 600 hours of video and was able to predict an activity accurately 43 percent of the time. This is quite high compared to previous algorithms, which could only produce an accurate prediction 36 percent of the time. May be in future, we will be able to predict all those mood swings and other emotions and take necessary precautions before hand to sustain successful relationships among one another. Truly, an interesting application of Machine Learning!

Predicting an Earthquake using Machine Learning!

Just take a moment to imagine how revolutionary this could be! How many lives can we save? How much damage can be avoided?

Researchers at Columbia University, have developed a ML model using earthquake datasets, that can help in the prediction and control of seismic activities. The dataset contained three-years of seismic recordings from “The Geysers”, a geothermal reservoir in California. This model is crucial for future energy production and embellishing the injection levels, which can lead to considerable gains in total energy outputs from geothermal sites.

Scientists can utilize the data to manipulate and boost injection levels to the safe maximum, that would lead to more cracks in the crust below a geothermal reservoir — and release more energy. The team also thinks their research could be applied to better model naturally occurring earthquakes. By applying machine learning to major earthquakes and Tsunamis in the past, seismologists be able to predict the next catastrophic quake.

Self-driving technology is expanding!

Self-driving is fundamental for the automobile industry and ensuring that autonomous driving is the most amusing, easy and safe way is what motivates us to explore the barriers of creation. Land Rover in association with a group of researchers in the University of Birmingham and Myrtle AI, aims to launch self-driving “off-road” vehicles, that can drive irrespective of the weather and terrain, that includes-ice, fog and rain, however steep, narrow, uneven or extreme. The enthusiastic project has been declared with a fund of £3.7 million and the main aim of this work is to develop the world’s first 5-D technology. The 5-D will combine audio, visual, radar, light detection and distance sensing (LiDAR) technologies with machine learning to help the vehicle advance over time. This empowers the vehicle to behave in a more sophisticated way, enabling it to handle any weather condition on any terrain.

Stock market prediction:

A research group used Azure Machine Learning Workbench to predict stock market performance, based on publicly available earning documents. Natural Language processing was employed to decipher the earnings, by removing stop words, punctuation and other ephemera. The system found the connection between language of the releases and its impact on stock price. Despite facing vocabulary issues throughout the training process, the team solved this drawback by comparing stocks on an industry-by-industry basis. The stocks were sorted into three groups-high, middle and low performing. The most accurate result was a 62% chance of determining a low-performance stock. This implies that there is plenty of room for improvements in the accuracy in the future.

Teaching chores to a Robot!

Teaching a machine fundamental human activities is much more complex than it looks. For instance, to teach a robot to bring you a glass of water, it has to not just detect the instruction to fill the water into a glass, but also to comprehend the many small facets of the activity that the human brain infers — like, walk into the kitchen, open the cupboard, and take a glass. Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory, in association with researchers at the University of Toronto have designed VirtualHome, a virtual 3D environment which can exactly teach robots how to complete simple tasks like filling water. In the near future, the team hopes to program the bot to perform specific tasks by watching a YouTube video. This technology could be increasingly useful for the elderly or those who may have limited mobility.

Influential advancements are being made in the field of machine learning, which is changing the world in a fast pace. In the near future, you can foresee the automation of literally every facet of your house.

(This Article was authored by Research Nest’s technical writer, Nivedha Jayaseelan)

The Research Nest

Team Research Nest

Written by

The Research Nest

Empowering humanity with exclusive insights.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade