Machine Learning and Patterns

By Pavithra Brahmananda Reddy

Machine learning is quite the buzzword these days. In simple terms, it’s used to predict what kinds of ads you should see, the products you would be interested in buying, and also what movies you would like to watch. There is seemingly no end to the list of applications that machine learning can have today.  

People keep finding new ways to make use of machine learning techniques in various ways. Recently, an article, “Chimpanzee face recognition from videos in the wild using deep learning”, by Schofield et al., detailed a way in which scientists have been further studying machine learning.  

Machine learning allows computers to perform tasks by identifying patterns from data sets. Their model was intended to use facial recognition technology to help the researchers “identify and track” chimpanzees that appear in video footage. For their dataset, they used annotated video footage of chimpanzees from Guinea that had been archived over a span of 14 years.  

This kind of machine learning is called supervised learning. Supervised machine learning is a type of machine learning in which the training set consists of labelled examples. The researchers’ model uses neural networks to achieve their goal. Neural networks are generally used to identify patterns in data sets (here, to differentiate between apes). The name comes from the fact they are modelled after the working of the human brain. In this case, the algorithms they use are called convolutional neural networks. These algorithms use different aspects of visual data to classify and distinguish between millions of images. The model is trained using visuals of chimpanzees in the wild. The visuals were labelled with chimpanzee identities. After the training is complete, the model will be able to identify chimpanzees based on the training data. The model used in the study helped researchers classify data(chimpanzees) into different categories(identities) based on specific examples the model was trained with.  

Upon testing the model under different conditions, researchers came to the conclusion that the model worked very well in identifying the apes even as their features changed with their advancing age. They also discovered that model was fairly generalizable. The rest of the article delves deeper into the implementation and training of the model. This model could be instrumental in the analysis of video footage to study animals. While the article discusses one application, convolutional neural networks could be applied to analyze other kinds of visual imagery. Neural networks, in general, have a wide range of applications. At the College of Wooster, neural networks are a topic of interest in research and several students make them a part of their Independent Study projects. 

If you have heard of machine learning but don’t know what exactly it is, this article could be a good way to ease yourself into it. The article was an interesting read because it reminds us about the various applications machine learning can have and how much of an impact it continues to have on the world around us. 

The article can be accessed at the following link: