Every Tuesday in April: Free Intro to Python Training Series for Business Professionals Register Now

Made at Metis: Two Music Lovers Build Neural Networks

By Emily Wilson • November 20, 2017

This post features two final projects created by recent graduates of our data science bootcamp. Take a look at what's possible in just 12 weeks.

Susan Fung, Ph.D.
Metis Graduate

When recent graduate Susan Fung says that Justin Bieber and Neil Young helped her learn about neural networks, she means it. She wanted to explore neural networks for her final bootcamp project, and as a music fan/neuroscientist, she thought to herself, what better way to do so than by training a neural network using Justin Bieber and Neil Young lyrics? 

"As a neuroscientist, it was too good an opportunity to pass up," she wrote. "What I wanted to do was generate new lyrics based on an artist's style. So at the highest level, train a neural network (NN) on a corpus of Artist X's songs, plug in a seed phrase, receive new text." 

Curious how she got the job done? Read about the entire process on her blog here. She writes about the tools and methods she used throughout, the technical process, and the results, including notes about future work. 
_____

Matt Murray 

Metis Graduate 

For his final project, graduate Matt Murray created a song recommendation engine with deep learning using Juno Download, a digital download website predominantly used by DJs that has a huge back catalog of tracks for sale. On his blog, Matt writes, "It’s a great music resource and they provide a generous 2-minute sample MP3 file for each song they have for sale. The only problem is…it’s really hard to find music on the site that isn’t a new release or currently top of the sales charts."

While it makes sense that newer music generates the most revenue, Murray's curiosity led him to wonder about the other 99% of musical content on the site. It was hard to find – so how could it be made easier? He felt that the site was missing a content-based "you might also like"-type recommender.

So he started downloading songs and converting them into spectrograms. After that, he trained a CNN (convolutional neural network) on the image data. "I needed to teach it to recognize what the different types of music ‘looked’ like in the spectrogram images, so I used the genre labels and trained it to identify the music genre from the images," he wrote. 

And that's just the start. Visit his blog to read a detailed post on his project. 
_____

Want to learn more about the data science bootcamp? Check it out!


Similar Posts

alumni
Bootcamp Grad Finds a Home at the Intersection of Data & Journalism

By Emily Wilson • July 03, 2019

Bootcamp graduate Jeff Kao knows that we’re living in a time of heightened media distrust – and that’s precisely why he relishes his job in the media. “It's heartening to work at an organization that cares so much about producing excellent work,” he said of the nonprofit news organization ProPublica, where he works as a Computational Journalist. Read Kao's full story here.

alumni
The Trifecta: From Bootcamp Prep Course to Bootcamp to New Career

By Emily Wilson • March 05, 2020

For Kari Davis, there’s before the bootcamp, and there’s after. A clear demarcation between two paces of professional life – one slower and more bureaucratic, the other fast-paced and constantly changing. Learn how a Metis Bootcamp Prep Course helped her prepare for the bootcamp and then land a new data science job.

alumni
Made at Metis: Smarter Pricing for Airbnb & a Karaoke Recommendation Engine

By Metis • February 26, 2020

This post features two projects from recent graduates of our data science bootcamp. Take a look at what's possible to create in just 12 weeks, including a project to improve Airbnb's smart pricing tool and another to enhance a karaoke app by creating an added recommendation engine.