WHAT IS THE MILLION SONG DATASET FREE

Feel free to skip it, but it is always here if you need/want it. ** This section is more tutorial than blog.

Which artistsвЂ™ songs are most similar to those of Nirvana? Is Kanye really as unique as he seems to think he is? Okay, We actually canвЂ™t say anything about Kanye as a person, but maybe we can see if his music is unique.

What aspects of music have and havenвЂ™t changed over the 90 years spanned by the data?.

Are specific audio features good predictors of a songвЂ™s popularity?.

There are several questions that one could try to answer using this sort of data:

The information is a mixture of audio features, e.g., loudness and timbre, and metadata, e.g., release year and artist. As the name suggests, this dataset contains information about a million contemporary popular songs.

The next couple of blogs will delve into a dataset that is nearly 300 gigabytes in size, the Million Song Dataset. Thankfully, my research groupвЂ™s computational cluster has 39 terabytes of hard drive space for me to play work with. That being said, my job often requires me to work with simulations that are upwards of 3 terabytes. My hard drive only holds a terabyte, so I wouldnвЂ™t expect this blog to cover datasets larger than a few hundred gigabyte anytime in the near future. However, more data also means more time spent loading, analyzing, and fitting. More training data means more predictive power in any fitted model.

Larger datasets mean more certainty in any analysis. What is the optimal size of a dataset? If you have an unlimited amount of time and hard drive space, then the answer is always вЂbiggerвЂ™.

What is the million song dataset