I'm having issues scrolling on your site
Building a Music Recommender with Deep Learning
51–60 of 74 posts
Re: Building a Music Recommender with Deep Learning
#52Warning: this comment has little to do with the article, beyond being a rant on the approach taken by all recommendation engines I've seen. This an interesting approach, but the objective is similar to most recommendation engines: "Find me something similar to something I like". Sometimes that's a good requirement (e.g. when trying to queue up the next song in a playlist, it's good to have some similarity to the song…
What you complained was exactly what I would complain about Spotify's suggestions some time ago. But as of, 3-6 months ago those daily mixes started putting some really interesting new songs that I wouldn't find otherwise. Sometimes it seems to go back to that "safe zone" but it's been such a much better experience I have been telling all my friends to try it. I really would like to know more about their process to i…
Couldnt' agree more. Spotify seems to be solving this problem. No other rec engine I've seen is as good at finding artists I've never heard of, who I really dig
Re: Building a Music Recommender with Deep Learning
#53Highly recommended further reading:
http://benanne.github.io/2014/08/05/spotify-cnns.html (Recommending music on Spotify with deep learning) uses CNNs trained on spectrograms + similarity data from collaborative-filtering to predict per-song vectors.
Re: Building a Music Recommender with Deep Learning
#54I'm having issues scrolling on your site
The page uses https://github.com/galambalazs/smoothscroll-for-websites, which is terrible IMO. It hijacks left/right swipe in Chrome, for no real benefit.
Re: Building a Music Recommender with Deep Learning
#55Music recommendation is a relatively easy problem on one level, and a huge problem on another. If you are recommending music to a neophyte of a certain genre, we've clearly been able to do this for awhile in a way that has real value. But if you're trying to recommend music for someone who is an expert/aficionado of a certain genre, this inevitably annoys that sort of person. For the 2nd type of recommendation, it's hard to provide results of actual interest. Instead, you wind up getting recommendations for pale imitations of things you like. The 2nd problem might require something close to hard sentient AI to accomplish.
Re: Building a Music Recommender with Deep Learning
#56My understanding of convolutions is that it's a way of extracting patterns from images. To convert audio into an image and then create convolutions from that seems... convoluted, if you will. I imagine a better way would be to think of what the equivalent of a convolution would be in the audio space? I.e. noise detection, treble/bass filters, etc.?
Convolution is generic signal-processing. It's quite common to use a one-dimensional convolution for audio filters, it would work perfectly fine as a bass filter for example.
However, 2D conv+maxpool is an image processing technique that gets you translation invariance. Fine for the time dimension of the spectrogram, but rather dubious for the frequency axis. Surely you'd want to distinguish if some feature happens at a high or low frequency?
Re: Building a Music Recommender with Deep Learning
#57This is interesting. My first thought was to wonder how a LSTM would do. Once might think it would be a better representation for music? There's some models which use convolutional layers along with a LSTM for video representation (eg [1]) and it would be interesting to see if convolutions are useful for capturing similar themes of music. I wonder if one could build a music embedding (word2vec style) and use similari…
An architecture like WaveNet could also be interesting here: https://deepmind.com/blog/wavenet-generative-model-raw-audio... (HN thread: https://news.ycombinator.com/item?id=12455510)
Re: Building a Music Recommender with Deep Learning
#58Very cool! One minor nitpick -- the author mentions that this is 'completely unsupervised'. It's true that the author didn't need to manually classify the data, but someone did. So, I believe that this is actually supervised learning, as the author is training a classifier on preexisting labels (the genres). I believe that unsupervised learning would not make use of a target variable at all. If the network architectu…
You're correct of course. But it's cool that you can learn a useful embeddings (in this case into a 128-dimensional space) with only relative few (in this case 9) binary labels.
Re: Building a Music Recommender with Deep Learning
#59Wow thanks for sharing + reading my blog post! I did this for my final project on the Data Science bootcamp at Metis [1] this spring. [1] https://www.thisismetis.com/
This is a really cool project. The hardest part of DJing is knowing which set of songs have similar sonic profiles, and would mix well together. I would love to see this put to use in personal music collections, or in a Traktor playlist, and be able to sort songs by their similarity.
Re: Building a Music Recommender with Deep Learning
#60This is interesting, but fairly easy to confuse. Esp. would be interesting to see what results come up when you use modified "artistic" spectographs like that of Windowlicker by Aphex Twin [1]. One thing I've learned from years of having worked with audio and images is that image representations of audio are horrible representations of it (other than for temporal changes). The results are good though! Good work! :D […
I'm curious to hear more about why image representations of music are horrible. What are the problems or limitations? Is there a better way to perform a similar kind of "dimensionality reduction" on music?