The Input table shows the tracks you
select as seed data to generate a recommendation
list. The engine is designed to work with
a very small number of input tracks -- often
as few as two to four tracks will produce excellent
results. To capture subtler musical characteristics
than genre and period, use a few more tracks
and experiment with the algorithm parameter
sliders in the middle of the page.
Weight is a measure of the influence that
each song has relative to the others in an input
set. Exposing this information enables you to
adjust for the naturally uneven distribution of
music relationship data, either by adding multiple
weak tracks with a similar flavor or removing an
overly strong track that may be distorting your
The Golf column contains icons for moving
tracks into and out of the input track set, and
for playing a sample (if available). The term
"Golf" refers to seeking the lowest number of
strokes in golf. Playlist Golf is designed to
help you generate interesting recommendation
sets with the lowest number of input tracks.
The Input Focus parameter allows you to
tune the algorithm to be more or less obsessive
in adhering to the theme of the input tracks.
Higher settings will produce more tightly focused
recommendations, but may become too constrained.
Lowering the value will give the engine more
freedom to explore. If you get no results at
all, lowering the Input Focus parameter may
The Breadth of Appeal parameter is difficult
to characterize. Roughly speaking, higher values
cause the engine to look at a broader range of
musical tastes when considering your input tracks.
The effect varies a great deal from one set of
input tracks to another; you can get the best
results by trying some different settings to
zero in on what works for your input tracks.
The Mainstreamness parameter adjusts
how heavily the engine weights the popularity
of recommendation tracks. Higher values will
give you the chart topping hits, while lower
settings will delve into more obscure tracks.
The Recommendations table shows the tracks
that the engine thinks best match your input
tracks using the algorithm tuning parameters
you have set. It will automatically reload when
you add or remove input tracks, or when you
adjust the tuning.
The Score column shows how confident the
engine is in each selection it made, relative
to the others that it found for your input tracks
and tuning parameters.
When the engine can't find any results, there
are a few ways to fix it. Lowering the Input
Focus parameter or removing "Not This"
anti-input tracks should solve the problem.
You can also try narrowing the focus of the
input tracks you have chosen.