correspond to 31034 MSD songs, an average of 1.45 midi files per song

should be msd_Id

The downbeat is the first beat of the bar,

This looks to me like the same as above.

root, subdir 1, subdir 2 = third, fourth, fifth character of msd_id (after 'TR')

If there is a C major key signature, most likely it was set by default and it is wrong.

That is: key signatures provided as meta messages.

I understand this meta message has two data bytes, but the second byte (minor, major) could have been replaced by a bit (as it may either be 0 or 1 only). Couldn't it have been defined as first data byte = root note, second data byte = one of seven modes (ionian, dorian, etc)?

This is by design of the MIDI specification

I understand each instrument should correspond to one channel in one track, as long as the channel doesn't change program. Correct?

Why does it need the numerator for this? Shouldn't the denominator be enough?

These first two are the same than last two

sections: large variations in rhythm or timbre, e.g. chorus, verse, bridge, guitar solo, etc.

0 or 1, minor or major

This link is broken. Go here instead: http://docs.echonest.com.s3-website-us-east-1.amazonaws.com/_static/AnalyzeDocumentation.pdf

But "Claro" offered in December 2018 data plans which only allowed access to Facebook, Instagram, Whatsapp and Snapchat.

Yeah, but don't they also do it to save money? If they favor services hosted in their own networks (such as a Netflix CDN installed in their facilities), aren't they saving money by not paying peer or transit?

But they both charge you based on how much you use. ISPs usually don't do that, or their transfer quotas are quite high.

Figure is missing

What does she mean?

Doesn't it make some sense that long distance calls cost more than short distance? The same way driving a longer distance means paying more tolls?

What are the real implications (poverty reduction, etc) of this delay in economy growth?

This reminds me of Mickey's broom in Disney's "Fantasia". Can I write "Mickey"? Or will I be sued by Disney Corporation?

Would this change if bpm changed through meta message?

This is called "ticks per beat" in Python's mido.

This seems to have been first implemented by Tim Berners-Lee, when he was trying to popularize his World Wide Web and Web browser (according to himself, in his "Weaving the Web" 1999 book).

Researcher working with vulnerable communities (as mentioned here, cited above).

Are they "copying" the websites?

Naturally, as she wouldn't know someone was commenting about her article in Twitter, reddit, etc, if she didn't know those sites existed.

"the" digital version? I'd rather say "a" digital version

catch-22: paradoxical situation from which an individual cannot escape because of contradictory rules or limitations.

Pero se divulgaron imágenes suyas? O de su ex esposo?

Why?

Why unspoken? Couldn't they be advertised in the group's page to make sure everyone is aware of them?

Open Knowledge Foundation

not suitable in public site areas

see https://web.hypothes.is/help/annotating-with-groups/ and https://h.readthedocs.io/projects/client/en/latest/publishers/config/#cmdoption-arg-services

toolbar buttons color is white on white background in slashdot

Fast Fourier Transform

It's talking about banks! Not presets :)

such as Greasemonkey?

I'm still not sure what the WebExtensions API is, but this may have been sent before its (widespread?) use

And what happens with versions before acceptance but after peer-review has started?

O creando servicios cortos entre Belgrano o Núñez y Retiro, pero harían falta o bien un playa de maniobras, o bien un cambio de vías

Este sitio tiene un problema. Calcula como año de menor popularidad el inicio del dataset. O sea, si hay varios años con cero registros para el nombre, devuelve el más antiguo como año de menor popularidad.

equivalente al tercer grado de la escala en la que vengo

I think we could explain how to do this using open source alternatives, such as gImageReader with Tesseract.

binds the "change" event handler to an @Output() or DOM event (?)

I understand this goes undefined (which is 'false') when i goes beyond the last element in the array

downgrades python to 3.5 and then psychopy fails because it needs >3.6. These can be installed with pip instead

Ex director del Museo Histórico Nacional. Habría sido destituido y luego restituido por denuncias en su contra. Finalmente en 2005 no habría sido renovado en su cargo.

dice "razas", ignorando el consenso mundial de hablar de "etnias"

free choice perception correlating with happiness. How does this relate to Erich Fromm ideas in "The Fear of Freedom"

http://www.memoria.fahce.unlp.edu.ar/tesis/te.1403/te.1403.pdf

La ONG se llama (o llamaba) "Nutrir la vida". No entiendo por qué en el artículo se sugiere que el nombre fue ocultado a drede en la tesis.

I was told the same thing was happening in Amsterdam: people buying houses just to rent them for tourism.

It returns a collection of paths. Each path in the collection describes the shape of the marker. For example, if the marker is a circle, the path would be that returned by mpl.path.Path.circle() (with some affine scaling if applicable). The collection of paths also has an offset value which describes where each path (marker) will be drawn.

I guess openvpn and http server can't be run on the same computer? or it can be forwarded to a different port?

url is https://speech.googleapis.com/v1/speech:longrunningrecognize if audio is longer than 1min: https://cloud.google.com/speech/reference/rest/v1/speech/longrunningrecognize

Seems to fail if environment variable GOOGLE_APPLICATION_CREDENTIALS is not set: https://cloud.google.com/docs/authentication/getting-started#using_a_service_account_in_your_application. Seems to work as well if run as gcloud auth print-access-token (without "application-default").

to pass your authentication as a bearer token in an Authorization header

not 58th Annual Meeting; but 3rd International Meeting

it should be false positive rate

it's true positive RATE

Cross-validation does not return any model with parameters tuned. It does, however, give a measure of out-of-sample (generalization) accuracy. It does so by averaging over several random partitions of the data into training and test (validation) samples.

According to Wikipedia, a random subset of features is chosen "at each candidate split in the learning process"

According to http://scikit-learn.org/stable/modules/ensemble.html#random-forests, ONE subsample would be used per decision tree. I can do bootsrap sampling without random forests: http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.BaggingClassifier.html

To pass the same seed to each DecisionTreeClassifier; https://datascience.stackexchange.com/questions/16800/why-max-features-n-features-does-not-make-the-random-forest-independent-of-num

"simply" training many trees; i.e. without selection of random subset of the features

The evidence p(x) can be evaluated (using the sum and product rules) in terms of likelihoods (probability of data given parameters) and priors (prior believes about parameters): \(p(x) = \sum_{i} p(x|\theta_i)\)

There must me a difference between "thinking thay they [Mr. Jovic as President of the Presidency of Yugoslavia] could help them in some way", and "being left [by them] in no doubt that Belgrade would help them".

See BBC's Death of Yugoslavia chapter 2: Road to War, 9m20s, here with spanish subtitiles: https://www.youtube.com/watch?v=BM5ewil4lws

Wikipedia's article on Geoffrey Nice: deputy prosecutor at the trial of Slobodan Milošević in The Hague and initiated the prosecution's initial case of linking atrocities committed in the former Yugoslavia to Milosevic

Cada non-zero entry es una palabra del diccionario (o feature) -una de las dimensiones del espacio vectorial- seguido de un valor proporcional al número de veces que la palabra aparece en uno de los documentos, pesado por la frecuencia de la palabra en el corpus (palabras más frecuentes tienen valores más bajos).

Observar que no usó una stoplist para ignorar palabras comunes. Probablemente, éstas perderan peso a causa del vector space model algorithm tf-idf.