Machine Learning and Music : Grace and Beauty (part V)
Updated: Jul 1, 2019
'Twas at the royal feast for Persia won By Philip's warlike son— Aloft in awful state The godlike hero sate On his imperial throne; His valiant peers were placed around, Their brows with roses and with myrtles bound (So should desert in arms be crown'd); The lovely Thais by his side Sate like a blooming Eastern bride In flower of youth and beauty's pride:— Happy, happy, happy pair! None but the brave None but the brave None but the brave deserves the fair!
Alexander's Feast; or, the Power of Music by John Dryden
In the previous article we discussed the timeline of the Artificial Intelligence from 1979 to 1981. We are moving forward, good.
In this article we are going to cover the next period of the fascinating history of the AI and music.
1982 : RNN, Stevie Wonder and Olivia Newton-John
In 1982 John Hopfield, an American scientist, started popularizing associative neural network, mostly known as Hopfield networks. Hopfield networks is a type of recurrent neural network that can be applied as content-addressable memory systems with binary threshold nodes. However, it is important to mention that the principle was described earlier by Little in 1974.
In 1982 Stevland Hardaway Morris, better known by his stage name Stevie Wonder, an American singer, songwriter, musician and record producer, presented "That Girl", that was the leading single from Wonder's album-era 1982 greatest-hits compilation. The most listened 1982 pop single was "Physical", a song by British-born Australian singer Olivia Newton-John for her twelfth studio album Physical.
1985: NetTalk, Careless Whisper and Lost in the Fifties Tonight
In 1985 Terry Sejnowski developed NetTalk. NetTalk was a program that learned to pronounce words the same way a baby does.
"Careless Whisper", a pop ballad written by George Michael and Andrew Ridgeley of Wham!, was released on 24 July 1984 on the Wham! album Make It Big and became the most popular single in 1985.
The song features a prominent saxophone riff, and has been covered by a number of artists since its first release. It was released as a single and became a huge commercial success around the world. It reached number one in nearly 25 countries, selling about 6 million copies worldwide—2 million of them in the United States.
quote from Wikipedia
The same year Lost in the Fifties Tonight, the seventeenth studio album by country music artist Ronnie Milsa, became the most listened country single.
1986: Backpropagation, Whitney Houston and On my Own
1986 was definitely one of the most outstanding periods for Whitney Houston.
In 1986, at the 28th Grammy Awards, Whitney Houston received four nominations; including Album of the Year and won one, Best Pop Vocal Performance, Female for "Saving All My Love for You".
Her album called "Whitney Houston" became the most sold album in both Pop and RnB categories. This album eventually topped the Billboard 200 for 14 weeks in 1986, generating three number-one singles.
The most popular RnB single was "On My Own" by Patti Labelle and Michael McDonald.
And what about Machine Learning? In 1986 Seppo Linnainmaa's reverse mode of automatic differentiatio was used in experiments by David Rumelhart, Geoff Hinton and Ronald J. Williams to learn internal representations.
Knowledge representation and reasoning (KR², KR&R) is the field of artificial intelligence (AI) dedicated to representing information about the world in a form that a computer system can utilize to solve complex tasks such as diagnosing a medical condition or having a dialog in a natural language. Knowledge representation incorporates findings from psychology about how humans solve problems and represent knowledge in order to design formalisms that will make complex systems easier to design and build. Knowledge representation and reasoning also incorporates findings from logic to automate various kinds of reasoning, such as the application of rules or the relations of sets and subsets. Examples of knowledge representation formalisms include semantic nets, systems architecture, frames, rules, and ontologies. Examples of automated reasoning engines include inference engines, theorem provers, and classifiers. The KR conference series was established to share ideas and progress on this challenging field.
quote from Wikipedia
However, it is important to mention that this mode of automatic differentiation was first applied to neural networks by Paul Werbos, an American social scientist and machine learning pioneer.
1989 : Q-learning, Bobby Brown and Supewoman
In 1989 Christopher Watkins developped Q-learning, a model-free reinforcement learning algorithm, which greatly improved the practicality and feasibility of reinforcement learning. The goal of Q-learning was to learn a policy, which told an agent what action to take under certain circumstances. It did not require a model of the environment that is why it was named model-free. Q-learning was able to can handle problems with stochastic transitions and rewards, without requiring adaptations.
Another huge breakthrough of that period was the stat of commercialization of Artificial Intelligence on Personal Computers. In 1989 Axcelis, Inc. released Evolver, a software package that allowed users to solve a wide variety of optimization problems. It was the first software package to commercialize the use of genetic algorithms on personal computers that marked the begging of new era in Machine Learning.
Don't Be Cruel, the second studio album by American singer Bobby Brown, was released in the United States on June 20, 1988 by MCA Records. It became the most listened pop album in 1989.
The most listened RnB single of 1989 was "Superwoman" by Karyn White. The single was presented in her debut album in 1989.
In the next chapter we are going to cover the 90s. Hope you enjoyed this.
Hopfield, John (April 1982). "Neural networks and physical systems with emergent collective computational abilities" (PDF). Proceedings of the National Academy of Sciences of the United States of America. 79: 2554–2558. Bibcode:1982PNAS...79.2554H. doi:10.1073/pnas.79.8.2554. PMC 346238. PMID 6953413. Retrieved 8 June 2016.
Rumelhart, David; Hinton, Geoffrey; Williams, Ronald (9 October 1986). "Learning representations by back-propagating errors" (PDF). Nature. 323: 533–536. Bibcode:1986Natur.323..533R. doi:10.1038/323533a0. Retrieved 5 June 2016.
Watksin, Christopher (1 May 1989). "Learning from Delayed Rewards" (PDF).
Markoff, John (29 August 1990). "BUSINESS TECHNOLOGY; What's the Best Answer? It's Survival of the Fittest". New York Times. Retrieved 8 June 2016.