ron with the strongest activation level
(the highest output value).
In the beginning, Mario had to be
controlled by a human player, and the
system would store vector pairs corresponding to the neural network input
(character positions) and the desired
output (human-like behavior). When
enough data was stored, the training
procedure could start and by the end of
this process the weights of the neural
network were fixed, the network would
have learned the right internal configuration needed to emulate a human
player. At this point Mario would leave
the game and Luigi would automatically start to play based on the recently
learned behaviors from the collected
data. We added some graphical representations of this internal weight configuration and its evolution over time,
finally we had a workshop ready to go.
The workshop took place on October
16, 2014. William Raveane, a colleague
of mine working on his Ph.D. with ANNs,
presented neural networks from a theoretical perspective. I developed the practical part of the workshop. At the end,
we were pleased with the results, and I
think our audience was, too. Since then,
we have held workshops on ANNs each
academic year, with more attendees at
If you are interested, the code used
for the workshop and the executable
.jar files can be found here: https://
Daniel López Sánchez is a predoctoral student at
the University of Salamanca. He holds a bachelor’s degree
in computer engineering and a master’s degree in
intelligent systems. Most of his work centers around low
computational cost machine learning, deep learning,
and artificial vision. He is one of the founding members
of the ACM USAL student chapter.
Early 1980s Researchers Paul Benioff, Yuri
Manin, and Richard Feynman independently
investigate computer models that are able to
simulate quantum systems, which are the first
conceptions of quantum computing.
1985 David Deutsch of Oxford University publishes a paper describing
the first universal quantum computer, which
uses “quantum gates” to behave similarly to a
universal Turing machine.
1994 Peter Shor, working at AT&T, proposes an algorithm using qubit
entanglement and superposition to find the
prime factors of any integer. Shor’s algorithm
works in polynomial time, and is thus far more
efficient than any other factoring algorithm at
2001 Researchers at IBM Almaden and Stanford University are the first to
successfully execute Shor’s algorithm to factor
the number 15, using a seven qubit computer.
2016 IBM Research announces the “IBM Quantum Experience”
for the public to work hands-on with a as
a cloud-based system composed of five
superconducting qubit computer.
— Jay Patel
The D-Wave 2X Systems allows for a search of 21,000
possibilities—which is higher than the total number of
particles in the universe—thanks to a lattice of 1,000 qubits.