Artificiële neurale netwerken veranderen de wereld. Wat zijn ze?

[Nr.10]

Artificial neural networks are changing the world. What are they?
_{12 okt 2015}

(...) a whole new type of computer, one based on the structure of the brain.

Een computer wiens werking gebaseerd is op de werking van onze hersenen.

When you ask yourself, “Can an artificial neural network do it?”
immediately after, ask yourself “Can I do it?” If the answer is yes,
then your brain must be capable of doing something that an ANN might one day be able to simulate.
On the other hand, there are plenty of things an ANN might one day be able to do that a brain never could.
The potential for ANNs is nearly limitless.

[Nr.10]

Neural Networks Rival the Primate Brain at Object Recognition
_{20 jan 2015}

A central problem in computer vision is that of designing algorithms that can competently accomplish tasks that humans can easily do, such as recognizing objects. Recently, neuroscientists at MIT found that the latest artificial deep neural networks (DNNs) are able to rival the performance of the primate brain at this task. (...)

(fig.: neurale netwerken en ML-algoritmen, gebaseerd op de werking van de hersenen)

[Nr.10]

Om te lezen:

Using neural nets to recognize handwritten digits
_{jan 2017}

[Nr.10]

Over het menselijke brein:

Het menselijke brein
Onze hersenen tellen 85 �* 86 miljard neuronen, volgens wikipedia.
Elk van die neuronen is verbonden met 10.000 andere neuronen.
Signalen tussen de neuronen worden uitgewisseld via synapsen.
Er zijn naar schatting 1.000 triljard synapsen bij een volwassene.
Bij jonge kinderen tien keer zoveel: 10.000 triljard.
De synapsen worden geacht de geheugenopslagplaatsen te zijn.
Het menselijke geheugen wordt geschat op 2,5 miljoen GByte.
Het gemiddeld volume van de hersenen = 1.450 cm³.
De energie die verbruikt wordt door de hersenen ligt ergens tussen de 10 en de 20 W.

Over SyNAPSE:

SyNAPSE is a DARPA program that aims to develop electronic neuromorphic machine technology
that scales to biological levels. More simply stated, it is an attempt to build a new kind
of cognitive computer with similar form, function,
and architecture to the mammalian brain.
https://en.wikipedia.org/wiki/SyNAPSE

SyNAPSE staat voor:

Systems of
Neuromorphic
Adaptive
Plastic
Scalable
Electronics

[Nr.10]

Op deze chinese site vind je een zeer goede, eenvoudige uitleg over hoe je een artificiëel neuron / artificiëel neuraal netwerk kunt omschrijven.

http://bbs.pinggu.org/thread-4192156-1-1.html

So what is a neural network? Let’s wait with the network part and start off with one single neuron. A neuron is like a function; it takes a few inputs and calculates an output. The circle below illustrates an artificial neuron. Its input is 5 and its output is 1. The input is the sum of the three synapses connecting to the neuron (the three arrows at the left).At the far left we see two input values plus a bias value. The input values are 1 and 0 (the green numbers), while the bias holds a value of -2 (the brown number). The inputs here might be numerical representations of two different features. If we’re building a spam filter, it could be whether or not the email contains more than one CAPITALIZED WORD and whether or not it contains the word ‘viagra’. The two inputs are then multiplied by their so called weights, which are 7 and 3 (the blue numbers). Finally we add it up with the bias and end up with a number, in this case: 5 (the red number). This is the input for our artificial neuron.

[Nr.10]

The neuron then performs some kind of computation on this number — in our case the Sigmoid function, and then spits out an output. This happens to be 1, as Sigmoid of 5 equals to 1, if we round the number up (more info on the Sigmoid function follows later). If this was a spam filter, the fact that we’re outputting 1 (as opposed to 0) probably means that the neuron has labeled the text as ‘spam’.

[Nr.10]

If you connect a network of these neurons together, you have a neural network, which propagates forward  —  from input output, via neurons which are connected to each other.

[Nr.10]

The Sigmoid function simply maps your value (along the horizontal axis) to a value between 0 and 1.

[Nr.10]

In november 2016 werd bij Google Translate
overgeschakeld van PBMT naar NMT.
Onder meer voor het belangrijke talenpaar Engels/Chinees.
In Google Translate zaten op dat ogenblik 103 talen.
Dit wil zeggen 103 in het kwadraat talenparen.
103 x 103 = 10.609 modellen
Die modellen worden allemaal met data gevoed.

• PBMT = Phrase-Based Machine Translation
• NMT = Neural Machine Translation

Neural Machine Translation gebruikt ANNs,
Artificiële Neurale Netwerken.

The Artificial Intelligence Behind Google Translate Recently Did Something Extraordinary
_{28 nov 2016}

[Nr.10]

Baidu Translate: The Inside Story
_{24 mrt 2016}

“Baidu Translate is based on deep learning technology,
a completely different method [from MOSES], the most popular
open source phrase-based machine translation system” [= PBMT]

“Deep Learning technology
performs much better
than PBMT” (Baidu Translate)

Confirmed: Deep Learning Is Coming to Google Translate
_{17 mrt 2016}

Google confirmed they plan to improve Google Translate’s accuracy through artificial intelligence called deep learning. Deep learning is an advanced model of machine learning where an algorithm takes what it has already “learned” (data previously processed) and uses it to form new ways to solve problems in a pattern.

[Nr.10]

In de eerste helf van 2014 maakte Andrew Ng de overstap van Google naar Baidu.

Why a deep-learning genius left Google & joined Chinese tech shop Baidu (interview)
_{30 jul 2014}

Baidu kondigt Deep Speech aan:

Baidu claims deep learning breakthrough with Deep Speech
_{18 dec 2014}

[Nr.10]

Wit.ai uit de tabel hierboven wordt omschreven als een voice recognition API provider.

Een filmpje om dit te demonstreren:

video
_{10 feb 2014}

Website:

https://wit.ai/