# <center>Computer component reacts similar as nerve connections in our brain.</center>

<br>

**Researchers have developed a computer component that works similar to the synapses in our brain. Depending on the number of optical data, the "photonic synapse" adapts its crystal structure and thus its conductivity. As with human learning, much-used solutions and ways of thinking gain more weight. The "learning" computer synapse could be an important step towards computers based on the brain.**

<br>

<center>![photosynapse.jpg](https://steemitimages.com/DQmYbg2rq5CVfCUrqAMTkJKQJG98HNWDwoqwQJmKHaHk2Zy/photosynapse.jpg)<br><code>Researchers have designed a chip device that changes like the synapses of the brain depending on the amount of signal. © Harish Bhaskaran</code></center>

<br>

Our brain is still superior to computers in many respects - and differs fundamentally from them: While in computing calculation takes place in the processor, and thus separate from the data store, our thought organ works integratively: the interplay of the neurons over their synapses and the plasticity of these connections are calculator and data storage in one.

>This unique structure makes the brain so energy efficient, says Harish Bhaskaran of the University of Oxford and his colleagues. The development of computers that work more like our brain has therefore been a kind of holy grail for decades.

<br>

### How do you make a computer plastic?

<br>

One way to mimic the plasticity of the brain is neural networks. In these, algorithms simulate the adaptation of the synapses to what they have learned: If a solution is found to be advantageous, this computing path will be preferred in the future. So far, however, this imitation of neural computation extends solely to the software. But there have also been initial attempts to replicate the synapses of our brain technically.

Now, Bhaskaran and his colleagues are making an exciting breakthrough here: They have designed a photonic synapse that can be integrated into optical computer chips. Similar to their biological model, this synapse adapts itself to how often and intensively it is used.

<br>

### Crystal varies according to light flow

<br>

<div class="pull-right">

<center>![photosynapse2.jpg](https://steemitimages.com/DQmNbZtbkHbiKtamzkUd8oviXMW6gX8xLPAYGzst5bZKrvR/photosynapse2.jpg)</center><br><code>Instead of the neurotransmitters and receptors of the biological synpase, discs are made from special material for plasticity. © Harish Bhaskaran</code>
</div>

The novel photonic synapse consists of a so-called phase change material ([PCM](https://en.wikipedia.org/wiki/Phase-change_material)) - a chemical compound that changes its aggregate state even with small changes in energy supply. Tiny discs of a germanium, antimony and tellurium compound (Ge<sub>2</sub>Sb<sub>2</sub>Te<sub>5</sub>) form the core of the synapse. In the rest state, this material is crystalline. If optical signals in the form of laser pulses travel through this material, its crystal structure changes.

This in turn affects how well the signals are propagated - in a controllable manner: 

>We see that the changes in synaptic weighting are exponential and monotonous on the number of incoming light pulses, according to the researchers. In this way stage "2" of the synapse is always achieved by applying 100 pulses of 50 nanoseconds duration - regardless of whether the component was previously in stage "1" or "3".

<br>

### Adaptation like a real synapse

<br>

>By simply varying the number of optical pulses, we have an incredibly simple and effective approach to creating systems with a continuously variable synaptic plasticity, Bhaskaran and colleagues say. This is because the photonic synapse is more effective and more conductive as the connection points in our brain become more effective.

Another advantage is that the photonic synapse can be easily integrated into optical computer chips. As it is controlled by light pulses as well as transmitting optical information, it is even a thousand times faster than our nervous connections in the brain.

<br>

### A way to a new era of computing?

<br>

According to the researchers, components such as their photonic synapse could pave the way for a whole new era of computers based on the brain. 

>Every computer of this kind has to be able to replicate the function of the synapses - and that's exactly what we have achieved here, says co-author Wolfram Pernice from the University of Münster.

His colleague, David Wright, from the University of Exeter, says, 

>Electronic computers are relatively slow and the faster they are, the more current they swallow. In addition, none of these computers has the learning capability and parallel processing of the human brain. Now, however, we begin with both of these points: We not only develop brain-like computer architectures, but also work with optical signals to save time and energy.

<br>

Source: [Science Advances, 2017, doi: 10.1126 / sciadv.1700160](http://advances.sciencemag.org/content/3/9/e1700160)

<br>

<center>https://steemitimages.com/DQmRhDtjokAZnGKi4QwheqksKTFo6m4fsjMYsNNrsitC1xk/U5dsnt2Y19qmNJ6S4rHWGSU4yi4wucH.gif</center>