Видео 24
Просмотров 76 491

1:34

Portia Part V: Spiking Neural Network with frequency-driven Syngen

1:09

Portia Part IV: Spiking Neural Network long-term Evolution

1:28

Portia Part III: Neural Network with evolving topology

3:05

Portia Part II: first real Neural Network, 12500 Neurons

0:58

Portia Part I: testing with Conway's Game of Life

0:57

Portia Part VII: Cortex Dynamics

Hi!
Here you can see how a cortex evolves under input pressure: in the first phase, synapses form and strengthen while inputs from two color channels of a webcam are fed to the input neurons (top left and top right).
In the second phase you can see how a lack of inputs causes weaker synapses to die and stronger synapses to weaken.
The stronger a synapse the lower the possibility of weakening.
Synapses can be excitatory (blue) or inhibitory (orange).
For more info:
Repo: github.com/Mathorga/portia
Github Page: github.com/Mathorga
eMail: michelettiluka@gmail.com

Видео

1:34

Portia Part VI: Input Mapping

Просмотров 322 года назад

Hi! Here you can find the result of Portia's new input mapping algorithm: it works by grabbing the input source as integer numbers, in this case the pixel values of a b/w camera and converting them to spike trains within the specified sampling window. On the right you can see a simplified view of the camera image, rendered in ascii, while on the left is the result as spikes in the net represent...

Portia Part V: Spiking Neural Network with frequency-driven Syngen

1:09

Portia Part V: Spiking Neural Network with frequency-driven Syngen

Просмотров 312 года назад

Here you can see a spiking neural network evolving through frequency-driven syngen, meaning new synapses are generated and deleted according to each neuron's spiking frequency. Source code: github.com/Mathorga/portia

Portia Part IV: Spiking Neural Network long-term Evolution

1:28

Portia Part IV: Spiking Neural Network long-term Evolution

Просмотров 382 года назад

This video show a simple neural network made of 24000 neurons evolving upon external (random) stimulation. In the top left corner of the screen you can see whether the network is being stimulated or not (fed), as well as how long its feeding state lasts, in steps count. Each frame of the video shows 5000 steps for a total of roughly 6.3M steps. The whole simulation was performed in around 6h (s...

Portia Part III: Neural Network with evolving topology

3:05

Portia Part III: Neural Network with evolving topology

Просмотров 332 года назад

Liath evolves a neural network in real time. The network is made of 6000 neurons with 30 inputs. Inputs (highlighted with a white circle) are randomly fed by the sample program. source code: github.com/Mathorga/liath

Portia Part II: first real Neural Network, 12500 Neurons

0:58

Portia Part II: first real Neural Network, 12500 Neurons

Просмотров 422 года назад

First spiking neural network simulation, working very well :D

Portia Part I: testing with Conway's Game of Life

0:57

Portia Part I: testing with Conway's Game of Life

Просмотров 862 года назад

First test for basic functions setup.

Imago: Simple Spiking Neural Network Evolution

4:46

Imago: Simple Spiking Neural Network Evolution

Просмотров 1103 года назад

Demonstration of evolution functionalities of a spiking neural network: neurons with high firing rates are more likely to generate synapses to others, while neurons with low firing rates are more likely to loose their output synapses. In the demo you can see a network of 100 neurons and 12 output synapses per neuron at the beginning. Input neurons are highlighted with a white circles and their ...

2:40

Spiking Neural Network Demo

Просмотров 6413 года назад

A topology-unaware spiking neural network demo showing 80 neurons and 10 output synapses for each neuron. You can see the net starting slow and then quickly increase in activity, as showed by the total number of spikes in the top left corner of the screen. Enjoy!