Work Groups
DVS/DAVIS event camera crash course
I brought along about 3-4 dynamic vision sensors (DVS) and dynamic and active pixel vision sensors (DAVIS) that can be used for independent projects.
Depending on interest, I have planned a set of 4x45m afternoon sessions to show
1. Using these sensors in jAER (viewing, recording, playing back)
2. Controlling ...
Spatio-temporal coding in the retinal circuit
The retina is more sophisticated than many physiologists and neuromorphic engineers think. In the retina, image of real world is first continuously transduced to spatio-temporal distribution graded responses (slow analog voltage changes). In ganglion cells (output neurons of the retinal circuits), the graded potentials are transformed to spikes following some ...
Exploring unsupervised online learning in digital spiking neural networks
In this workgroup, we will explore unsupervised online learning with the ODIN 28-nm digital spiking neural network, if possible applied to real-world examples (e.g., the MNIST dataset).
Compared to analog approaches, digital implementations of online learning rules:
- do not suffer from variations in operating voltage or temperature,
- allow one-to-one ...
SpiNNaker crash course
For New users and experienced users. We will:
1. walk through installations of both the latest release (through pip) and the cutting edge software stack if you feel lucky.
2. Give a quick description of SpiNNaker and its supprot
3. Describe PyNN and the differences between 0.75 / 0.8 ...
Winner take all behavior in continuous rate-based and discrete spiking systems
The goal of this project is to thoroughly characterize winner-take-all (WTA) behavior in 3 different systems and find a correspondence between them. WTA behavior is an ubiquitous (canonical) motive in neural computation and is used in many projects in our group and elsewhere, however, a thorough study of those dynamical ...
Exploring the classification of biomedical signals using plasticity in recurrent networks of ...
Continuous monitoring of physiological biomarkers could help to detect and diagnose several health conditions at their early stages, such as cardiovascular and neurological diseases and as a result real-time processing of biomedical signals in a low-power fashion is becoming increasingly important. Recurrent neural networks (RNNs) are capable of learning features ...
Relational Networks for goal-directed sensory-motor task
We will build a (relatively) simple spiking neuronal network implementing a goal-directed state-to-action mapping (SAG) unit. As a core architecture we will use a three-way relation network proposed by [1]. However, we would like to translate the rate-coding scheme into spike time coding scheme to exploit spatio-temporal sparsity of the ...
GPU-accelerated insect (and other!) brain simulations
The researchers from the University of Sussex at Capocaccia this year are part of an EPSRC-funded project called Brains-on-Board - now in its second year. Within this project we are developing insect-inspired models of vision, navigation etc to run in realtime on robotic platforms powered by hardware such as the NVIDIA ...
Filtering the noise from event-based vision sensors for learning patterns in spiking neural networks
The environment and the event sensor itself produce noise, however, the brain can distinguish and filter the noise before or during the learning. In this work-group, we try to understand in which way spiking neurons and synapses can be organized to filter the noise and build it in neuromorphic electronic ...
Hands-on working group “BrainScaleS-2”
As the next generation BrainScaleS-2 system is now available in prototype form we want to offer a hands-on working group to explore the new features of the next generation chip.
Programmable Plasticity
Besides its analog neuromorphic core the system features a digital plasticity processing unit (PPU) programmable in C and ...
Neuromorphic SLAM
Simultaneous localisation and mapping (SLAM) is one of the core tasks of mobile autonomous robots. Powerful solutions to the SLAM problem exist but fast and low-power implementations are still being sought for. Since even simple animals navigate effortlessly in complex dynamically changing environments, taking inspiration from biology can lead to ...
Spiking Neural Network Transfer to Asynchronous Analog Neuromorphic Devices
In this work group we will select a spiking neural network (maybe your favourite) and convert it to a form which will run on the DYNAP-se neuromorphic processor.
By going through this process we will attempt to extract general principles for converting networks made for 32/64 bit synchronous architectures ...
Using Spinnaker in the Neurorobotics Platform
The Neurorobotics Platform (NRP) of the Human Brain Project (HBP) is a software tool for simulating spiking neural networks embodied on virtual and real robots. It can be installed locally or used online. Currently, support for SpiNNaker is at the aplha stage.
In this workgroup, we would be demonstrating the ...
Power Management Applied to Neuromorphic Sensor Inputs
In this work group, we will explore how to exploit activity fluctuations in neuromorphic sensors for power management on the SpiNNaker2 prototype. The idea is that at lower input activity the core supply voltage can be reduced, saving energy consumed per spike. We have a DVS connected and support for ...
Neuromorphic Touch
This workgroup will focus on the development of methods and algorithms for neuromorphic tactile perception. We will use prototype optical tactile sensors with integrated DVS128 cameras developed as a collaboration between the University of Bristol and iniVation.
The project will include work on:
- Event-based tracking of the tactile elements of ...
Discussion Groups
Noise and variability in neuromorphic hardware (neurons and synapses)
On Monday Giacomo introduced the issue of inherent noise and variability in neuromorphic hardware, and the discussion---and also 6-words intros indicated there may be several people interested in using it for computation or finding ways to be robust against it. But what are the actual levels of noise/variability in ...
Tiny brains
This discussion group is about making a collection of very small SNN circuits that can be run on neuromorphic prototype chips with very limited resources. If you are aware of networks in insects / sea snails / worms or if you can construct just a nice Braitenberg vehicle, let's sit down, discuss ...
Biological inspired motor control - neurorobotics
how can we model and represent motions using mechanisms from biology to control robots?
of course using spiking neurons!!!
main concepts to discuss:
* M1, pre-motor cortex, cerebellum, spinal cord
* motion generation
* motion adaption
* motor hierarchy
* activation mechanisms
* motor primitives
* neuro simulators (Nest, Nengo)
* tasks: manipulation, grasping, locomotion
Recurrent connections in cortex and ANNs
Cortical computation seems to rely heavily on recurrent connections. In contrast, most ANNs and deep learning models only use feedforward connectivity and stateless neurons (an exception are recurrent networks used in sequence processing).
In this discussion group we will have a look at the similarities and differences between feedback signals ...
Nengo Tutorial
Due to popular demand, this will be a crash-course introduction to the Nengo software package. We will write Python code to train and simulate a SNN that can learn over time, incorporating both spike timing and synaptic dynamics, while allowing us to view the model at various levels of abstraction ...
Neuromorphic pictures, movies and more
A picture says more than a thousand words. Showing neuromorphic and related pictures, movies, ...
Bring your own, pictures and movies. You can also add interesting pictures / YouTube links / in the wiki.
Event-based Unsupervised Learning of Epipolar Geometry
The stereo matching problem is a well-known bottleneck in stereo-vision. Event-based stereo matching approaches make explicit use of precise interocular temporal coincidences. This computationally expensive process can be greatly improved by imposing constraints from the geometry of the stereo setup. However, obtaining this geometry information requires precise camera calibration procedures ...
Insect Inspired Neuromorphic Computing
For those interested in the incorporation of the architectures and computational principles studied in insect neural networks into hardware this discussion group offers an opportunity to share what model systems people are working with and are interested in. Some examples would be the elementary motion computation network studied in Drosophila ...
Dendritic Computation
Many leading neuroscientists and an increasing number of computational neuroscientists are coming to the conclusion that much of the interesting/relevant information processing which leads a neuron to spike is being done within the dendritic branches themselves before any signal reaches the neuron soma. This could be a broad description ...
Decision making continued
We will continue with the discussion of today's morning session
Multi-Scale and Multi-Level Brain Simulation
In the neuromorphic domain, most simulations and hardware implementations are done on spiking neuron level. This limits the spatial and temporal scales accessible under practical considerations. In this discussion group, we want to explore what abstraction levels in brain modeling exist, how they could interact, and which use cases and ...
Hands on YARP installation and use
We discussed the advantages and limitations of YARP/ROS for the neuromorphic community on Tuesday. This one is for a sitting at a laptop install YARP and running some code!
Tools for Hardware and Software Communication (YARP/ROS)
In robotics we have several tools for quickly connecting hardware and software modules that can be running on different computers, and in different languages. Using such a tool could reduce the amount of time required to engineer the communications between different modules that people have individually built, and then want ...
Primate recognition and tracking
We have data from 10 camaras observing a small group of monkeys that move in 3 dimensions in a >50 square metre area. Goal of this discussion group is to design a neural network architecture that can locate and track the animals based on all 10 camara inputs.
Evolving very small SNNs for simple computational tasks, robust to noise and damage
We are interested in using artificial evolution (that can change the topology and weights) to obtain very small spiking neural networks. Very small means 3-10 adaptive exponential or leaky integrate and fire. Simple computational tasks may include temporal pattern recognition, controlling an animat (a simulated robot), or a multiplicative operation ...
Time matters
How do we exploit timing in algorithms and networks of spiking neurons
Why do we need episodic long term memory?
Discussion group about long term memory
Please bring all your questions, ideas, theories and models you would like to discuss.
There will be a focus on episodic memory, but contributions related to long-term memory in general are also welcome.
We will discuss questions such as:
Why do we need episodic ...
Structural plasticity on neuromorphic hardware
We will discuss several different approaches to structural plasticity – implemented on different neuromorphic platforms. A discussion on the biological foundations might also come up.
Memristors in Neuromorphic Circuits
A group to discuss memristors and their use in neuromorphic engineering. The aim is to discuss the operation of devices that currently exist and in parallel brainstorm different applications where they could be used.
Predicting gene expression patterns
We have a database of gene expression (10'000 gene expression patterns) in specific cells at different stages of their development would like to predict future behavior of these genes based on their past expression (temporal series approach?)
Recreational Groups
Rope climbing
We want to go rope climbing close to Capo Caccia. Dates and details will follow/ But plan to bring your gear (Shoes, Harness, Ropes, Quickdraws etc.)
Yoga, acroyoga and slackline
Mens sana in corpore Sano..
Before or after intensive neuromorph sessions, let's try to find some balance (both internal and external).
Before breakfast with a yoga sessions at 7:15 am, in front of the sea, and after dîner with some slackline under the pine wood in front of the ...
Pizza in Alghero
For Pizza lovers!
Hotel dei Pini is one of the best Locations in Alghero with good quality food BUT they do not make Pizza's.
We will go to Alghero down town on Monday evening to enjoy a good Pizza or some typical Sardinian food.
The transfer will leave at 6 ...
Bridge / card games
from happy hour, to the evening meal, and after the evening meal to whenever. We will start by playing bridge in the bar, but will as always, fall into other card games as they come along. come learn new games, practice bridge, or just play the old favourites of go ...