Animals, when moving through the environment actively alter their visual perception, thus enriching the quality and quantity of information about certain objects and the entire scene. This information is used to recognize and but also to avoid objects. In order to avoid collisions and navigate through the environment, the basic ...

One of the essential properties of neurons and synapses is that they can change over time, in particular as a result of learning. In the newly developed HICANN-DLS chip from Heidelberg, we use an on-chip plasticity processing unit (NUX) to control that change. In this workshop we plan to implement ...

We will integrate the neuromorphic VLSI device ROLLS (256 spiking neurons with realistic neuronal dynamics, 128K synapses, including synapses with on-chip plasticity circuits) with neuromorphic vision sensor (embedded Dynamic Vision Sensor eDVS) and sensors and motors of robotic vehicles (2x"Pushbots", 1x"Omnibot") using a miniature PC "Parallella". We will use this ...

Thorough testing is essential during early development stages of neuromorphic hardware. However, transistor level simulations and prototype chips often do not allow the study of entire networks but only of smaller components such as single neurons. Therefore, it is of obvious benefit to have single neuron experiments which give an ...

Non-spiking computer vision with event-driven cameras.
YARP is the robot middle-ware used for the iCub robot. We use YARP with AER representation and can send packets of events over standard Ethernet network such that event-based sensors, computation, actuation and visualisation can be distributed among different computers and robots. We have ...

The idea for this project is to implement an algorithm to monitor the attention of a person (e.g. a driver of a car) by tracking the direction of view, the eye's movements and maybe also the blinking frequency using a DVS sensor. A second goal would be to implement ...

In this workgroup, I propose to work on algorithmic and circuit design ideas for designing memristor-based neuromorphic designs. Some of the interesting problems to be considered could be variability-tolerant algorithms and circuits, synapse design, synaptic learning rules, etc.

Neuromorphic hardware has reached a point where it is possible to use a script-based interface to specify your network and have it configured by an automated tool chain to run on a hardware platform. In this workgroup we will explore systems for large-scale simulation that let you do just that ...

The Modular Event-driven Growing Asynchronous Simulator (MegaSim) is a tool designed for simulating Address-event-representation (AER) systems. In MegaSim the user may define a topology by utilizing modules that process input events, nodes to connect modules with each other and input stimuli sources for example data recorded from a DVS retina ...

We have many devices out there generating and using AER packets of some form in various ways. Wouldn't it be nice if we could link them together to form larger, multi-system simulations? Up until recently, the main problem was that everyone was using a different 'flavour' of AER, with different ...

This time, we will make use the ROLLS Neuromorphic Processor as our event-source and compute both the spatial and temporal correlation of events. The Idea is to set up the Neuromorphic Processors Long-Term-Plastic Synapses in a predetermined fashion that is unknown to the participants, and then to come up with ...

In this work group, I propose to design and implement a network on the neuromorphic VLSI device CxQuad to do real-time interactive decision making. The goal is to implement an AI agent on the neuromorphic substrate to play the Wisconsin Card Sorting Test using the DAVIS camera as the visual ...

Rotating machines need sensory-motor systems to stay healthy and perform well. Although the motor degrees of freedom of machinery like a wind turbine or cargo train wheelset are different compared to typical robots, most machines exist in complex environments and their characteristics change over time due to wear and maintenance ...

The number of neurons of the brain is not fixed at birth. Neurons keep being produced throughout life in some regions of the brain: among which the dentate gyrus of the hippocampus. Adult neurogenesis has a physiological relevance, newborn neurons are integrated in circuits and this plasticity of circuits seems ...

We will discuss possibilities of supervised training of deep spiking architectures. In the context of spiking networks back propagation is not applicable. Therefore, alternative approaches have to be followed. The goal of the group is to come up with different alternatives. One possible approach is recently proposed "target propagation", which ...

Milestones of AI dropping years before their expected due date (AlphaGo) and prominent figures warning of the nearing Robot Apocalypse give the impression of AI standing on the brink of achieving its ultimate goal of surpassing human intelligence.

In this discussion group I would like us to come up with ...

Deep neural networks (DNN) and Convolutional Networks (CNNs) offer a great opportunity for lots of classification and recognition tasks in machine learning, such as handwritten digit recognition (based on MNIST dataset) or image classification (based on Imagenet dataset). The weight between two consecutive neurons within the deep network is stored ...

We will continue from Yves Fregnec's presentation to further discuss the view of cortex as a nonlinear spatiotemporal integration machine, aiming towards improved algorithms and architectures for neuromorphic engineering.

Possible projects range from using STDP rule in MegaSim to form or modify receptive fields, to improving the real-world utility of ...

Convolutional Neural Networks (CNN) have shown remarkable classification performances on many different datasets. However, these networks need couple of days to train for example on ImageNet and even during inference the processing time of one image takes too long to run state-of-the-art networks in real-time, especially in context of a ...

Second meeting for the workgroup

The learning principles used by the nervous system are subject to many constraints imposed by the nature of its computational substrate. Could it be that these constraints are useful for determining canonical learning mechanisms that can be used in multiple areas and domains, or for building cognitive agents that interact ...

How do we evaluate whether a spiking neural network is performing 'well' or 'badly' on a task? Up until now, the methods used have been a mix of some formal definitions mixed with ad-hoc measures. The result is a lot of metrics that are difficult to compare, difficult to establish ...

TU Dresden and the University of Manchester are currently working on the next-generation SpiNNaker system. It will be a radical redesign using a state-of-the-art 28nm process so there is considerable room for more features. Things we are looking at include dedicated co-processors (e.g. for computation of exponentials and random ...

While most artificial neuron networks continue to assume a point model for spiking in neurons (a single weighted sum of all of its synapses, who play standard roles), several new neuroscience studies suggest that the spatial proximity of synapses relative to the neuron body (eg proximal, distal, apicall), is surprisingly ...

Demonstrations of the DVS (Dynamic Vision Sensor), DAVIS (Dynamic and Active-pixel Vision Sensor) and DAS (Dynamic Audio Sensor), including how to work with them using "jAER" the open-source framework for visualisation of address-event-based data and rapid development of address-event-based algorithms. "cAER" was also demoed.

There are various videos of our ...

CapoCaccia is fortunate to be right next to one of the major climbing venues in Sardinia. There are actually 2 areas: 'Placche di Peyer', within walking distance of the hotel and featuring short, sharp vertical lines on compact limestone in a peaceful environment, and 'Casarotto', an easy drive or taxi ...

I am planning to organize an easy bike tour on the free Sunday. We can discuss about the route (nice way to a paradise beach...). All levels are welcome. Let me know in advance if you want to join, we will have to organize the bike rental (around 10 euros ...

A session to enjoy and explore different types of tea. We will try to see if our taste buds can detect the subtle flavours of the teas.

Register here to arrange for tennis on the court behind the hotel.
We have 3 rackets and new balls.

Register here to arrange times for poker or other gaming activities

The plan is to get some good local cheeses and try them together with nice local wines too. Feel free to bring some cheeses from your country!

Today iniVation will lay on drinks from 18:00 to 19:00; there'll be an announcement at 18:15

The plan will be to meet at the pool to start (its a little warmer) and than afterwards hop into the sea. If the weather is good and warm.
Every one is wellcome to join.

Tuesday 16: The swimming pool is sadly closed, but i think it will open soon ...

Ultimate frisbee is a non-contact team sport which involves running, throwing and catching. I teach and play in Genova and brought my disc with me here. During enforced sport time we can find some space on the beach and throw. If we have enough people I can teach you how ...

Returning again: the annual chocolate recreational group. We will arrange both 'tutorial' sessions designed to introduce people to fine chocolate bars and bonbons, and 'competition' sessions where chocolate bars are 'tasted off' to pronounce a winner! Please bring the best chocolates you can find locally. I will bring plenty more ...