Lectures | Work Groups | Discussion | Recreation | Social | Create New Session

Lectures

Theoretical Computation

What do theoretical models of computation bring to the understanding of biological information processors?
Can we forward design a 'biological' style computation?

Go to group page

Biological Computation

What is Biological Computation in the universe of all Computation.
How would you recognize one if you saw one?

Go to group page

Navigation 2

The journey continues...

Go to group page

Setting Goals 1

What are the challenges, how should we evaluate, which should we choose, and how should we implement?

Go to group page

Setting Goals 2

continues...

Go to group page

Long-term memory & plasticity 1

State and time

Go to group page

Long-term memory & plasticity 2

continues...

Go to group page

Inference & Noise

What was that??

Go to group page

Inference & Noise 2

continues,,,,

Go to group page

Balanced activity & working memory 1
Balanced activity & working memory 2

continues...

Go to group page

Cooperative & Distributed decisions 1

But not global? :-)

Go to group page

Cooperative & Distributed decisions 2

continues...

Go to group page

Control & Error prediction 1

There are no two ways about this...

Go to group page

Control & Error prediction 2

continues...

Go to group page

Navigation 1

Where are we? Where and how should we go?

Go to group page

Neuromorphic Computation

What is Neuromorphic Computation, in the universe of all computation?

Go to group page

Hardware show-case and workgroup presentation

Participants who brought along hardware can showcase it to others and form workgroups. Proposals for workgroups are presented.

Go to group page

Classification & Clustering 2

Giacomo Valle
Enea Ceolini

Go to group page

Classification & Clustering 1

Chairs

Chiara Bartolozzi
Germain Haessig

Speaker 1 - Thomas Nowotny

Speaker 2 - Simon Thorpe will concentrate on Classification and Clustering under conditions where neurons only get to generate a single spike, and where synaptic connections are binary. Such simplified systems can nevertheless allow quite sophisticated behaviors.

Go to group page

Self-introduction of the participants

Self-introduction by all participants. Your introduction is limited to 6 words! Choose them for maximum information. Collective construction of a map of participant's interests.

Go to group page

Workgroups scheduling & Discussion groups presentation

We will introduce the discussion groups and schedule non-overlapping schedules for the first meeting of each workgroup.

Go to group page

Work Groups

Accelerated neuromorphics on BrainScaleS-2

BrainScaleS-2 JupyterHub: http://10.2.17.58

We bring along multiple BrainScales-2 systems and invite everyone to try some of our demos or implement their own neuromorphic experiments.

The most recent generation of BrainScaleS-2 ASICs features 512 analog multi-compartment neuron circuits with 256 plastic synapses each.
The analog circuits are tightly coupled to on-chip …

Go to group page

Spiking network synthesis for a resting vs Grasp (orthogonal gesture) using sEMG on a Neuromorphic hardware

This workgroup will investigate Surface Electromyography (sEMG) data for rest position in collaboration with the workgroup "Regressing individual finger position from sEMG using a spike-based approach" using the Excitatory-Inhibitory network using the DYNAP-SE 2. The approach aligns with the studies on using balance networks to investigate epilepsy vs resting state. …

Go to group page

Classification of time-dependent data with SNNs

Even though many datasets are time-dependent (videos, audio, biological signals, DVS streams, etc.), not all of them actually require machine learning models to refer to time-dependent features for an effective classification. For example, in videos, a single frame can already give out a lot of information, which can be sufficient …

Go to group page

Routing information through oscillations

Future neuromorphic systems will integrate multiple sub-systems where information should flow through different sub-systems in a context-dependent manner. In this project we will try to create such routing by using an hypothesis from neuroscience called communication-through-coherence. We will use it for iplement it in dynapse (other hardware is welcome too). …

Go to group page

Temporal augmentation for SNNs

Augmentation is an important part of training deep neural nets these days. Whereas convolutional neural networks typically use spatial augmentation techniques such as rotation, shifting and other affine transforms, data for spiking neural networks often has an important temporal component. I'd be interested in exploring specifically temporal augmentation techniques on …

Go to group page

Bloggers and blogging

A channel for volunteer workshop bloggers to communicate.
https://cne2018.blogspot.com
for 2018 blog.
We need a few energetic bloggers for this year's CCNW.

Go to group page

Playing with DYNAP-SE1

Get access to the reconfigurable asynchronous mixed-signal realtime neuromorphic processor to emulate (feedforward or recurrent) networks of AdEx LIF neurons.

For tutorial, clone the following repository, checkout branch samna-dynapse1 and search for the tutorial jupyter notebook:
https://gitlab.com/neuroinf/ctxctl_contrib.git

Go to group page

SpiNNaker, event-based sensors, and BitBrain

This workshop is intended for people interested in working with SpiNNaker. Tutorials for using SpiNNaker: https://spinnakermanchester.github.io/workshops/eighth.html/.

Our particular aim is to implement BitBrain on SpiNNaker neuromorphic platform and learn event-based stimuli from sensors. This work was presented during NICE conference 2022: https://flagship.kip.uni-heidelberg.de/jss/HBPm?mI=235&publicVideoID=8944.

BitBrain is a learning algorithm based upon a …

Go to group page

Loihi and Lava tutorials

In this workgroup, we will answer any questions you have about Lava and Loihi and run a couple of Lava tutorials.

You can grab Lava here: https://github.com/lava-nc/lava and get started yourself. Lava is an open-source project.

In the first session, we will go through the high-level structure and idea behind …

Go to group page

ADP: Activity-driven perception and control

A general workgroup to collect people interested in activity-driven (event-driven) perception, computation, and control.
Volunteers to coordinate this group would be very welcome, please let Tobi know

Tobi is bringing a few DAVIS event cameras, plus a PDAVIS polarization camera prototype. Also some arduinos and prototyping boards, servos, LEDs, etc. …

Go to group page

Olfaction in humans, animals, and machines

Our aim is to discuss the current state of neuromorphic olfaction, relate it to the state of the art in machine olfaction, explore the links to olfaction research in humans and animals. We could also discuss the development of neuromorphic/event-based gas sensors, sensor technologies. Special interest could be devoted to …

Go to group page

Mobile Robot Competition (event-vision moving-obstacles avoidance)

This work group focusses on perception, reasoning, prediction, and motor action in a simple closed-loop system. Use your own computing hardware, or use our neuromorphic SpiNNaker or GPU setup.

We provide small mobile robots, equipped with high-resolution event camera and some additional sensors. We setup a soft-ball on a string …

Go to group page

Fast and energy-efficient neuromorphic deep learning with spike times

Deep Learning with spiking neural networks is a hot topic in general and specifically in the neuromorphic field. In this workshop we will work together on applying and extending an algorithm that allows exact error-backpropagation in networks of LIF neurons (https://www.nature.com/articles/s42256-021-00388-x respectively https://arxiv.org/abs/1912.11443).
As a starting point, we provide insights …

Go to group page

Machine Learning for Neuromorphic hardware

Several of us are bringing new Neuromorphic chips along to Capocaccia this year:
- Xylo: a new digital SNN inference chip
- Speck: a single-die event-based vision sensor plus spiking CNN processor

These are integrated into open-source machine learning pipelines based on Python, PyTorch and others, for training and deploying …

Go to group page

Regressing individual finger position from sEMG using a spike-based approach

How to design an EMG-prosthesis system that can mimic the natural hand movement remains an open question. One approach is through classification-based control; the user motor intent gets decoded and mapped to a finite set of motions. However, this approach suffers from one main limitation: the user can only perform …

Go to group page

Predicting future behaviour using event-based digital activity

On its surface human behaviour in the real world appears complex, and highly different from one moment to the next. The log of behavioral episodes captured on the smartphone touchscreen can span multiple years, enabling the discovery of hidden behavioral structures. Recent work in our laboratory suggest that day-to-day smartphone …

Go to group page

GPU-enhanced neural networks

Fancy running your SNNs 10x faster? Our GPU enhanced Neuronal Networks (GeNN) library is freely available from https://genn-team.github.io/ and provides an environment for GPU accelerated spiking neural network simulations. GeNN is capable of simulating large spiking neural network (SNN) models at competitive speeds, even on single, commodity GPUs. In GeNN, …

Go to group page

Inhibition-Stabilized Networks: if we can control them, what can we do with them?

Cortical networks have the remarkable ability to self-assemble into dynamic regimes in which excitatory positive-feedback is balanced by recurrent inhibition. This inhibition-stabilized regime is increasingly viewed as the default dynamic regime of the cortex and believed to underlie many cortical computations, such as input amplification, working memory or motor control. …

Go to group page

Discussion Groups

User friendly and efficient neuromorphic software

We will create an overview of software frameworks that are being used and developed by the community to program neuromorphic hardware, list their key assumptions, design principles, and goals, as well as limitations, challenges, and unmet user needs. We will then discuss the possibility of converging on a common programming …

Go to group page

Benchmarking SNNs. Choosing tasks and metrics

ANNs can be compared to each other in terms of number of parameters, energy consumption, classification accuracy and many other things.
In order to showcase benefits of SNNs over ANNs, we have to identify the right tasks, network architectures and metrics such as energy-delay product. Ideally we also create some …

Go to group page

Exploiting diversity of dynamical neuron types in spiking neural networks

Biological neurons are diverse. They do not only appear in different cell types, but also in different "dynamical types" that is different transitions between resting and spiking state. For example, class 1 neurons (with a continuous f-I curve) correspond to one dynamical type (SNIC) whereas class 2 neurons (with a …

Go to group page

Issues faced with Published datasets

In this workgroup, we will discuss some of the problems and issues we face while working with published datasets in general. Most of the times, we end up using these datasets for our work without realizing that there might be some problems hidden in the dataset. Sometimes, our work proves …

Go to group page

Why don't androids dream of electric sheep?

This is a freeform discussion to explore the potential mechanisms involved in sleep and evaluate their benefit in the generation of learning architectures.

All beings with biological brains exhibit some form of sleep, from ants to humans. It has been suggested sleep is necessary for the cleaning of chemicals in …

Go to group page

Quo vadis neuromorphic hardware?

Let's continue the discussion from Monday morning:

- Where are we going with neuromorphic hardware?
- Should we start from scratch and design generic building blocks?
- What can we learn and how can we benefit from existing systems?
- Are we limited by the actual hardware or missing (software) …

Go to group page

Simulating the entire human brain today?

You can now buy an Apple MacStudio with an M1 Ultra chip. It has 128 GB of low latency unified on chip memory that can be directly accessed by 20 CPU cores, 64 GPU cores and 32 Neural Engine Cores. The memory bandwidth is 800 GB/second, and the chip is …

Go to group page

Medical Applications for Neuromorphic Technologies

As discussed during the first day of the workshop, in order to keep neuromorphics alive and relevant, we must push the technology to have more of an impact in peoples' lives. I am starting this discussion group so we can specifically compile and discuss past and emerging medical applications for …

Go to group page

BitBrain and other methods for single-pass and continuous learning

Many contemporary machine learning/AI methods require extensive computation for training which consumes a great deal of time and energy resource. Their inference mechanism also tends to have a large energy-latency product. We will discuss methods that can learn in a single-pass. This usually also promotes an ability to learn continuously …

Go to group page

Theme A: Neuromorphic Perception for/and control

Discussion group wrapper for theme A

Go to group page

Theme B: Learning

Discussion group wrapper for theme B

Go to group page

Should we pay attention?

Can we define attention and how we can understand the visual perception that allows us alertness. How can we leverage event based processing to inform salient features? Because we are event based species, we want to discuss what more we need to closely relate saliency and noninforming features. Can event …

Go to group page

Theme C: From single neuron to network dynamics

Discussion group wrapper for theme C

Go to group page

Theme D: Neuromorphic Hardware. From devices to systems

Discussion group wrapper for theme D

Go to group page

Abstracting principles and concepts as a prerequisite for reasoning

Abstracting principles and concepts (e.g., shape, color, learn to count) is essential for reasoning. In this session, we can discuss learning rules, and architectures that would allow our systems to learn these principles. See this paper: https://arxiv.org/pdf/1705.03633.pdf

Go to group page

Combining Hebbian and predictive plasticity yields invariant object representations in deep nets

Discriminating distinct objects and concepts from sensory stimuli is essential for survival. Our brains accomplish this feat by forming meaningful internal representations in deep sensory networks with plastic synaptic connections. Experience-dependent plasticity presumably exploits temporal contingencies between sensory inputs to build these internal representations. However, the precise mechanisms underlying plasticity …

Go to group page

Tutorial: introduction to PyTorch

Some people who have not worked with deep learning before are interested in an introduction to pytorch. We will give a very introductory tutorial on what PyTorch is, how to train your first (more or less) deep neural network with it, how to use it with a GPU, and how …

Go to group page

Beating our cortical conundrums

Featuring Florian ‘the Fish‘ Engert, Georg, ‘the Gorilla‘ Keller, and Valerio ‘the Monk‘ Mante in their final joint performance at CCNW22,. See them escape from their chains before your very eyes..! Let them show you how your brains could be... (Supported by Matthew, Michael, Yves, and a cast of thousands).

Go to group page

Neuromorphic engineering: where are we and how can we move towards AGI?

Neuromorphic engineering field started about 3 decades ago with the goal to understand the organizing principles of the brain and implementing it.
- Which organizing principles do we know today which have proven to have a technological advantage?
- What is missing in the mainstream AI that these organizing principles …

Go to group page

Spiking networks with PyTorch and Norse

After Martino's presentation on PyTorch, we'll learn how PyTorch can be extended to spiking neural networks.
We'll build small neural networks, train them with gradients, and explore some community-based tools that makes our lives easier and much more productive.
We'll end the session by briefly discussing biological learning.

For the …

Go to group page

Exploiting inference capabilities of memristive arrays for neuromorphic hardware

Memristive (memory + resistor) arrays are matrix-like structures of a special kind of electrical resistors, with adjustable resistance value. As they allow vector-matrix multiplication or Multiply-Accumulate functions in one single step, they are considered highly intersting for for ANN accelerators.
But can we further exploit this unique feature for neuromorphic …

Go to group page

Learning, causal reasoning, creativity

What does it mean for an agent to be creative? And is creativity a precondition for causal reasoning and learning, or rather a strategy that agents follow to reach general intelligence?

This group has been created with the aim of starting an open and interdisciplinary discussion on how the answers …

Go to group page

Plugging into real world behaviour

Can we explain real-world behavior? Our laboratory captures discrete real-world behavioral events on the smartphone, and some of these records are accumulated in conjunction with invasive and non-invasive neural recordings. The behavioral logs - spanning multiple years - reveal (novel) structures underlying real-world human behavior (in health and disease). Furthermore, …

Go to group page

Recreation Groups

Sailing group

The hotel has a small number of dinghies that are available to hire.
Windsurf boards (I think 25 euros/hour).
Lasers (30 euros/hour, 1 person, maybe 2)
Hobie cat 15 (40 euros/hour, 1-4 people crew), available later this week.
I will try it out already on Monday at 5 if there …

Go to group page

CapoCaccia Climbing Crew

CapoCaccio offers one of the most beautiful climbing sites right where the ocean meets the land. If you are into rock climbing and/or want to have one, two or more relaxed afternoons on the rope or in the hammock make sure to bring your climbing shoes and harness.

We do …

Go to group page

Chocolate Tasting

It has been too long since the last chocolate tasting in CapoCaccia...

Luckily, one of the participants is co-founder of Allpa Kula, a tiny chocolate company in Switzerland.

This is your chance to try sustainable, high-quality bean-to-bar chocolate, made from some of the tastiest and most eco-friendly cacaos around the …

Go to group page

Sketching

Do you draw, or would you like to start? This session is to get together and spend an hour sketching. It might also be interesting to discuss how perception for drawing differs from everyday perception. Today , given the weather, we can do this indoors but you might want to …

Go to group page

Football

Let's gather to play some football

Go to group page

Tennis

To plan tennis hitting

Go to group page

Gaming, esp. Poker and magic talent

To let people know we want to play and to plan our own workshop magic or talent show

Go to group page

Cycling to capo caccia

We will cycle to capo caccia for those who can still get a bike.

Go to group page

Afternoon hiking in the beautiful countryside

We will walk/hike on some of the beautiful trails around Capo Caccia. Lush vegetation, steep cliffs, stunning views, and plenty of ruins await. Join us to spend some quality recreation time with like-minded hikers.

Some spots can be reached from the hotel, others are a within a few minutes drive. …

Go to group page

Scuba Diving

Are you interested in doing some Scuba Diving?

Paolo Micarelli from Diving Alghero https://en.divingalghero.com/dove can take up to 6 people around 17h (earlier if possible) on Saturday 7th May and can do training dives for beginners on Sunday 8th in the afternoon. €50 per dive, plus €20 for equipment.

Please …

Go to group page

Watching Eurovision together

The first semifinal of Eurovision is happening during CapoCaccia. Let’s watch it together

Go to group page

Social Events

Morning Coffee Break

This morning, and every morning. Go for-coffee :-)

Go to group page

Happy Hour!

Today, and every day...

Go to group page

Workshop Dinner

And 2021 Misha Mahowald Prize presentation.

Go to group page

Welcome Apero

Meet old friends again, and greet new ones.

Go to group page

Lunch

Every day

Go to group page

Dinner

Every day

Go to group page