View allAll Photos Tagged Neuron
#deepdream code informatique de l'intelligence artificielle de Google spécifique "Fractal DDC " développé et dédié pour un nouvel art à La Demeure du Chaos - The Abode of Chaos ou comment les machines perçoivent La Demeure du Chaos - The Abode of Chaos
et si leurs regards étaient ce qui se cache derrière la matrice que nous percevons en tant qu'humains? ces multiples miroirs sont peut-être un autre monde plus réel ou plus éthéré... NB thierry bonne lecture de ce post et ses images dantesques.
Depuis quelques temps vous avez peut-être vu circuler sur les réseaux sociaux des images étranges, affublées d'un hashtag (mot-clé) #deepdream.
Deep Dream est un programme d'intelligence artificielle mis au point par les ingénieurs de Google. Ces derniers travaillent à la reconnaissance d'images pour, entre autres, améliorer la pertinence des recherches dans Google. Le 17 juin dernier ils ont publié un billet intitulé : "Inceptionnisme : plus loin dans les réseaux neuronaux".
Dans ce post ils expliquent comment ils ont réussi, dans leurs recherches, à faire analyser une image mais surtout générer des formes par l'ordinateur. Pour que l'intelligence artificielle puisse mieux reconnaître ce qui compose une image, les ingénieurs ont commencé par lui montrer des millions de photos.
Plusieurs couches de neurones
L'intelligence artificielle fonctionne ici en un ensemble de réseaux de neurones qu'il faut se figurer comme différentes couches. La première est chargée de regarder les bords et les angles d'une image.
Les couches intermédiaires cherchent quant à elles les formes et les différents éléments présents dans l'image comme une feuille ou une porte. Les derniers réseaux assemblent toutes ces informations pour en fournir des interprétations complexes, comme des arbres ou des bâtiments.
Pour comprendre au mieux comment fonctionnent ces couches, les ingénieurs ont tenté de pousser l'analyse de certaines. Ils résument ainsi la commande faite au système : "Quoi que tu vois, on veut le voir encore plus." C'est alors que l'intelligence artificielle a généré des formes au sein des clichés.
"Si un nuage ressemble un petit peu à un oiseau, alors le système va le faire ressembler encore plus à un oiseau, expliquent les ingénieurs. En réitérant l’action, le programme va reconnaître un oiseau plus fortement et ainsi de suite jusqu’à ce qu’un oiseau très détaillé apparaisse, comme sorti de nulle part."
"L'inceptionnisme"
Les images varient selon le réseau neuronal qui est amplifié. Par exemple, plus on sollicite les couches inférieures, plus des traits vont apparaître. Si on stimule d'avantage les couches supérieures, ce sont des objets qui émergent de l'image.
Les ingénieurs précisent d'ailleurs que comme l'ordinateur a enregistré beaucoup de clichés d'animaux durant son entraînement, il en reproduit souvent. Et parfois en les mixant, ce qui crée des créatures étranges.
Pour ces chercheurs, le Deep Dream a ainsi créé un mouvement artistique qu'ils appellent "l'#inceptionnisme", en référence à l'architecture des réseaux neuronaux.
Au début, cette expérimentation ne cherchait qu'à améliorer l'intelligence artificielle. Mais lorsque les ingénieurs ont posté ce billet, de nombreux internautes se sont intéressés à ce Deep Dream.
Google a donc rendu public le code utilisé pour générer ces images. Des informaticiens s'en sont emparés et ont mis au point des logiciels et des interfaces pour que les internautes puissent s'en servir.
Ce qui ne manque pas de plaire à Google. Les chercheurs encouragent à taguer les images #deepdream sur Twitter, Facebook ou Google+. "Il sera intéressant de voir quelles images les gens arrivent à générer", écrivent-ils.
Regeneration or sprouting of injured axons improves with anti-RGMa antibody treatment (top) compared to controls (bottom). (JCB 173(1) TOC2)
This image is available to the public to copy, distribute, or display under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported license.
Reference: Hata et al. (2006) J. Cell Biol. 173:47-58.
Published on: April 3, 2006.
Doi: 10.1083/jcb.200508143.
Read the full article at:
Transgenic mouse lines engineered at the Allen Institute allow researchers to see neurons in new ways. Here, excitatory neurons in the cortex are labeled in green and red.
O Neuron conta sempre com uma equipe de plantão para atender qualqer ocorrência envolvendo problemas de neurologia e neurocirurgia. Os tratamentos do Neuron incluem acompanhamentos clÃnicos e cirúrgicos.
Foto: Fernanda Acioly
In Naked Knotted Neurons, a group of protesters, some injured, some choking on tear gas, escape violent confrontation with police and other forces by staggering into a safe house they find in the midst of chaos. Strangers to one another, they soon discover they are from different worlds: all were involved in protests, but in different places and times. A trio of dieties, Fate, Chance and Destiny, have gathered them together to charge them with a task: to create a new Hero to solve the world’s most intractable, knotted problems. How to get this message across? Puppets, riddles, and audience participation reveal the secrets the protesters need to fulfill this task.
Following their run at the International Fringe Festival in Edinburgh, Scotland, the Penn Theatre Ensemble presented the company-devised piece Naked Knotted Neurons at Annenberg Center Live on September 4th and 5th, 2015.
Photo by Dyana Wing So.
Mirror neurons and social cognition – Gold nuggets for advertisers or just too complex to be practical?
Even if you limit your reading these days to Stieg Larsson, the free newspapers on the underground, and the trade press you probably have come across a few scientific concepts that many advertisers get very excited about: Mirror neurons, social neuroscience, and theory of mind. Understanding how these social systems in the human brain work and how we use them to make sense of our social environment might hold the secrets for triggering empathy in people, for evoking emotions in ad viewers and ultimately how to create more effective campaigns. If only the bloody science behind these things wasn’t so complicated and one knew where to start in understanding them!
Just last week the academic publishing house Cell Press organised a one-day workshop at the University of London’s Birkbeck College on Social Cognition, the scientific area into which mirror neurons and all those other exciting discoveries broadly fall. The contributors were all first-rank scientists and luminary figures in their respective fields delivering very high-level overviews on the state of knowledge for each topic. Of course, the workshop wasn’t targeted at advertising people or marketeers (academics normally don’t feel the need to explain anything to the business world). Instead, the audience was mainly academics from a broad range of disciplines (biology to neuroscientists to social scientists). But the workshop was all free and advertising professionals could have attended (if they hadn’t been too busy to put the last touches to the very important deck for that really important pitch next week … you know how it is). So, I as the Scientist in Residence for DDB UK, was probably the only one in the audience who actually listened with an advertising ear to the latest developments in social cognition.
Since the initial publication by a research team around Giacomo Rizzolatti in 1992 mirror neurons have had a steep career and are today very widely postulated as a neural system that can explain a variety of phenomena from consciousness to the understanding of what other people are thinking (the so-called theory of mind) and the learning of complex motor and social behaviour. Really, if you do a literature search for mirror neurons these days it seems like they can explain almost any interesting human behaviour. Essentially, mirror neurons are ensembles of nerve cells in probably three different areas of the human brain that are active when we perform a certain action (like grab a cup of tea or hit a ball with a tennis racket) but also when we see other people perform the same action. By this very behaviour they could explain how we make sense out of the world around us and why we primates can learn so quickly: whenever we see someone performing an action it is a bit like we are doing the same thing ourselves. You have to admit that this sounds very elegant (and may trigger all sorts of philosophical speculations if you are that kind of guy), but as James Kilner from University College, London explained, a) they are only directly proven to exist in monkeys, b) it is unclear what they ‘mirror’ when an action that you see is ambiguous, and c) people with lesions in presumed mirror neuron areas (to be precise: the inferior frontal cortex, the inferior parietal lobe, and the super temporal sulcus – if you want to show off at the next Christmas party) can still understand the intentions behind the actions they see. Kilner thus reckons that many claims of what mirror neurons actually do have been a bit bold. They indeed seem to be active when we observe actions predicting what is going to happen next when someone is whacking a tennis racket towards a flying ball (or towards a referees head). But mirror neurons might actually not encode the intention of an action – why the tennis racket is on collision course with the referees head (and you could clearly see that there might be several reasons that caused an incident like this to happen). Thus, Kilner suggests that the function of mirror neurons is at a lower level, being more concerned with how an action is performed (eg its kinematics) rather than why it is being performed.
Does this make them less attractive to the advertiser with a curious scientific mind? Take a look in the mirror and decide.
Cultured dorsal root ganglion neurons lacking the E3 ubiquitin ligase PHR1, which targets the kinase DLK for proteasomal degradation, display high levels of phosphorylation of the MAP kinase JNK (green) in their axons. Huntwork-Rodriguez et al. show that a rapid increase in the abundance of DLK protein occurs after neuronal injury and is important for initiation of downstream signaling events and neuronal apoptosis. Nuclei are labeled red.
Image courtesy of Huntwork-Rodriguez et al.
Reference: Huntwork-Rodriguez et al. (2013) J. Cell Biol. 202:747-763
Published on August 26, 2013.
doi: 10.1083/jcb.201303066
Read the full article online at: jcb.rupress.org/content/202/5/747.full
Neurons and epigenetics and something or some such. There was a lot of impressive translation happening.
Submitted caption:
Bright field image showing the AFM tip probing the biomechanical property of neuron cells in culture.
##
This image was submitted to the University of Chicago's 2022 Science as Art competition. From neurons to nanoparticles, the entries display the gorgeous landscape of scientific research going on every day at the University of Chicago. More than 100 images were submitted to the contest, from undergraduates, graduate students, staff, postdoctoral researchers, and faculty members. Read more.
Image may only be reprinted with credit to the authors and the University of Chicago.
Neuronal migration requires N-syndecan, which FRET reveals clusters with the EGF receptor. (JCB 174(4) TOC2)
This image is available to the public to copy, distribute, or display under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported license.
Reference: Hienola et al. (2006) J. Cell Biol. 174;569-580
Published on: August 14, 2006
Doi: 10.1083/jcb.200602043
Read the full article at:
Activated neurons in an image from the Allen Brain Observatory, a window into single-cell neural activity in the mouse visual cortex.
Neurons have a soma (or cell body) and processes, which are extensions from the soma that receive or carry information. Processes can be diversified as dendrites which receive information or axon which carries information. The axon shown here is covered by insulating sheaths of myelin. (Image credit: "Labeled structure of a neuron" by Chiara Mazzasette is licensed under CC BY-SA 4.0 / A derivative from the original work)
