Skip to main content

Home/ Advanced Concepts Team/ Group items matching "novelty" in title, tags, annotations or url

Group items matching
in title, tags, annotations or url

Sort By: Relevance | Date Filter: All | Bookmarks | Topics Simple Middle
Luís F. Simões

Encouraging Behavioral Diversity in Evolutionary Robotics: An Empirical Study - MIT Press Journals - Evolutionary Computation - 2 views

www.mitpressjournals.org/...EVCO_a_00048

evolutionary computation neural_networks robotics behaviour

shared by Luís F. Simões on 04 Mar 12 - No Cached
  • several papers recently proposed to explicitly encourage the diversity of the robot behaviors, rather than the diversity of the genotypes as in classic evolutionary optimization. Such an approach avoids the need to compute distances between structures and the pitfalls of the noninjectivity of the phenotype/behavior relation; however, it also introduces new questions: how to compare behavior?
  • In this paper, we review the main published approaches to behavioral diversity and benchmark them in a common framework.
  • The results show that fostering behavioral diversity substantially improves the evolutionary process in the investigated experiments, regardless of genotype or task.
  • Luís F. Simões on 04 Mar 12
     
    paywall skipping: http://www.isir.upmc.fr/files/2011ACLI2061.pdf The most complete study I've seen so far on a new approach (Novelty Search) that has been gaining a lot of attention lately. And they even use parallel coordinates to visualize the results!! ;)
Juxi Leitner

Real-Life Cyborg Astrobiologists to Search for Signs of Life on Future Mars Missions - 0 views

www.dailygalaxy.com/...e-on-future-mars-missions.html

technology novelty detection inf science

shared by Juxi Leitner on 03 Nov 09 - Cached
LeopoldS liked it
  • EuroGeo team developed a wearable-computer platform for testing computer-vision exploration algorithms in real-time at geological or astrobiological field sites, focusing on the concept of "uncommon mapping"  in order to identify contrasting areas in an image of a planetary surface. Recently, the system was made more ergonomic and easy to use by porting the system into a phone-cam platform connected to a remote server.
  • a second computer-vision exploration algorithm using a  neural network in order to remember aspects of previous images and to perform novelty detection
  • Juxi Leitner on 03 Nov 09
     
    well a bit misleading title...
santecarloni

Artificial Braneworlds Made to Collide In Lab - Technology Review - 4 views

www.technologyreview.com/...27537

metamaterials cosmology PHY

shared by santecarloni on 31 Jan 12 - No Cached
pacome delva liked it
  • santecarloni on 31 Jan 12
     
    Physicists have simulated two universes colliding inside a metamaterial--  Now, this is cool (if it is true...)
  • ...1 more comment...
  • LeopoldS on 01 Feb 12
     
    we... the article is a bit overblown in my view ... except maybe the last paragraphs: "The collision between universe's is a variation on this theme. "The "colliding universe" scenario can be realized as a simple extension of our earlier experiments simulating the spacetime geometry in the vicinity of big bang," he says. He simulates an expanding universe using concetric rings of gold separated by a dielectric. "When the two concentric ring ("universe") patterns touch each other ("collide"), a Minkowski domain wall is created, in which the metallic stripes touch each other at a small angle," he says. Being able to recreate these exotic events in the lab is certainly interesting but it is beginning to lose its novelty. The problem is that this work is not telling us anything we didn't know--the universe behaves the same way inside a metamaterial as it does outside. What Smolyaninov needs is a way of using his exotic materials to do something interesting. In other words, he needs a killer app. Any ideas? "
  • Luzi Bergamin on 05 Feb 12
     
    Hm, they use more or less everything I don't especially like. They are nonmagnetic, so the relation materialGR is already rather weak. Usually, experimentalists prefer nonmagnetic media, since they are cheaper and broadband. At least the broadband is no argument here, since the frequency defines the "mass", which I find a rather strange point of view. And finally, they use strong anisotropy as a model of "time", which is rather problematic. Of course, the spatial direction with eps<0 appears in the wave equation with the same sign as time. But this does not mean that it behaves like time. But to teach material physicists that time is more than just a different sign in the wave equation seems to be as hopeless as to teach them that a black hole is more than something that absorbs all light... SIGHHH
  • LeopoldS on 10 Feb 12
     
    Luzi I miss you ...
Kevin de Groote

Gartner Adds Big Data, Gamification, and Internet of Things to Its Hype Cycle - 2 views

www.readwriteweb.com/...ner-adds-big-data-gamifica.php

shared by Kevin de Groote on 24 Aug 11 - No Cached
LeopoldS liked it
  • LeopoldS on 27 Aug 11
     
    nice idea to bring some structure to the list of IT novelties ...
jcunha

When AI is made by AI, results are impressive - 6 views

m.techxplore.com/...2017-12-ai-results.html

technology google nasnet ai

shared by jcunha on 11 Dec 17 - No Cached
LeopoldS liked it
  • dharmeshtailor on 12 Dec 17
     
    This has been around for over a year. The current trend in deep learning is "deeper is better". But a consequence of this is that for a given network depth, we can only feasibly evaluate a tiny fraction of the "search space" of NN architectures. The current approach to choosing a network architecture is to iteratively add more layers/units and keeping the architecture which gives an increase in the accuracy on some held-out data set i.e. we have the following information: {NN, accuracy}. Clearly, this process can be automated by using the accuracy as a 'signal' to a learning algorithm. The novelty in this work is they use reinforcement learning with a recurrent neural network controller which is trained by a policy gradient - a gradient-based method. Previously, evolutionary algorithms would typically be used. In summary, yes, the results are impressive - BUT this was only possible because they had access to Google's resources. An evolutionary approach would probably end up with the same architecture - it would just take longer. This is part of a broader research area in deep learning called 'meta-learning' which seeks to automate all aspects of neural network training.
  • dharmeshtailor on 12 Dec 17
     
    Btw that techxplore article was cringing to read - if interested read this article instead: https://research.googleblog.com/2017/05/using-machine-learning-to-explore.html
1 - 5 of 5
Showing 20▼ items per page