10.6084/m9.figshare.247218.v1
Randy Olson
Randy
Olson
David Knoester
David
Knoester
Arend Hintze
Arend
Hintze
Risto Miikkulainen
Risto
Miikkulainen
Christoph Adami
Christoph
Adami
Research proposal: Darwin vs. DARPA: Evolution of a Neural Controller to play Tetris
figshare
2013
synapse
evolution
DARPA
Markov Network
evolutionary computation
Software Engineering
Applied Computer Science
2013-02-19 23:31:06
Journal contribution
https://figshare.com/articles/journal_contribution/Research_proposal_Darwin_vs_DARPA_Evolution_of_a_Neural_Controller_to_play_Tetris/247218
<p>The goal of the DARPA-funded SyNAPSE project is to build a microprocessor-based machine that mimics many of the characteristics of biological neural networks (i.e., high connectivity, synapse plastic- ity, and scale). The proposed device is termed a neuromorphic machine, and recent simulations have indeed approached biological scale. However, while SyNAPSE has been demonstrated on a variety of tasks, from optical character recognition and classification to control of quadrotor helicopters, it has done so at a cost of $42 million. In contrast, the BEACON-funded DvD (“Darwin vs. DARPA”) project has demonstrated the evolution of logic circuits that outperform SyNAPSE on optical character recognition for 1/1000th of the cost. In this phase of the DvD project, we now turn our attention to a more complex task: Evolving logic circuits that play the computer game “Tetris.”</p>