Discipline: Computer Sciences and Information Management
Subcategory: Computer Science & Information Systems
Aswan Benjamin - Hampton University
Artificial Neural Networks are made to mimic the function of neurons in the biological system and can be combined with evolutionary algorithms to allow nano-robots to exhibit learning capabilities. There are two main types of artificial neural networks, the Hopfield neural network and the backpropagation neural network. The Hopfield network employs a group of neurons connected to every other neuron by negatively and positively weighted synapses. The main benefit is communication between every neuron, which can be used to store information. The Backpropagation neural network uses separate layers of neurons connected by synapses, allowing the network to achieve higher efficiency by propagating the error percentage backwards after running the network.
The network that I modeled was of the Hopfield variety and it is created by plotting a group of neurons that are connected by randomly generated synaptic weights. Each weight can be negative or positive and this contributes to the total efficiency. The strengths of synapses are shown through thickness of synaptic connections and negative or positive value is shown through color of synapses. Although the efficiency in this model does not keep rising due to random synaptic weights, if the network was combined with an evolutionary algorithm the robot would become more efficient because it would be able to learn from mistakes and previous experiences due to communication between each neuron. In future work, other neural networks will be investigated and combined with evolutionary algorithms to find the best network to employ as the primary learning tool for nano-robots.
References: Tani, J., Maniadakis, M., & Paine, R. W. (2014). Ten understanding higher-order cognitive brain mechanisms by conducting evolutional neuro-robotics eperiments. The Horizons of Evolutionary Robotics, 219. Karayiannis, N., & Venetsanopoulos, A. N. (2013). Artificial neural networks: learning algorithms, performance evaluation, and applications (Vol. 209). Springer Science & Business Media. Pomerleau, D. A. (2012). Neural network perception for mobile robot guidance (Vol. 239). Springer Science & Business Media. LudoBots. (2013). http://www.uvm.edu/~ludobots HechtNielsen, R. (1989, June). Theory of the backpropagation neural network. In Neural Networks, 1989. IJCNN., International Joint Conference on (pp. 593-605). IEEE.
Funder Acknowledgement(s): This research was funded by NanoHU: Nanoscience Transforming STEM Education at Hampton University, funded by National Science Foundation HBCU-UP ACE Implementation Award HRD-1238838.
Faculty Advisor: Moayed Daniel Daneshyari,