YONSEI BK Seminar
2021. 11. 10. 3PM
Research in Neural Networks Leads to New Combinatorial Identities
Eric Dolores Cuenca
To understand artificial neural networks we study the flow of signals on biological neural networks. If we use an external stimuli, like a sound, to force a network of neurons to fire signals, the feedback coming from the neurons will perpetuate the flow of signals. Can we predict if after 2 minutes a fixed neuron will still be firing without replicating the whole network? Using a mixture of topology and combinatorics we classified a family of cascades of signals. Among the consequences to our work are new identities of binomial coefficients and efficient algorithms to predict if particular neurons will fire after a fixed number of iterations.