Research area: Development and application of network reconstruction methods.
Networks produce dynamics: the spike trains of neural networks, the protein expressed in a cell, the number of visits to a youtube page, and the stock-share prices, are of the more familiar examples. In many cases, the dynamics are observed, but the underlying network is unknown. What can we infer from the observed dynamics about the connectivity of the network? This is the central question in network reconstruction.
While it is an important and interesting question (its application range from better investments in the stock market to actually understanding how the brain works), it is also a very hard problem to solve analytically, as the number of possible networks that can produce the same dynamics grows exponentially with the number of nodes. There exists, however, approximated methods in use, and an abundance of new methods that are being developed. It is important to note that these methods, as well as improving accuracy and efficiency of the network reconstruction, also address the interesting question causality, and the relationship between effective connectivity to real connectivity.
Navit is currently working on methods to improve accuracy and efficiency of network reconstruction. She is also collaborating with Dr. Paolo Bonifazi and Prof. Ari Barzilai from Tel Aviv University to apply these methods to calcium imaging of in vivo neural networks.
- Dori, N.; Menon, M.; Kilian, L.; Sokolowski, M.; Kronik, L.; Umbach, E.; (2006) Valence electronic structure of gas-phase 3, 4, 9, 10-perylene tetracarboxylic acid dianhydride: Experiment and theory, Physical Review B 73(19):195208