Artificial Nervous Networks Can Not Be Modular Until Expected

Artificial intelligence research continues to question the basic acceptance of how nerve networks work. A study tests the opinion that these networks are generally accepted in the form of “no lice” and offers interesting findings. The researchers reveal that networks actually have a more complex modarite structure.

In traditional sense, individual neurons in nerve networks ** contribute to more functions** leads to counting these networks not modular. The new research examines this situation in three different perspectives:

• Unit-level modarite: It controls whether different calculations are performed completely in separate neurons
• Ortogonal modarite: It researches that the network weights do not appear and exit the independent subspaces
• Herarchy structure: It defines and tests three levels as an inmelically nested hierarchy

The fundamental claim of the research is to look at the very narrow perspective of existing analyses and to illustrate the level of modarite in nerve networks. While conflicting at the neuron level is observed, it is possible to have functionally split zones of the network, looking at the wider matrix.

This finding is important for us to better understand the architecture of artificial nerve networks. The presence of modular structures can facilitate interpretation, controlling and developing networks.