Assessing where and how information is stored in biological networks (such as neuronal and genetic networks) is a central task both in neuroscience and in molecular genetics, but most available tools focus on the network’s structure as opposed to its function. Here, we introduce a new information-theoretic tool—information fragmentation analysis—that, given full phenotypic data, allows us to localize information in complex networks, determine how fragmented (across multiple nodes of the network) the information is, and assess the level of encryption of that information. Using information fragmentation matrices we can also create information flow graphs that illustrate how information propagates through these networks. We illustrate the use of this tool by analyzing how artificial brains that evolved in silico solve particular tasks, and show how information fragmentation analysis provides deeper insights into how these brains process information and “think”. The measures of information fragmentation and encryption that result from our methods also quantify complexity of information processing in these networks and how this processing complexity differs between primary exposure to sensory data (early in the lifetime) and later routine processing.

Deep learning (primarily using backpropagation) and neuroevolution are the preeminent methods of optimizing artificial neural networks. However, they often create black boxes that are as hard to understand as the natural brains they seek to mimic. Previous work has identified an information-theoretic tool, referred to as $R$⁠, which allows us to quantify and identify mental representations in artificial cognitive systems. The use of such measures has allowed us to make previous black boxes more transparent. Here we extend $R$ to not only identify where complex computational systems store memory about their environment but also to differentiate between different time points in the past. We show how this extended measure can identify the location of memory related to past experiences in neural networks optimized by deep learning as well as a genetic algorithm.

Summary form only given, as follows. An intense-beam gyro-traveling-wave amplifier (gyro-TWA) experiment has been carried out at 35 GHz, using a 1-MeV, 200-A beam from the VEBA pulseline accelerator. The experiment was carried out in a 10.8-mm-diameter drift tube and used a 6-mm-diameter solid electron beam with low initial traverse momentum and very low initial parallel velocity spread. The electron transverse momentum required by the gyro-traveling-wave interaction (β⊥~0.3-0.4) was induced by transit through a one-period untapered bifilar helical wiggler magnet with a 4-cm period. A 35-GHz signal was launched in the TE₁₁, mode using a directional sidewall coupler. A massive iron collar (6-1/4-cm diameter, 6-cm long) was placed around the 1.27-cm outer diameter of the stainless-steel drift tube as a means to exclude the applied axial magnetic field and thus to control and vary the axial extent of the interaction. In experiments operating at 8 kG, near grazing incidence between the dispersion relations of the beam and waveguide modes, the the gyro-TWA demonstrated linear growth rates of ~2 dB/cm, total gain exceeding 30 dB, and an output power of ≳20 MW

Summary Form only given, as follows. A 35-GHz gyroklystron experiment has been designed using a 1-MeV, 100-A, 55-ns electron beam from the VEBA pulseline accelerator. The 10-MW design uses two TE/sub 111/ bunching cavities and a TE/sub 121/ output cavity. A helix gun approach is used to generate an electron beam with a low velocity spread (<5%) at a design pitch ratio of approximately 0.75. The operation of the oscillator is described in detail. The locking signal is provided by a 20-kW K/sub a/-band magnetron and is used to drive a slotted cylindrical TE/sub 111/ bunching cavity with a Q of approximately 200. The experiment has been fabricated and cold tested. In initial experimental operation, free-running oscillation at 35 GHz has been observed at about 5-MW output power. The TE/sub 112/ mode of the bunching cavities has been observed to oscillate early in the pulse, but generally not during the voltage flattop, so that it should not interfere with the phase-locking process.