Synapse Information Processing
In the Neural Engineering Framework (NEF), synapses are:
- Leveraged to provide filtering of spikes at different time scales. This allows for managing heterogeneity in neurons, as well as allowing to trade-off speed of response for information integration over time.
- Used to create dynamical systems using these filtered spikes including oscillators, integrators and temporal memory networks.
In all of these cases, information can be lost. However, this is always framed as a trade-off between informational integrity (which can be measured various ways, such as how well a signal is classified, how well a dynamical system is approximated, how much information is compressed over time) and resource use.
Retransmission as Information Recovery
In the context of the brain, TCP doesn't really make sense. TCP assumes there is a copy of the information that is desired to be transmitted somewhere, but that's usually not the case in the brain. That being said, there is something called the phonological loop which appears to be a short term store of what was heard. Consequently, you could interpret this replay as a retransmission and consider it to be information recovery.