The way to understand learning and memory

In György Buzsáki's 2003 nature paper, he proposed that except for spatial prediction, firing pattern of peer cells will add to the accuracy of the prediction. They found that neurons get sychronized with their peer cells, forming cell assemblies. Basically these cells have a preference to while cell assemblies to join in, in turn you can actually catergorize the other cells into two: positive-weighted cells (cells that this neuron will always join to fire), negative-weighted cells(cells that this neuron don't join to fire).

The paper also related EEG to the predicability. I don't know how the EEG will contribute to the firing possibility or how much the firing of this neuron will change EEG. But it seems that cells only fire at the trough of theta wave. So to my knowledge, theta wave is the collective activity of the brain neurons or hippocampal neurons, which to some degree, can be understood as the background neural activity. Does the firing lying at the trough phase of the theta wave means distinct from background activity, which in turn can be dictinct and received by the subordinate neurons?

Also after reading and discussing this paper, I can draw a propose about how to understand learning and memory. Well, this in the end actually turned out to be the general thing everyone is trying to look up, but still, I'd like to describe it here:
Neurons tend to fire in assembly, or the firing of neurons tend to be sychronized or dissynchronized by peer cells. When several different sensory inputs get into the brain, different sets of neurons get fired. At first there's nothing specific happening between this set of neuron and that set. But when this simultaneous happening came up again, neurons from different sets are kind of getting related, or associated. With repeats, the synchronization set up. That's associative learning. If we try to understand in the molecular or cellular level, that is, the synchronization made the originally weak synapses or silent synapses got strengthened (namely LTP?), which vice versa, the dissynchronization make the originally strong synapses weaker (namely LTD?).

The good thing we assume here is that neurons tend to set up synapses with any other neurons. They lie there, do something, or not doing anything. Frequent electrical signals make them strengthened. When no electric signals coming to knock at their door, they just lie there silently. They are ready to go whenever the electrical order come.

But what about a different propose here? Synapses assemble in response to the need and disassemble when no longer needed. So how can synapses between different sets of neurons set up? Think about this, the electric signal go through the whole neuronal processes, which elicit chemical/physical changes in the processes. Once both the two neighboring neurons undergo the chemical/physical changes, synapse tend to form between them.

Well, I can go on and go on, become a total theorist. I need to go to read more literatures to support or dissupport my idea. (Only if someone can favor me more time.)