This is an excerpt from my email I wrote regarding next steps:
1. Make a diagram of the conception of the system as it was before, including how I was thinking of links between high level neurons having different weights for each high level feature (which may be made up of a vector of lower level features, eg. histogram). This would be the context in which to situate discussions of machine learning systems.
2. Read more about LIDA to understand better how percepts and knowledge from memory stores are structured, and if a subset of the architecture can be used.
3. Wish list for machine learning / clustering (in no particular order)
• Ideally statistical (stores a distribution of patterns, not all the patterns themselves, but this is flexible)
• Output could be manifest in connection weights between neuron-like structures that represent higher level concepts. (concept as distribution of activation) Each item should belong to a concept to a degree, not Boolean.
• If it is deep learning and I can forgo the feature extraction, it must be elegant (simple enough for me to understand) and be feasible for high resolution colour input images.
• If it is deep learning it should be generative: can learn high level abstractions (concepts) from sensory data, and also construct sensory data from those high level abstractions.
Thank is all I can think of off hand.
I was thinking about working memory and long term memory in relation to dreams. Does it make sense that latent activation in working memory could “cue” (LIDA parlance) memories in long term memory (or transient long term memory), and continue to activate long term memory structures? How does this effect memory? Presumably short-term memory is impoverished in dreams, could this be explained by activations happening only in long-term memory? Or does this make no sense at all, as long-term memory can only be activated through its transfer to working memory? Kosslyn’s proposal is that memories are encoded using “population coding” in LTM and only “explicitly” available once they are decoded into the VC. Perhaps dreaming is a ping-pong between LTM and working memory, where LTM activates working memory structures, which activate LTM structures, and so on.
Now I’m getting confused between cognitive and neurological levels of description. Now that I think about it, all the dreaming models I have come across are either simplistic (dreaming as no more than random activation of memory) or described on the neurological level. A model of dreaming at the cognitive level would certainly be worth looking at, which may make sense of some of these issues with LTM and working memory. Is working memory equivalent to short-term memory?) If dreaming is a loop reading from LTM and writing into working memory, what makes some dreams remembered and others not?
Concrete example A: I rarely remember dreams, and I presume this is due to not waking up during REM much (but my dreams do appear quite NREM like in general). When my partner had a concussion I had to wake her up every 4 hours to make sure everything was ok. During that period I almost always remembered my dreams, since I woke up during them.
Example B: I remembered a dream this weekend, it was a platform video game, not one I had seen before, but one constructed in my dream. It had a particular visual design and some strange elements, like blue force fields. The thing is, I don’t play video games much, maybe supertux once a month (and it had been weeks since the last time), and this game looked nothing like it. Also I had not even thought about video games for some time. Perhaps this dream could have resulted from something similar to the following:
1. Day time experience primed working memory
2. The dream was initiated and a loop before LTM and working memory started.
3. At some point the integration of LTM memories in working memory constructed the video game dream (based on long term components, but a long way, in associative space, from the primed memories).
4. Perhaps I awoke during the dream, moving it into transient LTM, and as I thought about it while waking, into LTM proper. (Which begs an empirical question, can you remember a dream without waking up around/near REM sleep? what about NREM dreams?)
It seems to me, the degree to which this dream is divorced from daily experience shows that dreams are inspired or composed of components from LTM, not just working memory.