ensemble methods, sexual reproduction, regularization, order and chaos

https://www.youtube.com/watch?v=DleXA5ADG78

one idea is, the balance between order and chaos in giving rise to life.

another idea is ensemble methods performing really well - like the netflix competition winner. (is this related to monte carlo methods too? lots of weak solutions thrown together?)

in this talk hinton talks about why it might be good to have sexual reproduction in evolution. sex keeps reassorting chunks of the genome, so you never have a "single learner" in that evolutionary sense, but rather an ever-shifting ensemble of individually maybe not all that amazing learners, but always thrown back together into the same system.

he shows an example of how you can do the same thing within a deep neural network. you basically constantly turn on and off bits of the network as it's learning, so subsets of it are trying to learn some of the structure in the data, but the whole thing is never getting too committed, it's finding bits of *relatively* shallow solutions. and all these bits are learning in the context of each other (so they regularize each other, in that larger generalized space).

for one thing, it would be cool if the brain works this way, essentially living on the edge of chaos so the dynamics are always getting thrown around of which subcircuit is online during some experience (even at a microscale of time, like individual theta cycles for example). so the brain is keeping online a sort of chaotically scattered but somewhat balanced subset of stuff, always flitting between different hypothesis spaces and driving updates in those, within the global weight-space of all the learning it's done so far.

this might partly explain why individual trials of brain activity look so random, and doug garrett's data where dynamical *variability* in brain signals is related to behavioral performance.

and fundamental balance between chaos and order- the emergence of beauty and life is this being tilted with your weight ahead of your feet, the unresolving energy of the sampling dynamics kind of bootstrapping itself.

No guests remain within the dragon gate

regularization

"generalizing from finite samples in extremely high dimensions" is an amazing problem. think of life expectancy. let's say men born today expect to live 82 years and women expect to live 84 years (don't know the actual numbers). that's based on a lot of samples of men and women (plus some projections, but let's say for now it was just based on raw statistics of past events). OK. so if you're a man, your best guess for how long you'll live is 82 years, and if a woman, your best guess is 84 years. in other words, if you were doing optimal decision-making or whatever, this is the number you should factor in (well actually the variance could be asymmetric so you ought to integrate over the whole distribution but let's ignore that for now too). OK, *but*, let's suppose the expectancy for white-men is 85 years and for non-white-men is 77 years. maybe for some reason it's reversed for women; white-women 82 years and non-white-women 86 years. NOW, if you know you're a white man, you should use 85 years as your best guess. OK, but what if you factor in height? smoking? educational history? city of birth? environmental toxin exposure? genetics? .... pretty soon, there won't even be a *single sample* of experience to draw on, to make an estimate of the life expectancy in your particular category.

one way around this problem is to assume that each factor operates linearly and independently. then, you solve a big linear regression problem, and linearly interpolate/extrapolate to points in the space where you don't have any samples. note that even under these simplifying assumptions, there are serious problems. for example, when you have many many dimensions, some will by chance have extreme slopes -- which will push your extrapolated prediction into crazy values. but even ignoring those problems, there's the bigger problem that the factors aren't actually linear or independent. they interact very strongly in fact: sometimes one factor might even reverse its direction of effect depending on another factor. so the linear regression is not a good solution.

what can you do? you have to regularize the problem. this is, you have to use some prior knowledge to drastically pare back the number of effective dimensions. for example, you might know some variables are strongly correlated, so you can treat them as one. or you might know that some variables are dominated by others, so you omit the weak variables. or you might know that the underlying function ought to be at least locally smooth, so you impose a smoothness constraint.

but what is the right regularization? brains are somehow really good at this (algorithms are getting better quickly though..)

i was just thinking that the regularizing prior knowledge could be a good place to fit in scale-invariance and holo-stuff. i haven't read enough, but i haven't seen yet much in machine learning on using a single learning framework to accommodate all different kinds/levels of data. i imagine that's what the brain is doing (maybe it's related to the relatively "unitary" consciousness that we seem have subjectively) -- because we're essentially shifting around the focus of more or less the same global model to apply to massively disparate "kinds of things". so we have more samples to draw on for any given problem. we can even turn this global model partially on itself, which perpetually gives us even more samples and might also produce other weird features.

somehow, the circuitry of the cortex must be cleverly set up to do this kind of learning over space and time. (like hawkins' and others' "hierarchical temporal memories" type ideas).

in a friston-type framework, i'm picturing that the structure of the organism at all levels already encodes a deep regularization (for example RNA only interacts with some molecules). since there's no separation between what is "inference" and what is just dynamical structure of the "inferring entity" itself. if every dynamical system can be thought of as doing prediction (which means vastly generalizing from finite samples), then its whole structure from the ground-up is scale-invariantly encoding regularization for all the external states it's exposed to.

new ideas

Part 1

Some models suggest that the basal ganglia are used to select between actions. Some candidate actions are generated, and the circuitry of the basal ganglia allows one to win and be executed (like Jeff Wickens). I wondered if the basal ganglia are also used to select between "internal actions", like what to think about, what to attend to, etc. This could be connected to habits of mind. 

Part 2

I was thinking about Karl Friston's view of sleep. If I understood correctly, sleep is a time of lower top-down precision, when the experience collected during wakefulness can be used to drive big changes to the world-model. It makes sense to isolate your brain from your muscles when you're doing this deep restructuring: when you're awake you need high top-down precision to make your best active predictions.

[As a side note, I wonder how this fits with the data about interstitial volume increasing during sleep to allow "clean up" of metabolic products. Maybe a sleep-state is simply a useful time to do multiple unrelated things. Or maybe the physiological implementation of precision actually has to do with neuron-glia interactions and interstitial space (e.g. glutamate reuptake).]

The most obvious thing that happens during sleep, physiologically, is change in oscillatory activity. High-frequency activity mostly drops away, and low-frequency oscillations become prominent.

Does low-frequency activity correspond to the "subtle" state described by Eastern mystics?

In terms of oscillatory activity, a child's brain looks like a sleeping adult's brain. Roughly, a deeper stage of adult sleep corresponds to an earlier stage of child development. In childhood we also have high plasticity. This is when things seem really meaningful, like they can in dreams.

The adult mind can be pretty locked in place. Old feelings, even from early childhood, are still there somewhere, but maybe don't find much expression. Really deep *meaning* seems to be locked in there somewhere, too. Things can seem a bit abstract until it's in your gut. Truly believing in love isn't something the rigid adult mind is very good at.

What about meditation and the subtle stage of development?

Maybe meditation peels the layers off. Perhaps during development, something happened to hurt that level of the self: suffering (free energy) caused by direct contact with not-self. That then becomes a scar or a rigidity: some loss of dynamic flexibility like Eve Marder's ion channels. Specifically it's a hyper-assertion of the self-pattern at that level.

Meditation brings awareness to the thing that's creating the rigidity. Bringing awareness is the same as nesting the thing in a larger context, allowing it to healthily integrate and discharge some of its exclusivity. In meditation we also see slow oscillatory states.

This suggests that top-down precision is something we should look at in multiple layers. There might be a different kind of precision every time we peel off a layer.

Something like the aPFC or hippocampus might correspond to top-levels of the adult hierarchy. Interestingly, you also grow to a bigger model as you relax levels, because that slice of the self gets integrated with a bigger context. This might be like how unexpected uncertainty could convert to expected uncertainty as the model grows.

A tricky issue is how to reconcile this with the psychedelic state. Psychedelic states have been likened to dream states, and sometimes include a feeling of vast lovingness and meaningfulness. Is this playing the same role as sleep? If so, do psychedelic states show the same oscillatory signature?

Some studies have described increases in low-frequency oscillatory power with psychedelics, but more recent work (Carhart-Harris et al, 2013) showed decreases in oscillatory power across the spectrum (in default mode). Carhart-Harris suggests this corresponds to "destabilization" of brain activity, possibly allowing it to leave attractor states that it's been stuck in.

Part 3

Oscillations, again. Subjectively, people sometimes describe "the present moment". To some extent, every moment seems to be the "same moment". It's always the present moment. Of course in some ways this is vacuously true; it follows from the definition of "now". But I think it might also be describing some aspects of how the brain works.

Clark and I talked about "ontologeny" (he calls it "microgeny"), which is the "waking up" of the universe in each moment. It's possible that it sort of ultra-quickly recapitulates all of the development of the universe, up until now.

Clark wants to extend this to fundamental physics, which I don't necessarily disagree with, but being more conservative for a minute, let's say this is just an attribute of the brain. Within a brain rhythm, you keep doing the same thing over and over again. If this brain rhythm is one signature of the dynamic quasi-stable attractor state you're in, then it's probably repeating the same pattern over and over again. That's "you" -- the stability of that attractor is what makes up the coherence and consistency of your experience. But it's constantly "refreshing", going through all the phase-angles of that space. So in every moment it would feel like you're constantly waking up to that moment as you go through that same cycle. The attractor can gradually (or sometimes quickly) change, both through its own unstable dynamics and through sensory inputs, and there's also long-term plasticity. A wakeful attractor state would have some signature in alpha frequencies in the brain, for example.

This could explain why sometimes it seems like mental events are perpetually happening "before" the present moment of awareness. A simplistic explanation would be that there is some particular phase-angle that actually corresponds to the "waking up" or the present moment (and I suspect that's true in some partial ways), but a more complete explanation probably involves interactions at all the phase-angles. In either explanation, bringing those things into awareness (into the present moment) might involve an attentionally-driven shift in the relative phases of firing of different neural populations.

For me, subjectively, this corresponds to things that I feel were always "there" inside me somewhere. With some attention they can be brought up to "near" awareness, but it sometimes feels like they're still lurking before the edge of the present moment. Sustained attention brings them into the present.

I think this is also a key to the "hierarchical temporal memory" idea from Jeff Hawkins. Microcircuits in the brain must be set up to dynamically mirror aspects of the sensory inputs / the world. But as you go up the hierarchy of the brain, they extract structure. A lot of work has focused on the "spatial" structure they extract (like going from spots to oriented lines to edges to objects to recognizing people, etc), but they are implemented as oscillatory dynamical systems, so it makes sense that the real mirroring and abstraction is deeply spatiotemporal.

[Also, I think it's worth mentioning that we often use the visual system as an example for this type of thing, but a huge area of the brain is devoted to somatosensory, visceral, vestibular, etc processing. Further, I only talked about the sensory side of the hierarchy but the brain can at least as well be thought of as organized hierarchically around action, on the motor side. There are always rhythms in motion, too. Together I think this fits embodied cognition and the body-mind naturally into the framework.]

"Commit not a single unwholesome action,
Cultivate a wealth of virtue,
To tame this mind of ours,
This is the teaching of all the buddhas."
-- attributed to Shakyamuni Buddha

somehow this meant to me that we create the world (in its partial, divided, samsaric sense) through karma. or, duality *is* precisely our own karmic burden. which is why buddha said "i and all beings are enlightened".

zen & inference

i think i've said the same thing before, but this felt more clear as i was writing it.

one suggestion is that zen is about building a better model. that the vast majority of your model of the world is the sort of moment-to-moment dynamics of sensory prediction (including interoceptive, meta-cognitive activity, etc) ; and the model basically sucks because it ignores the vast majority of the actual structure of the world (you see what you expect to see). so zen practice is about relaxing the top-down priors to allow the model to incorporate more of the dynamics.

that doesn't feel like the whole story, but it's an idea to start from.

This discovery will be the one that tells me what invented me

Like any dealer, he was watching for the card that is so high and wild he'll never need to deal another

identification

one issue that's been bugging me is what "identification" means. i was watching another one of diane hamilton's videos, and she's talking about how identification with the small self is just one of many perspectives.

it seems to be true that "i" suffer when i identify with a particular small-self to the exclusion of other phenomena. when "i" release into a more inclusive perspective, the suffering-as-partiality of the smaller-self is revealed to just be a transient object, whereas before it felt like an existential/unsolvable problem.

but, what does it *mean* to expand identification like that? three broad categories of possibility:

1) both the smaller system and the larger system already existed before you released your perspective, and your change in perspective didn't really change either system. instead, it has something to do with how the dynamics of these systems are coupled to the aspects of manifestation that we agree to associate with the individual (e.g. control of your speech muscles to say "i am experiencing xyz").

2) the smaller system is being changed. dissolving exclusivity structures, so that it can arise and release and transform more spontaneously in the context it's in.

3) the larger system is actually being created in the process. when the smaller system is allowed to enter into increased dynamic coupling with other phenomena, then the system that calls itself "zeb" now consists of fundamentally new dynamics that only exist in the interplay between the previous small-self and the other phenomena. *however*, i'm not sure how this fits in the case of releasing to "identification" with all phenomena / formless awareness.

i feel like all of these are true, but i'm still missing something.

making pot illegal

http://www.nytimes.com/2014/01/03/opinion/brooks-weed-been-there-done-that.html?src=me&ref=general

i love brooks, and this is a fucking awesome article in a lot of ways.

i can't even say i categorically disagree with "seatbelt laws". we want to make policy that does the most overall good, and the disutility of restricting personal choice is just one component of that.

i think he misses two things-

1) some of personal growth comes from making stupid choices- how do you do that if all the stupid choices are illegalized?

2) it's true that you can find some good in illegalizing pot (and he does a brilliant job of outlining the shape of that good), but how can you have a complete argument without comparing the magnitude of that good with the magnitude of the concomitant harms?

when it comes to illegalizing pot, the magnitude of the good is really pretty meager. like one comment points out, the usage doesn't skyrocket when pot is legalized in places like holland. yeah, if you make it illegal people might do it a bit less. that is a good, but is a big one? i believe illegalizing heroin reduces the number of addicts. that's probably a bigger good. (still, how big?)

and he doesn't talk about the harm. making drugs illegal obviously creates shit-tons of practical problems. then there's the difficult-to-estimate disutility of inhibiting personal freedom. and another nebulous factor is slippery-slope into 1984. i'd have to think very carefully about how to put this harm in a common currency to compare directly to the good. yet my intuition is that the harm outweighs the good by several orders of magnitude here.

i want to be SURE

free energy hierarchies and pedagogy

last night somebody was spraying the sidewalk with water to clean it. but he was doing it at 6:30pm, the busiest time. he turned off the hose whenever someone was walking past (to avoid spraying them), and this meant he had the hose off 90% of the time. he was just spraying in little bursts when there was a gap in people.

i thought, shouldn't they do this at a different time of day? it's probably just poor management. if i were in charge, i would set it up differently.

then i thought, it's possible that they know something i don't know, and the best solution is actually to do it at 6:30pm for whatever reasons.

but this made me think about free energy and pedagogy.

in order to learn, i have to take action based on my current model and see the consequences. there may be a prediction error that drives an update of my model; then i involute this structure and the model gains richness.

people who take control of all aspects of their life in this way, fully inhabiting their current model to produce action and then incorporate the updates, probably involute the most of the world.

and "teaching" should based on this. getting students to take action based on their current model, in situations where they can incorporate the prediction error.

dopamine, progress, and tantrums

the real "dual-system" issue is probably the one that plato and buddha and whoever else understood from the beginning.

virtue or responsibility, that's the hard thing that's right to do.

the unvirtuous thing is perseverating the goals of some smaller system.

the relatively virtuous thing is what o'reilly said -- you're unhappy when you're frustrated in making progress toward your goal, so the way to be happier is to let go of that goal. in a little sense that means the smaller system you're currently identifying with has to "die" or "dissipate".

the subjective experience of this is "awareness of" that smaller system. identification with a perspective which is not limited to being *inside* the smaller system. because the goals are determined by the priors of the system, and now the dynamical system in question has another hierarchical level, so it behaves as if it's trying to minimize surprise at *that* level. (still unvirtuous from a broader context, but moving in the right direction!)

that's why i was excited by o'reilly's talk about tantrums. along with karl's model it feels like a step toward a scientific view of wtf is going on...

how stuff works

As these peaks in the holomapping and their concentric boundaries coalesce, stable conscious binding flourish-develops as manifold harmonic attunement thru complexificationary unitive intraperfusilusion in dynamical introjective identity with the All as it can be 'known' as consciousness.

-- Clark Potter

intelligence

intelligence is the ability to make accurate predictions about the

world. you can think of the mind as a *model* of the outside world. but of

course it's never a perfect model. the self is the model, and the not-self

is the aspects of the world not captured by the model. we're perpetually

afraid to relax the self, to allow the model to capture more aspects of the

world. there's a moment of "little dying" when we do this. the self never

wants to let go of the idea that *it is the whole world*, or in other words,

it is already a perfect description of the world. fundamentally, it has to

be this way, because the self is a homeostatic system that maintains its

integrity because it constantly enforces its own pattern against the chaos

of the outside world. but fortunately, the world forces little bits of the

self to relax, so the self can grow and become a more honest and faithful

model of the world (it might be fair to call this the universal principle of

"love").

that's why taking care of yourself emotionally makes you smarter. 

I have a sheep doing roofing over at my house.

Come and drop in. We'll put on Zeppelin and eat cheddar cheese.

free energy and suffering

the free energy framework provides a really natural explanation for what suffering is. "good" vs "bad" feels like such a basic quality of subjective experience, but neuroscience says the material correlate is dopamine or opioidergic neuron firing or whatever. that doesn't make sense, it's way too arbitrary. in the active inference framework, suffering is exactly free energy. it's the same principle for any system. to the extent that the system is trying to assert that "it is the universe", and the rest of the universe is shooting that theory down, that's the subjective quality of suffering.

of course, the concept of a "system" is just something we lay down post hoc to try to understand. you can draw a boundary anywhere you want and call the inside a "system" (although this might get increasingly pointless for weird boundaries). and then our concept of "suffering" would match up with the free energy of that "system".

this is a nice explanation for what happens in addiction, i think. when you look at the system-boundary around your relatively shallow ego, free energy actually decreases when you take drugs (or do whatever addictive behavior), because the drugs help this shallow system to live in a more isolated dream, where it's not exposed to the truth of the rest of the world which it's denying. but simultaneously, if you look at the system-boundary around some deeper identity of the individual, the drugs are making things worse -- creating more suffering. so at the same moment, there's either more or less suffering depending on what system-boundary you're considering.

inconsistencies

whether particular beliefs are inconsistent depends on your model of the world. fundamentally all beliefs must be inconsistent in some way because they're just partial truths. evolution in general, i think, is the process of steadily burning away these inconsistencies when they're forced to interact with each other. or in another sense, the inconsistencies remain (since, after all, they are what manifestation is), but their reconciliation also exists (which itself is another kind of inconsistency). so depending on how you look at it, more-and-more "stuff"/form is being created, or more-and-more "stuff"/form is being burned away to leave what it originally was, truth.

the transition from not-life to life

Q: do you think the "no-life" to "life" transition... that first step to the first living cell.... was a bigger and more complex step than all of the evolution of lifeforms that has happened since?

A: the free energy way of looking at life suggests that, 1) there isn't any kind of border between life and not-life. homeostatic dynamics arise within all systems, and just progressively increase in richness when possible. there were probably lots of intermediate steps between autocatalytic RNA sets and modern "cells". and 2) each additional step is only meaningful in the context that it arose in. the organizational level of, e.g. what we call "life" is just the tip of an iceberg of systems dynamics. the right way to think of the non-living world isn't as a static or dead -- it has its own rich dynamics, upon which the steps toward life are just little nudgings that reorganize those dynamics.

extending science to mind, heart, and spirit

this kind of science is fundamentally a bit different from classic science, for at least three reasons:
1) observations may be internal rather than external. in science, i do an experiment and i see that the litmus paper turns blue. then you do the same experiment and you see that the paper turns blue. we talk to each other, and we agree that we observed the same thing. but what if the experiment is to meditate for one year, and the observation is whether i feel calm? it's harder to agree on what we mean, because these internal observations are nuanced and deeply coupled to the act of perception itself that looks at them.
2) observations require more practice to see. we can never see the real world, only the inside of our own brains. if we "see" a particular phenomenon, like "a chair" or "an act of bravery", this means we have a predictive model that hasn't yet been shot down by the sensory evidence. but we can only see things that we have predictive models for. if caveman witnesses a traffic jam, he won't see "a traffic jam". at best he might be able to recognize that each "car" is an autonomous unit. likewise, autistic people probably can't see "extracting oneself from an awkward social situation", yet for other people this is as clear as "a chair". in a science of spirituality, observations of things like "mindfulness" would require the observer to actually have the predictive models for the things being observed, which requires a lot of experience and practice.
3) finally, these experiments might require people with strongly developed compassion and mindfulness to perform the manipulations, for example to provide therapy to psychopaths. such people are rare, and it's difficult to identify them.

when i got back from the airport last night, everything was closed (the 26th, "boxing day", is a holiday in britain), so i ate at subway. in the middle of my sandwich, a young employee came up to my table and said, "you have to move in 30 seconds". i was like, "what?". he said, "you can't sit here now, the chairs have to be moved". my first reaction was, ok, i see what you mean, but you're being really rude about it. i was preparing to be a dick back to him, but then i looked at him and it struck me that he seemed like he might have asperger's. for him, it's just a mechanical job: move the customers so he can move the chairs. he knows when he needs to move the chairs (i.e., 30 seconds from now), and i'm in his way. so i collected my luggage and as i was leaving, i looked in his eyes, gave him a smile and said "have a good evening" as warmly as i could. this is on the theory that no matter how weirdly the brain is wired, there's something human inside most of us, which you can find if you look hard enough. and it felt like there was something in him that responded to being treated nicely. probably he normally gets very bad reactions from people, and this just reinforces his isolation. in this sense, mental illness is a deeply social phenomenon, involving interpersonal feedback loops.

but what i think is even more interesting about this is why i decided not to be a dick to him. it was because i thought "he has asperger's, so he needs some special help". but the really important thing is this: *anybody* who's being a dick to you, needs special help.

100 years ago, we didn't know about autism or schizophrenia. naturally, people with these mental illnesses would do things that hurt us (i.e., rudeness, socially inappropriate behavior, violence, etc). and our reaction was to label them as fools, people of low moral character, imbeciles, etc. we would lock them up or even execute them for their crimes.

now, we recognize that it's an illness and they need help. but the point is that there's nothing magical about our current diagnostic categories. in another 100 years, we'll have probably identified the neural underpinnings of psychopathy, antisocial personality disorder, and even the things that make people commit mass killings. then, rather than saying, "you're fundamentally a bad person and you don't deserve another chance", we'll say, "at the core you're a human being, you're struggling with extraordinary suffering, and we'll do everything we can to help".

but even right now, if we want to, we can act based on this general understanding. when someone does something bad, they are the ones who are hurt the most. the suffering that a victim feels is nothing compared to the suffering that the criminal feels. i know this from my own experience-- doing something bad arises from a deep suffering. it's the exact opposite of being kind to yourself. and having done the bad thing, it becomes even harder to admit this weakness to yourself, because you have to take responsibility for what you've done. in the people who commit mass killings, i think this guilt is locked away so deeply that they literally feel like they would die if they let themselves feel it.

"Let me be thankful, first, because he never robbed me before; second, because although he took my purse, he did not take my life; third, because although he took what I possessed, it was not much; and fourth, because it was I who was robbed, not I who robbed." -- Matthew Henry