1st January 2006 Repetition
Based on 15th July 2005 idea to always reflexively react to boring conscious repeated stimuli a series / pair of A’s never becomes an S-Habit. In the sequence
Loc. 1 2 3 4 5 6 7 8 9 10
Stim: A A A A A B A C A A
Resp: a b x x x
At locations 4 and 5 it continues to respond randomly with a or b (=x). It made A permanent at location 4. At location 7 the A starts S-Habit=1 expecting another A but this fails. At location 9 the A starts S-Habit 1 expecting another A and this succeeds but because the next stimulus was boring and consciously repeated it responded reflexively. But now have 2 A A’s in memory. Recall of an A needs to not process previous A A’s once one found. It could store this AA as S-Habit =1 with output response x. When next AA occurs it will start S-habit=1 expecting second A and it will succeed and there will be again a reflexive response.
1 A A -> 1 1
x x x x
Each time you store the 1 with a reflexive response there is no recall done on 1 and so can have many S-Habits=1 in memory.
Good and bad are graduated stimuli. I.E. feelings are generated stimuli because a change in the stimulus is also used as a stimulus – i.e. getting better and getting worse are recognizable. So pain and pleasure are graduated! What about novelty and familiarity? If you introduce similarity and have multiple features then similarity could be a graduated scale.
4th February 2006 Thinking
“Thinking about what to do” I am in the living room and feel thirsty. I think about what I can do to get a glass of water – I think I could get up and go to the kitchen and get one. – I then think of going to the powder room which is closer and drink from the tap. I then remember that yesterday I called to my spouse who was in the kitchen when I was feeling hungry to bring me a cookie and she did. Depending on my current context, how I feel (energetic or lazy) and the recency of which of these solutions I have done in the past I will decide to act.
Play is a more mature form of exploration.
11th February 2006 Responses
Output consists of amounts of changes to make to the output device – not absolute positions to move to – continuous feedback is used to determine if the goal (position) has been reached.
28th February 2006 What to do next? See 12th Jan 2005 list
- - find code not being used
- - get good bad feelings working
- - fix recognition of multiple simultaneous stimuli – kinesthetic
- - add graduated stimuli
- - multiple actuators – output devices
- - sense of time
- - thinking
- - thinking before acting or concentrating override reaction.
- - subconsciously running A-Habits.
I decided on a slightly different simultaneous P-Habit structure from June 17th 2005.
3rd March 2006 Parallel Habits
Question: When does similarity occur with recognition of P-Habits where the source stimuli are quantized – absolute (not relative or graduated)?
1st occurrence A & B -> AB
N N N N=Novel
2nd occurrence A & B -> AB
N N F F=Familiar
A and B individually are still novel because not conscious of them – not paid attention to them – not in LTM.
A later occurrence A & C -> AC
N N N
AC P-Habit is novel – newer experienced before.
Next occurrence A & C -> AC
N N F or similar? to AB?
If A becomes familiar the 1st occurrence of
A & C -> AC
F N N
And 2nd occurrence
A & C -> AC
F N F or similar? to AB?
Question: Do all the component stimuli need to be familiar to have the P-Habits labelled as similar? Or is AC similar to AB even if A is novel?
Note that a P-Habit can be familiar while its parts are still novel individually.
4th March 2006 Primitive stimuli
When a primitive sensory stimulus is novel it will always be different – unexpected.
When a P-Habit is novel it’s the combination of its parts that is novel and at least it or some subset of its parts are unexpected.
When an S-Habit is novel it failed to get any of its next expected stimuli.
If a primitive sensory stimulus is unexpected it could be because it’s novel or familiar – it’s just different from expectation.
If a P-Habit is unexpected then all of its parts were unexpected – the parts may or may not be novel.
If an S-habit is unexpected it failed to get any of its next expected stimuli – even though what was received may or may not be novel.
5th March 2006 Change and unexpected
A change is detected by getting the unexpected. For a P-stimulus this can be done by comparing it with the previous stimulus. It’s equivalent to having an S-Habit that is a loop expecting the same stimulus again active on every sensor. When it succeeds (in getting the expected) it has looped and no change has occurred. When it fails we have a change in expectation. This is worth paying attention to.
For a P-habit, change is detected for the largest part of the P-Habit that has changed from the previous one. This is the AND of the P-stimuli that have changed.
For an S-Habit change is detected when all (AND) of active (expecting) habits get something different from what they are expecting. As long as one completes we get the expected. If none complete, not all fail i.e. at least one is continuing then we are still expecting – doing something.
6th March 2006 Attention
So final attention algorithm is:
- any novel S-Habits
- any unexpected S-Habits
- any completed S-habits or A-Habit result
- any unexpected P-Habits or sensory Primitive Stimuli (novel ones are unexpected)
8th March 2006 S-Habits
Because we generalize before we specialize when it comes to S-habits we only react after we have re-experienced the sequence. So instead of:
The first scenario was based on when getting the second “A” we know what to expect and it would be boring to get it again. But the AB sequence is what is being expected – recognizing the general pattern before either of its parts – just like the 1st repeat of a P-Habit. So in the second scenario the 2nd B confirms the sequence – got expected sequence – it is now ready to be recognized as a unique pair – collapsed into an S-Habit if occurs again.
9th March 2006 S-Habits
The 2nd scenario is actually:
This is because the AB has to be learnt and on the 2nd B the BA has to be learnt before starting to react to A.
10th March 2006 S-Habits
I realized that when an S-Habit is started based on a permanent trigger then it produces a list of S-Habits to expect all the possible results. If all fail in the group then we should store the trigger and start recognizing (pay attention to – start a habit based on) the result stimulus that was unexpected.
The first set of stimuli from sensors should be used to “prime the pump” so to speak. They should populate the current stimulus list so that differences / unexpected stimuli can be detected. This 1st set of stimuli should not be used as a novel stimulus and not put in LTM.
Should only have to produce higher level P-Habits if the 2 source stimuli are different – not produce all the combinations of source stimuli. This is the same strategy being used for Sequential S-habit recognition.
16th March 2006 Permanent Stimuli
Right now a pair of stimuli are collapsed into a recognized pair when the 1st in the sequence is permanent and the 2nd one has occurred a second time. (At this point no action is done based on 2nd stimulus if it is not permanent because it is absorbed into the collapsed pair.) Maybe should not collapse pair until the 2nd is also permanent? Or perform action based on the 2nd stimulus as the action for the collapsed stimulus?
17th March 2006 Unexpectedness
From a recent Reader’s Digest article:
“This difference between the genders is, researchers say, a result of expectations and mindset. The women in the test group – and, by extension, women in general – were generally sceptical and didn’t expect the cartoons to be funny; when they did laugh, they were surprised, setting off a “reward” in the brain. Men, on the other hand, expected to be amused; when they weren’t, they experienced disappointment.”
Novelty is currently a feeling rated at 2 and 0 is familiar. But decisions & attention are based on unexpectedness – not novelty. Is it possible novelty is a feeling only generated dynamically when an experience / stimulus is not found in LTM and learning takes place based on expectedness and getting the unexpected? – Not reducing novelty to familiarity? If so the change of state from unlearnt to learnt is not a change from novel to familiar and it can be removed from the feeling property associated with all stimuli.
22nd March 2006 Expectation – Reward
Possible Expectation – Reward scenarios
1st Time 2nd Time
Trigger Result Experience -> Expectation Same Trigger Result Experience
Novel NorF Novel No Exp. 2 Famil s->y Famil Learnt 1 Expect
Famil s->x Novel Novel Surprise 2 Famil s->x Novel Novel 2
Famil s->y Famil Exp. Learnt 1 No Do
Famil s->x Famil Unexp N 2 Famil s->x Famil Unexpect N 2
Famil s-> Reward R 4 Famil s->x Novel Novel 2
Novel Reward R 4 Famil s->y Famil last reward 3
4 Famil s->x Famil Novel 2
S -> = doing A or S-habit 4 Famil s->y Reward R 4
y or x=succeed or fail 4 Famil s->x Reward R 4
xResult Novelx, (Trigger Novel) or Habit fails = Experience Novel
Result Rewarding = Experience Rewarding unconditional
Habit succeeds = Experience Learnt – reduce by 1 level
Reflexive response = Experience Novel at least to start with
24th March 2006 Algorithm
A slightly modified algorithm is:
- Trigger Novel or Reflexive Response – No habit, Experience Novel but then changed to feeling of result if it is reward/wanted/ feared or punishment & Trigger gains same -1 / +1
- Habit fails – Experience Novel but then changed to feeling of result if it is rewarded / wanted / feared or Punishment
- Habit succeeds – Concentration -1 unless result of feeling is better
- All Habits continuing at same concentration level.
If performing a habit has a concentration level equal to expectation feeling of last time it was done.
If result is different from expected – have at least novelty even if result familiar or novel. If result has a feeling compare with expected – if greater then habit mark wanted, If equal to expected feeling mark the same – if less than expected reduce experience / habit feeling (to similar / learnt) by one level.
Else If result is same as expected then experience / habit is same as result feeling.
When you change the experience / habit feeling then also set the trigger feeling to the same level.
25th March 2006 Playing a card game
Question: Why do we play a computer card game? Obviously to win! But what reward is in winning? It’s the avoidance of failure! or the feeling of success. But success at a task alone is not good enough. This just means it has been learnt. Winning means we avoided “not winning” or avoided loosing and loosing is punishing (not wanted) because we don’t get the expected result. So we try and avoid getting an unexpected result.
There are nine levels / degrees of feeling (emotional)
4 Good rewarding stimulus food - pleasurable
3 Wanted expected reward penny
2 Novel not in LTM new experience – 1st time
1 Interesting expected novelty TV, book, entertaining source of novelty
0 Neutral Permanent no longer react to
-1 Uninteresting expected boring Opera
-2 Boring Familiar – in LTM repetitive experience e.g. TV Ad
-3 Feared expected punishment house fire
-4 Bad punishing stimulus painful cut
“Ignore” is something done, like avoid and pursue / do. Ignore is not a feeling. “Similar” is partial familiarity that occurs with P-habits that share one or more features.
26th March 2006 Like and dislike
What about Like and Dislike or Love and Hate. Interesting level of feeling should be given to any P-Habit that can be done (learnt) but for which have no motivation. So learnt does not end up with boring unless it is learnt and boring and do not want to do. Therefore two kinds of Familiar – Familiar learnt and can do = Interesting and Familiar learnt but boring = don’t want to do.
28th March 2006 Feeling Scale
The new feeling scale is:
1 Interesting = Novel
-1 Uninteresting =Boring
30th March 2006
27th April 2006 Interest
I realized today that the difference / change in a stimulus results in interest – thus the change is the feeling of interest. Every stimulus comes with a feeling that is either interesting or wanted and the level of interest is the amount of change or difference from the previous stimulus when graduated stimuli are being considered.
Novelty / change and difference all cause interest and the magnitude of the interest is the same as the amount of change. Some stimuli instead create a feeling of good / bad whose magnitude is the same as the amount of wanted-ness / feared-ness.
30th June 2006
Details about the Canadian Revenue Agency’s SR&ED tax incentive program are here.
31st July 2006 Emotions
While at CogSci2006 conference I realized that emotional feelings like punishment and reward are attached (as a result of the brain interpreting them) to absolute values of stimuli and not to changes in the values of stimuli where it is a graduated stimulus. Changes are differences and these are interpreted as interesting / boring.
Now ready to separate out emotional feelings from interest from measure of success / failure of a habit. A habit can succeed and be marked as learnt because it got its expected result (succeeded in its own right). It can succeed because the result was interesting (caused a change) or it can succeed – be learnt because the result was rewarding. And similarly for failure in reverse / reflection. This gives 3 mechanisms for marking a habit as learnt / worth repeating. So upon executing a habit there are 3 pieces of information to process with respect to the result.
- meets expectation or not – absolute value
- the change in the result
- the emotional feeling of the result
No, it is more complex than that. The habit has an expected absolute stimulus result and an expected change amount. Or does it have one and not the other or vice-versa?
There are two types of values that can be associated with a stimulus. A penny is a physical thing that has an emotional feeling value of being good and can be used to signal success in an informational way.
A clicker associated (using classical – Pavlov conditioning) with reward is not physical but occurs as an event (not collectable) but conveys only an informational signal that can be used to signal success – Same as the words “yes” and “no” do to a child after they have done something.
You can also signal yourself to learn an action (mark success of a habit) by using the matched expectation feeling of success – The “yes” I did it or the “no” I didn’t get the expected result.
So if we only have one sensor that produces a graduated stimulus every cycle, the sensor will produce 2 stimuli – The absolute value of the stimulus (symbol) and the difference from the previous (change +ve or –ve with an amount). Adaptron can attend to either or both as a P-Habit.
Concentration level is the amount of change expected when performing an A-Habit and if the change received / resulting is greater then attention is attracted away from the habit resulting sensor.
3rd August 2006 A-Habits
Most systems using production rules save the ones that have proven to work and reuse them (bring them into working memory) if they have a chance of being applied to satisfy the current goal. In Adaptron an A-Habit that has got its expected result has been learnt and there is no interest in repeating it since the outcome is known and the Habit proven to work. They will become reused when the trigger stimulus is a P-Habit and can be used as in generalization when the newly experienced stimulus partially matches the trigger and the action is performed.
The clicker or the “yes” can gain its signalling potency because it is associated with a reward via classical conditioning. It occurs at the same time as the rewarding incident or it precedes the rewarding incident. At the same time it forms a P-Habit and gets generalized as indicating a reward. Preceding the reward incident it forms an S-Habit which also should get generalized as indicating a reward! Currently 2 sequential stimuli forming an S-Habit or not, don’t generalize such that when the 1st of the 2nd occur, the action of the second is performed. What about the 1st being given the “feeling” from the second since the 1st is a clicker it only needs to provide an expectation of reward. The clicker loses its positive signalling when it is misused – occurs at random after no action was done or when it was associated with punishment of some kind.
Trying to sort out these different emotional feelings
S Trigger see it and want it (good) and if action done expect it.
Action can get it, own it, can then use it (read / eat / trade)
S Result Sensory rewarding – pleasurable or interesting – source of novelty
- wanted (good)
A penny can see it - wanted – expect wanted if used
Get it - owned (rewarding)
Use (trade) it - results in wanted result
A book can see it - wanted – expect novelty if used
Get it - owned (rewarding)
Use (read) it - results in novel / interest / different
A food item can see it - wanted – expect pleasure if used
Get it - owned (rewarding)
Use (eat) it - result in sensory reward / pleasure
A click can hear it - signal success if S result
Can’t get it
Can’t use it
28th August, 2006 A-Habit Learning [not in red book]
The trigger for proceeding with the execution of an A-habit could be just one of the two stimuli that comprise the P-Habit trigger being experienced, provided the last A-Habit experience was rewarded or interesting. This is the basis of generalization. The reward / interest of the A-Habit is equally assigned to its trigger stimuli.
If the trigger is an S-Habit then maybe the first of the two source stimuli can trigger the A-Habit (i.e. Pavlov). Maybe only if the S-Habit is learnt?
Success and failure are the information signals that are used to mark an A-habit as learnt or novel. They are based on whether the expected result occurred. Success says the expected result did occur and failure says it did not. However a success does not guarantee that an A-Habit is worth repeating. It is the reward obtained from the result that determines this. In the case of interest/novelty success may mean that the result obtained is now familiar and the A-Habit is not worth repeating. On the other hand a failure to get the expected result may produce a novel A-Habit which is worth repeating.
So it is the reward / interest obtained the last time that makes it worth doing again when triggered. If we just deal with interest due to unexpectedness the 1st time the result is unexpected and we have a novel A-Habit. The second time it is done because there is interest in the novelty. Success occurs when the same result occurs but the interest level gets reduced because there is no novelty – the result is familiar. Thus the only result is exploratory behaviour. Once the environment is explored boredom sets in. However before the second occurrence occurs the trigger stimulus of the novel A-Habit is wanted. Should there be a previously learnt A-Habit that results in this trigger stimulus and this A-habit’s trigger occurs, then the A-habit should be done because its result is wanted.
E.G. Trigger A, Action X, Result B the 1st time is a novel A-Habit.
Then Trigger A is followed by Action X, and Result B the 2nd time marks the A X B, a learnt and not interesting A-Habit.
Then Trigger B, Action Y, Result C the 1st time is a novel A-Habit.
Should Trigger A occur then the A X B A-Habit should be done due to the interest in getting to B in order to repeat / learn B Y C.
This is using the learnt / successful A-Habits.
The trigger B must be given a wanted feeling when the BYC sequence 1st occurs. If C is rewarding (not just interesting) the B is given a rewarding feeling. This is what happens to make a penny have value. As a trigger in A-Habits a penny continues to have results that are rewarding. This is referred to as the backward propagation of feelings. The same backward propagation of interest takes place.
Thinking on the other hand is a forward simulation (propagation) that explores learnt A-Habit sequences seeking out interesting or rewarding results.
29th Aug, 2006 From Steve Grand’s website: [not in the red book]
Action sequencing occurs using yang [downward / outward] signals, which specify a desired state (in the co-ordinate frame of a given map). This map’s task is then to specify desired states in lower maps, or directly in muscles or sub-cortical structures, to act as “subroutines”. Completion of each step in the subroutine triggers the next step (perhaps via loosely coupled rules); eventually resulting in an actual state that matches the desired state. This then signals the end of that step to the higher map, and so on. Sensory attention and anticipation operate in an essentially similar way.
30th Aug. 2006 Expected / unexpected rather than familiar / novel
[not in red book]
Given several sensors they all detect a stimulus each cycle. If a detected stimulus matches the same stimulus detected by a sensor in the previous cycle then the sensor does nothing because the sensor is expecting no change in the stimulus – the sensor is expecting the same stimulus repeated and only “speaks up” if there is a change in stimulus. The greater the change / difference the greater the “speaking up”. The sensor that detects the greatest change speaks the loudest. This may overcome the concentration level of any conscious habits and thus attract attention.
The same is true of any active S or A-habit, it is in a state expecting its result stimulus and it is being done at its concentration level. If the habit gets the expected result the pair has been detected, expectations met and there is no unexpected stimulus sequence of interest that needs learning. If there are several simultaneously active S-habits (all started by the same trigger stimulus) with expected results as long as one gets its expected result there is no unexpected stimulus sequence of interest that needs learning. If the trigger stimulus started an A-habit then the action has been done and there can only be this one A-habit active. However with multiple output devices and multiple sensors multiple A-habits can be active in parallel each with its own expected result.
So when the trigger stimulus occurs it is matched against previous experiences. It matches the trigger stimulus of one or more past habits. Of all these habits there may be some A-habits whose (1) result is interesting or (2) sequence was unexpected. The most recent one of these will be done at a concentration level equal to (1) the interest level of the expected result or (2) the change experienced. If no A-habits are interesting then the next reflexive action will be done at no concentration level and expecting a result the same as the trigger stimulus i.e. no change.
So when the result stimulus occurs it is compared with all expected results. If it matches an active A-habit the habit is learnt and no longer of interest. If it matches no expected result then we store the current trigger, the action and the result with an interest level in being repeated equal to the change experienced between the result and expected result, i.e. (2) above. The trigger stimulus is also assigned this interest level so we get backward propagation.
31st Aug, 2006 From the Internet [not in the red book]
A reflex is a simple form of behaviour and clearly illustrates the logic of a sensory-motor system loop.
A reflex is an involuntary and relatively stereotyped response to a specific sensory stimulus. Two features of the sensory stimulus are particularly important in shaping the reflex response. First, the precise location of the stimulus determines in a fixed way the particular muscle that contracts to produce the reflex response. Second the strength of the stimulus determines the amplitude of the response. Reflexes therefore are graded behaviours.
Some authors exclude from the definition of learning very simple experience-induced behavioural modifications, such as sensitization or habituation, although most neuroscientists identify them as non-associative learning. Non-associative learning is a behavioural change brought by repeated presentation of one stimulus. In the case of sensitization, a stimulus that originally elicited a weak or no response, starts evoking stronger responses after several presentations, or when the presentation of one very intense stimulus evokes stronger responses at other stimuli. For example, after a very loud crash sound, smaller noises can startle a person, which otherwise would go almost unnoticed. In the case of habituation, repeated presentation of the same stimulus produces decreasing responses to it. In the example of the loud crash, if it keeps sounding repeatedly every twenty seconds, the startle to it will decrease in further presentations.
1st September 2006 Environment for testing graduated stimuli
A simple environment for testing graduated absolute and changes in stimuli would be a 4 square world in which Adaptron moves. Responses of turn left 90 degrees, turn right 90 degrees, move forward one square and graduated stimuli which are range finding distances to the wall in front of it which would be 0 distance or 1 distance. Changes in the stimulus value would be -1, 0, and +1.
[This is the end of the written comments in the red book]
20th Sept, 2006 From Wikipedia
Reinforcement causes a response to occur with greater frequency.
Punishment causes a response to occur with less frequency.
Extinction is the lack of any pleasant or unpleasant stimulus and the response occurs with less frequency.
Reinforcement, and punishment are either “positive” (add a stimulus following a response), or “negative” (remove a stimulus following a response). This creates a total of four basic consequences, because what is added or removed may be either reinforcement or punishment. A fifth procedure is known as extinction (i.e. nothing happens following a response).
The four possibilities are:
- Positive reinforcement occurs when a response is followed by a pleasant stimulus that increases the frequency of that behaviour. In the Skinner box experiment, a stimulus such as food or sugar solution can be delivered when the rat engages in a target behaviour, such as pressing a lever. An increase in the level of a pleasant stimulus.
- Negative reinforcement occurs when a response is followed by the removal of an unpleasant stimulus thereby increasing that behaviour's frequency. In the Skinner box experiment, negative reinforcement can be a loud noise continuously sounding inside the rat's cage until it engages in the target behaviour, such as pressing a lever, upon which the loud noise is removed. A decrease in the level of an unpleasant stimulus.
- Positive punishment occurs when a response is followed by an unpleasant stimulus, such as introducing a shock or loud noise, resulting in a decrease in that behaviour. An increase in the level of an unpleasant stimulus.
- Negative punishment occurs when a response is followed by the removal of a pleasant stimulus, such as taking away a child's toy following an undesired behaviour, resulting in a decrease in that behaviour. A decrease in the level of a pleasant stimulus.
A type of learning in which a certain behaviour, either through its performance or restraint, avoids a punishing stimulus is termed avoidance learning. For example, performing the behaviour of shielding one's eyes when in the sunlight will help avoid the punishment of having light in one's eyes (decreases the level of an unpleasant stimulus). However, refraining from going outside into the sunlight will also result in the avoidance of this punishment. (Must have experienced an increase in the level of an unpleasant stimulus and thus no longer performs the response)
Extinction occurs when a response that had previously been reinforced is no longer effective. In the Skinner box experiment, this is the rat pushing the lever and being rewarded with a food pellet several times, and then pushing the lever again and never receiving a food pellet again. Eventually the rat would cease pushing the lever.
21st Sept, 2006 Interest Level = Concentration
If we are processing S-Habits, the level of interest (i.e. concentration level) results from the amount of change in a stimulus when compared with the previous stimulus from the same sense/feature. In an R-Habit all the stimuli that have changed from the previous one describe the “object”. When performing an S or A-Habit which has an expected level of interest (amount of change in the resulting stimulus which determines the concentration level) if the amount of change is greater then it is rewarding. If it is less than expected it is undesirable?? If it matches it is learnt and thus not interesting. Set the concentration level so that something else that happens in parallel on another sense has to be of greater interest to interrupt the S or A-Habit being done.
8th Oct, 2006 Action Sequences
An A-Sequence is the RSR that is the natural subconscious “collapse” of an S Trigger, R-S-R, S Result into S-A-S when both S-R-S’s are learnt. Once we have one or more A-sequences that are triggered by an S then they should be chosen before any primitive Responses to try when there is a need to perform a reflexive response. This concept does not apply when there is only one sense/feature and its values are symbolic. This is because a particular stimulus once followed by an A-sequence will continue to be followed by this A-sequence if rewarded and if not the next primitive response will be reflexively chosen. Using learnt A-sequences as possible reflexive responses before primitive responses works when the trigger stimulus partially matches the A-sequence’s trigger stimulus. This is possible either due to it being a P-Habit or the stimuli from one sense/feature is graduated and the change in the trigger stimulus matches but not the value or vice-versa.
18th Nov 2006 Graduated stimuli
When attention is switched to a particular sense/feature that is graduated, only the absolute value is obtained immediately. This is why when you look at (change your attention to) a clock with a second hand it seems to stand still for a long time until the second hand moves. It is also possible that when attention is attracted to a graduated stimulus the amount of change that triggered the attention shift is not recognized, just used to attract attention. This seems to imply that only conscious attention paid to changes are pattern matched / recognized with remembered experienced ones.
Isn’t the S-Habit formed from two sequential graduated stimuli on the same sense/feature the same thing as the change between the two stimuli? No, the S-Habit consists of two absolute amounts and the difference. An equivalent change / difference should trigger /match the S-Habit as should an equivalent 1st stimulus absolute value. When the difference matches there is no expectation of the second stimulus, the S-Habit fires / is recognized???
When it lays down a sequential track of three graduated stimuli it must form two S-Habits from the absolute values and also in parallel lay down a sequential pair (S-Habit) of change amounts that can be recognized separately from the absolute values. Thus the tune “Happy Birthday” can be recognized at a very high pitch where the pitch of the notes can not be recognized absolutely but the sequence of changes in notes can be recognized. Timing is also important in reality but if no time intervals were recognizable, i.e. time was not an available stimulus this is how it could work. I don’t believe we need to hear the tune twice to lay down both tracks of S-Habits.
28th Nov 2006 Architectural definitions
Stimulus: The signal / impulse that results from the recognition of something. This could be a primitive stimulus from a sensor detecting something or a stimulus produced by a Recognition Habit (R-Habit) when it recognizes something.
Response: The signal / impulse that triggers the performance of an action. This could be a primitive response which is sent to an actuator / device to do something or a response that triggers an Action (either a Sequential or Parallel Action).
Recognition Habit (R-Habit): An R-Habit is implemented as a Binon which receives two input stimuli and produces one output / resulting stimulus. The resulting stimulus signals the recognition of something. There are two types of R-Habits, Sequential (S-Habits) and Parallel (P-Habits). Note that the input stimuli may be produced by either S or P Recognition Habits.
Binon: A Binon is the architectural component that implements a Recognition Habit. It requires two input stimuli and produces one output / resulting stimulus when the two inputs stimuli have occurred.
Sequential Habit (S-Habit): An S-Habit is a Recognition Habit that is triggered by an initial stimulus and then is expecting a second stimulus. If the second stimulus occurs immediately after the initial one then the S-Habit has recognized the sequence of stimuli and it produces its output / resulting stimulus. If the expected second stimulus does not occur next then the S-Habit fails and does not produce its output / resulting stimulus.
Parallel Habit (P-Habit): A P-Habit is a Recognition Habit that produces its output stimulus when both of its two input stimuli occur at the same time, i.e. in parallel, concurrently or simultaneously within the time resolution of the sensors. Only primitive stimuli and other Parallel Stimuli can be input stimuli.
Action Habit (A-Habit): An A-Habit is implemented by a Binon and functions similarly to an S-Habit in which the output / resulting Stimulus becomes a Response. It only works sequentially. After the initial trigger stimulus occurs the A-Habit produces the output / response and then is expecting its second stimulus (S-R-S). If the expected Stimulus occurs the A-Habit has been learnt else it has failed and some other A-Habit has been formed.
Sequential Action (Action Sequence): An S-Action is triggered by an initial sequential Response. It then produces its first Response, obtains its expected feedback Stimulus and then produces its second Response (R-S-R). It is performed subconsciously. If the expected feedback is not obtained it fails and does not produce the second Response. The unexpected Stimulus should attract attention depending on the current level of concentration.
Parallel Action: A P-Action is triggered by an initial parallel Response. It then produces its two Responses at the same time, i.e. in parallel, concurrently or simultaneously. These two responses can be primitive responses or ones that trigger other Parallel Actions.
3rd Dec 2006 Thinking
A coin fell out of my pocket while driving and went down between the seat and armrest. At that time Ann made a “mental note” to look for it once we stopped. When we stopped her “mental note” “kicked-in” and she looked for it.
How did the thinking correspond? She thought about the Coin C and its desirability. This caused her to think about the Action A to search for C. However the context of riding in the passenger seat of a moving car did not allow her to perform the action. She then thought about an appropriate event / context which should happen in the future that would allow for a search, i.e. the stopping S of the car and then re-thought about A and its desirability (note; not the desirability of C). This formed the “mental note”. When the actual S occurred it matched up with the thought of S and produced the associated thought about the desirable A. The thought about A would have been a thought about the stimulus feedback obtained during performing a search between seat and arm rest.
18th Dec 2006 Changes in a graduated stimulus
A visual object is recognized because all of its edges move together. All the changes to the edges, which are changes themselves, occur at the same time and of the same amount. Thus when two or more changes occur simultaneously, usually on the same sense but that is not necessary, we tend to group them and interpret them as having a common cause which is the same “object”. An object is symbolic and is identified as a thing not a change in a graduated stimulus. Take for example two lights that are one unit of distance apart, the first is on, the second is off. If the first turns off and the second turns on we get the impression that the light moved. In the direction of alignment between the two lights, when the first one was on it presented two edges, its left edge is the change from dark to light and its right edge is the change from light to dark. When the second light turns on and the first turns off the left edge moves one unit over at the same time that the right edge moves one unit over. The two edges move the same amount in the same direction and we interpret it as an object (the light) moving over one unit. I’m not sure if the same would be true for a single edge such as a transition from dark to light. So a summary would state that any two or more changes that are occurring simultaneously could be viewed as an object. A single light turned on is viewed as an object because it has two edges. When the light appears to move we have a motion event “object”?
I believe the same is true for two changes that occur sequentially and get repeated in that same sequence later on. We recognize the pair of changes as a single symbolic thing. Thus “Happy Birthday” sounds like the same tune even if played in many different keys. Note that since the change quantities are also graduated the change in the changes could also be perceived as a change but I can’t think of an example of such a stimulus chain being recognized as a result. An example would be the stimulus reading from a range finder that measures the distance to a second object. The second object moves toward us and the range decreases by a certain amount. In fact each range reading is made up of its absolute value and the change from its previous reading. This change is the velocity of the object, which should be recognizable independent of the absolute distance to the object. Two readings in a row provide a sequence of two absolute values which could be recognized again and the two changes provide a rate of change which indicate slowing, steady, or speeding up of the object i.e. changes in its velocity which I believe we use as a stimulus.
Range reading can be represented as numeric values as can the differences / changes. So the sequence 20, 19, 18 produces the changes -1, -1 a steady velocity which does not attract attention because there is no change in the changes.
20th Dec 2006 Graduated stimulus and changes
The following sequences of graduated stimuli produce the same acceleration values:
Stimuli: 20 20 19 18 16 16 6 6 5 4 2 2
Change ? 0 -1 -1 -2 0 ? 0 -1 -1 -2 0
Acceleration: ? ? -1 0 -1 2 ? ? -1 0 -1 2
Stimuli: 20 21 21 21 20 21 6 8 9 10 10 12
Change ? 1 0 0 -1 1 ? 2 1 1 0 2
Acceleration: ? -1 0 -1 2 ? -1 0 -1 2
Consider the sequence that has a single event in it.
Stimuli: 1 1 1 1 2 1 1 1 1
Change 0 0 0 1 -1 0 0 0
The change of 1 is the left edge of the event and the -1 is the right edge if these were visual and all received simultaneously instead of sequentially. Same in the following 2 examples:
Stimuli: 1 1 1 1 2 2 1 1 1 1
Change 0 0 0 1 0 -1 0 0 0
Stimuli: 1 1 1 1 2 2 2 1 1 1 1
Change 0 0 0 1 0 0 -1 0 0 0
The change attracts attention.
The magnitude of the change overcomes (or not) the concentration level.
The patterns in the changes are recognized (2 changes in a row) where the stimulus sequence (3 in a row) is familiar.
What if we just have a single change event in the sequence of stimuli?
Stimuli: 1 1 1 1 2 2 2 2
Change: 0 0 0 1 0 0 0
The changes 0, 1 will form a binon representing the change from steady to up one. This is a left edge if these were visual and all received simultaneously instead of sequentially.
21st Dec 2006 Changes in Stimuli
The change of 1 above is the same as currently occurs when a change is noticed in a repeating symbolic stimulus. Now consider a change of 2.
Stimuli: 1 1 1 1 3 3 3 3
Change: 0 0 0 2 0 0 0
This is more significant than a change of 1 and would attract if doing a habit at concentration level 1.