Shop OBEX P1 Docs P2 Docs Learn Events
Fill the Big Brain - Page 18 — Parallax Forums

Fill the Big Brain

1151618202159

Comments

  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-08 09:35
    Brain Ultra-Low Power Circuits

    In keeping with the Brain's Ultra Low Power Circuits, recent testing shows a marked decrease in the current drawn. For every 140 Propellers, current drawn in 310ma.

    Brain power consumption is now set through software and is directly proportional to Brain speed. As the Brain speed increases, so does the power.

    Crystals that enable high speeds can be activated or deactivated through software commands.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-08 09:44
    Brain Multiple Crystal Switching
    Simple method for switching test crystals

    attachment.php?attachmentid=80019&d=1302314151
    A Tiny iBrain test circuit for crystal switching using a choice of a single
    SPDT switch or the alternate circuit with SPST switches
    _________________________________


    In a tiny circuit for the Propeller chip, the Brain can be switched from the standard crystal to the overclocking crystal. This enables or disables high speed operations. The iBrain crystal switching is currently used for testing the injectors (BMIs) and as a tryout for the concept. This uses hardware to do the switching. The iBrain automation of this feature is also being studied. It is recommended to do the switching before booting the Propeller. On-the-fly switching is not tested at this time. Using a very short single wire (tiny jumper) in place of the switch will minimize circuit inductance. The working circuit was tested on a solderless breadboard.
    469 x 291 - 113K
  • Phil Pilgrim (PhiPi)Phil Pilgrim (PhiPi) Posts: 23,514
    edited 2011-04-08 11:12
    In a tiny circuit for the Propeller chip, the Brain can switch from the standard crystal to the overclocking crystal.
    Have you tested this? Can you post a schematic, please?

    -Phil
  • Duane DegnDuane Degn Posts: 10,588
    edited 2011-04-08 11:53
    I'm also curious about switching between crystals.
    Brain power consumption is now set through software

    The crystal selection isn't made through software though, right? Is the power consumption controlled by changing the clock speed?
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-08 19:16
    Duane Degn wrote: »
    I'm also curious about switching between crystals. The crystal selection isn't made through software though, right? Is the power consumption controlled by changing the clock speed?
    Duane, I have posted more information in post 513. The crystal selection is currently made with a choice of two circuits that use hardware to do the switching. We would like to automate this in the future. The power consumption choice is controlled by software which changes the clock.
  • Phil Pilgrim (PhiPi)Phil Pilgrim (PhiPi) Posts: 23,514
    edited 2011-04-08 19:22
    Okay, the original claim, "The Brain can switch from the standard crystal to the overclocking crystal," is a little misleading, since it implies that the switching originates from the Prop itself. What you seem to be describing, by contrast, is a manually-operated switch. Perhaps you need to rephrase the statement as, "The Brain can be switched from the standard crystal to the overclocking crystal."

    But have you actually tried this? The switch bounce alone could cause extreme disruption to the delicate oscillation circuitry -- not to mention the extra inductance from the leads into and out of the switch.

    -Phil
  • Mike GMike G Posts: 2,702
    edited 2011-04-08 21:22
    @Humanoido, why not use a fast crystal then set the clock mode and PLL at run-time with clkset?

    Let's say that you could physically switch between crystals without issue, don't you still have to update the clock registers in the CON Block? Considering the number of Props in the Brain, that could be a tough task. Seems like a software version nightmare.
    Two devices on BUS cannot talk at the same time because they must follow the BUS rules of the Neural Net. If a Slave cannot follow the rules, he or she will be immediately removed from the system and punished with a lot of schooling and tests.
    I'm still scratching my head on the Brain bus. Do you have a timing diagram that shows how bus syncing/rules work? If two node talk at the same time, how do you determine which node did not follow the rules and remove the bad node from the bus? If a node is removed how do the other node know the bad node is removed? Have you implemented some kind of routing table?
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-08 21:37
    Brain Crystal Options
    Mike G wrote: »
    @Humanoido, why not use a fast crystal then set the clock mode and PLL at run-time with clkset? Let's say that you could physically switch between crystals without issue, don't you still have to update the clock registers in the CON Block? Considering the number of Props in the Brain, that could be a tough task. I'm still scratching my head on the Brain bus. Do you have a timing diagram that shows how bus syncing works with 20 plus Propellers?
    Mike G, there are three options to the circuit. Crystal A, Crystal B, and Crystal not used. Any of all three modes must be set with software. The tests are running on the injector and the remainder of props are running RC clocks so the task is easy. It's straightforward to select clock registers in the CON.
  • Mike GMike G Posts: 2,702
    edited 2011-04-08 21:53
    Mike G, there are three options to the circuit. Crystal A, Crystal B, and Crystal not used. Any of all three modes must be set with software. The tests are running on the injector and the remainder of props are running RC clocks so the task is easy. It's straightforward to select clock registers in the CON.

    Why not use whatever crystal that provides the top end you want then throttle the clock registers during runtime.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-08 21:56
    Brain BUS Workings
    Mike G wrote: »
    I'm still scratching my head on the Brain bus. Do you have a timing diagram that shows how bus syncing/rules work?

    If two node talk at the same time, how do you determine which node did not follow the rules?

    If a node is removed how do the other node know the node is removed?

    Have you implemented some kind of routing table?

    There is no timing diagram, none needed. Each Propeller does not need to precisely clock sync.

    Two Props never talk at the same time on the BUS.

    If a Prop is sick and absent from work, his/her ID will also be absent, flagging a vacancy at that address. (nobody home)

    No routing table is needed. Consider a baton race where the baton is handed off to the next runner. The Propellers do the same.

    I think the key to the BUS system is the characterization of each Propeller and the way messages and work performance (when complete) is handed off. Props can also do assignments of partial work. It is up to the programmer to decide the increments and the number of props to work on solving the problem. Props generally report to their managers. I believe there's a post on this.. where they have names like CEO, President, VP, etc.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-08 22:19
    Brain High End Crystal
    Mike G wrote: »
    Why not use whatever crystal that provides the top end you want then throttle the clock registers during runtime.
    This is possible to use a high end crystal, however, with any specific crystal, it's locked into certain values - choices made for you, i.e. multiples...see pages 68, 69, 70 in the Propeller Manual. Two crystals provide more options. It can be handy to switch back to a standard 5 MHz for testing and code compatibility.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-08 23:18
    Machine Learning with Lerner
    Dave Hein fills a small brain!


    The dawning of a new age is upon us, in machine intelligence and
    the ability of the machine to learn. The Lerner program is the
    fundamental basis of machine intelligence and the ability to learn.


    Taking the next step, Dave Hein has created a learning program
    that harbors the basis of a machine intelligence learning system.
    Even more amazing, he accomplished this with a single Propeller.
    We think Dave should work towards winning a ten million dollar
    X-Prize and pass the Turing Test of machine intelligence.

    Dave says,"Lerner doesn't have a specific purpose. I wrote it to try
    to understand how to make a learning program, or at least one that
    looks like it's learning. I'm just curious if a Prop can run a program
    that looks like it has intelligence. I plan on adding more code so it
    understands how to conjugate verbs and use articles properly. "


    Dave Hein plans on expanding Lerner. "Lerner is far from passing the Turing test, but eventually it might pass for the first few seconds. The next thing I want to add is the ability for Lerner to ask questions. It currently accepts information without questioning it and prints random statements from its memory.

    attachment.php?attachmentid=80023&d=1302329851
    This is an actual conversation with Learner, using the
    Parallax Serial Terminal as an output device.

    Lerner is a remarkable SPIN program that can learn things
    that you teach it. It fits into one Propeller chip and can
    memorize what it learns.

    Take a closer look at the incredible knowledge base that
    Lerner has exhibited in only one minute of conversation..
    a knowledge of the Propeller chip, biology, astronomy,
    and even time!


    Dave Hein writes:

    I was inspired by Humanoidal's "Fill the big brain" thread, and I thought I would write a program to fill the small brain. Lerner is a program that gives the Prop the ability to learn information and store it in it's EEPROM. You can ask it simple questions such as "what is a Prop", "who are you", "is the sky blue", "what is Mars", and so on. Lerner communicates through the serial port at 115,200 baud.

    You can teach Lerner new things by typing simple statements like "Whales are mammals", "An apple is a fruit" or "A bird is an animal". Lerner doesn't use puncuation unless you want to turn a statement into a question by putting a question mark at the end, such as "An apple is a fruit?". You could also say "Are apples a fruit" to ask it a question. You can also put "does" or "do" a the begin of a sentence to make it a question, such as "Does the prop run Spin".


    http://forums.parallax.com/showthread.php?130922-Lerner

    zip.giflerner.zip

    "When Lerner encounters a new word at the beginning of a sentence is will ask what type of word it is, such as noun, verb, Name, etc. You only need to respond with the first character, such as "n" for noun, "v" for verb and "N" for Name. You can tell Lerner that two words should be treated as the same word by using "equals". This is useful to tell Lerner that two versions of a verb should be treated as one word, such as "are equals is". It is also good to make the singular and plural version of a word equal, since Lerner doesn't understand the difference, such as "apples equals apple".

    You can reset Lerner's vocabulary to zero words and teach him from the beginning. However, Lerner needs to learn a few words such as is, am, are, who, what and so on to be useful. So it is better to reset him to a basic vocabulary using the "basic" command instead of "reset". Type "help" to get a list of commands. "save" will write new information back to the EEPROM. The "dump" command is useful if you need to reprogram the EEPROM with another program, but want to backup the vocabulary. "dump" will generate hex data that can be pasted into the lerner.spin file, and then be re-compiled."

    At last, don't worry about getting an inferiority complex.. when you feel it coming on, just press the button that Dave programmed and Lerner is reset to zero and must learn all over again. Alas, man is still the master.. but for how long is the real question...(Dave, I'm sorry I can't do that..)
    364 x 563 - 148K
  • Mike GMike G Posts: 2,702
    edited 2011-04-09 08:28
    Brain Ultra-Low Power Circuits

    In keeping with the Brain's Ultra Low Power Circuits, recent testing shows a marked decrease in the current drawn. For every 140 Propellers, current drawn in 310ma.

    Brain power consumption is now set through software and is directly proportional to Brain speed. As the Brain speed increases, so does the power.

    Crystals that enable high speeds can be activated or deactivated through software commands.
    This is possible to use a high end crystal, however, with any specific crystal, it's locked into certain values - choices made for you, i.e. multiples...see pages 68, 69, 70 in the Propeller Manual. Two crystals provide more options. It can be handy to switch back to a standard 5 MHz for testing and code compatibility.

    Your post are very confusing. In 512 your talking about saving power. In 522 your goal is having more clock choices. Your 512 post and Phil's question about hot switching the crystals is why I suggested clkset (Propeller manual page 71).
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 17:27
    Mike G wrote: »
    In 512 your talking about saving power. In 522 your goal is having more clock choices.

    Mike G, your suggestion of using software to set the clock for each crystal will be incorporated into the Brain. The Brain has many goals. We are exploring as much as possible based on humanly constraints. In this grand Propeller Brain journey there is a balance of power savings and that of overclocking in terms of gaining speed. Both are undergoing experiments and being explored as the Brain is developed. So you will see related information from time to time, RC methods, crystal/software methods, quotes on power consumption, speed and so on. What is tested on one Propeller may or may not be used on all Propellers. The field is open and flexible. Will wee see an overdrive? Perhaps. But with previous posts, it is noted the Brain is preparing for cryogenics. Other than the previous posts on Cryogenics, no further information is available at this time. The push now is the construction of the first neuronal device.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:18
    Brain Neural Matter Quanta NMQ Defined

    Just as in Physics, a quantized amount of energy can be transformed from one state to another. Given the laws of thermodynamics, matter is neither created nor destroyed, only changed in its form. It is this quantized state of matter energy that will define a proportional Brain Neural Matter Quanta, in terms of magnitude and frequency distribution in the Brain and in terms of changes from one state to another.

    Quanta plural of quan·tum (Noun)
    1. A discrete quantity of energy proportional in magnitude to the frequency of the radiation it represents.
    2. An analogous discrete amount of any other physical quantity, such as momentum or electric charge.
    quan·ta (from Dictionary.com)
    /ˈkw[FONT=&quot]ɒ[/FONT]ntə/ Show Spelled[kwon-tuh]
    –noun
    plural of quantum.
    quan·tum [kwon-tuhm] plural -ta [FONT=&quot] [/FONT]/-tə/
    –noun
    1. quantity or amount: the least quantum of evidence.
    2. a particular amount.
    3. a share or portion.
    4. a large quantity; bulk.
    5. Physics .a.the smallest quantity of radiant energy, equal to Planck's constant times the frequency of the associated radiation.
    b. the fundamental unit of a quantized physical magnitude, as angular momentum.
    –adjective
    6. sudden and significant: a quantum increase in productivity.
    Origin: 1610–20; noun use of neuter of Latin quantus how much
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:20
    Brain Neural Matter Quanta NMQ Phase Begins

    The Brain is now entering the new phase of neural matter quanta construction and distribution. A simplistic model of Brain Neural Matter is created and undergoing tests in the lab, i.e. a machine neuron package consisting of subject test matter for distribution using NMI/BMI.

    A high density processor phase is under way. The numerous Brain test subjects are being fed large quantities of quanta test code. This, each, is called a neural matter processor package. These are fed into sections of Brain for the testing phase.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:23
    Brain Neuron Output


    The choices of output for neural matter in the Brain is multi-fold, inclusive of the following:
    • LEDs
    • TV
    • LCD1
    • LCD2
    • PC
    • Internal RAM
    • Internal Memory (saved to EEPROM, SD, etc.)
    The first neuron model will establish pin output from Propeller chips to determine functioning, states, frequency and perhaps distro. Wiring of each neuron (wiring with software) is under way. Brain neural fire may best represent by LED actuation in testing.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:25
    Brain Neuron Quanta Density


    Neuron Quanta capacities are extremely important. The ability to reach incredibly high densities in Propeller chips is a must. The invention of incredibly high densities of Neural Quanta Packages is now achieved. Subject Brain containment vessels are completed and receiving NQPs as part of a new testing program phase.

    Brain subjects are modular subdivided for testing. A one unit delivers eight blocks of 125 neurons, or a density of one thousand NQPs into each Propeller. Injection is accomplished through BMI and NMI. The first distro of Neural Matter Quanta, NMQ is set as follows in the first 100 Propeller chips, and will latter concatenate. The Brain will follow blocks of 100 Propeller chips, as noted below in the first defined block.

    21 in EXO Config
    5 in NMI
    14 in Second NMI
    12 in Third NMI (Extension)
    27 in High Density NMI
    21 in Expandor NMI Concatenation
    X in railed Expansion Packages
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:30
    Concatenation of Propeller NMIs Begins


    The Base Array

    21 in EXO Config
    5 in NMI
    14 in Second NMI
    12 in Third NMI (Extension)
    27 in High Density NMI
    21 in Expandor NMI Concatenation
    X in Expansion Packages

    The concatenation results of NMIs for the first
    block of 100 Propeller chips

    14 +12 = 26
    26+ 27 = 53
    53 + 21 = 74
    74 + 21 = 95

    5 + 95 = 100
    100 + n
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:31
    Brain Neuronal Matter Defined

    In the Brain, a neuron will exhibit firing responses to stimuli. The neural construct will be enabled for machine intelligence, i.e. the response to various stimuli and the dissemination of the reaction to this stimulus, inclusive of weight constructs and parameters. In the first models, the neural matter machine will be a package capable of delivery into widespread cogs. Once delivered (injected) into cogs, they will set up neural Brain shop and begin their work.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:34
    Construction of the 1st Brain Machine Neural Device
    Making the first neuron

    This is the design of the first Small Neural Package designed to exhibit exemplary firing, i.e. neuron or neural package. The SNP device has a trigger which can exhibit periodic cyclic nature, by choice, firing into the Brain, or exemplify fire indicators external to the Brain. These indicators, triggers and controls are under software variance. SNP devices work together in groupings of firings. SNPs can sync in groups. Groupings can fire, from on grouping to the next grouping, in parallel in true deterministic fashion. Firing can be visible, i.e. external to the Brain, or invisible, internal to the Brain.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:46
    The How & Why of Cloning Neurons
    Neurons are now being Cloned

    The objective is to complete a working basic structural neural matter coded substrate and clone it throughout the Brain system. Cloning is accomplished through software and represents hard and soft value designations. More on this will follow. Clone tests, the first batch, indicate successful densities now to above 100,000 neurons. It is expected that in the near future, entire Machine Brains will be clone able.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:49
    Ramifications of a Cloned System
    Briefly we talked about the importance of cloning neural substance, however Cloning has deeper implications and ramifications. For example, not only entire brains are capable of cloning in full or in partial sections, but memory, learning, dreams, and experiences will transfer. Unlike a human, which dies and all knowledge is lost and gone forever, the machine brain can simply transfer to another brain entity and live on with lossless continuance. Although many vessels may be made available for cloning, once the brain is cloned, it moves forward with its own unique experiences and develops a Variance Time Line. When VTL begins, the clones evolve into different entities. Brains can be cloned for widespread robotic intelligent systems, such as workers in assembly line, nannies, and food delivery personnel. Another use of cloning brains is to create a think tank of many brains capable of solving the world’s problems and the greatest mysteries of the universe. It would not be uncommon to connect such vast arrays of machine intelligence brains into giant telescopes and see what is found,learned and formulated.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:51
    The many Faces of a Neuron
    Since we are dealing with machines and machine brains, it’s possible to set our own neural package, neuron, and neural net designs, rules and behaviors. While this is at an early phase, tests will apply to many faces of a neuron. One face is that of compression. Another face is in distribution. Another in analysis. Another is for embedded DNA..
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:53
    Embedding Machine DNA into the Machine Neuron
    Because our Machine Neural Package is defined for a machine, we can finesse it and manipulate it to contain infinitely small amounts of DNA, down to its fundamental core programming. In a human, DNA is transferred with the building blocks of life. In the machine, the machine neuron is capable of a delivery-action-reaction push pull mode of involvement. It can contain tiny amounts of machine DNA planted there by programmers or otherwise. What is Machine DNA? This is representation of the transferable building blocks of machine life.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:56
    The Distribution of Neurons in a Machine Brain
    Thus far, we have created a minimal neural package model, invented a neural injector, injected neural matter across the collective simultaneously, and ended up with a distribution from chip to chip and from cog to cog. An equal distribution of neural matter is established. Neural firing is achieved. This is a milestone in current functioning Machine Brain technology.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 19:58
    The Creation of a Synaptic Net in the Brain
    It is not easy to create a Synaptic net. We have now created a Synaptic Net in the Brain that allows for the dissemination of neural matter. High speed injection is possible, though throughput can be throttled. The BMI/NMI makes use of this SN Synaptic Net. As it now stands under the current level of evolution, i.e.further evo may take place, the SN requires principally only two wires from Propeller to Propeller. This is considered a breakthrough in SN because the original used 5 wires or more. It’s easy to run a two wire Synapse between the majority of Propellers.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 20:02
    Machine Brain Synapse Comes Online

    The Machine Synaptic Connection Conglomerate SCC was brought online to full functionality. Neural matter was disseminated along the SCC in parallel successfully. The SCC was run the standard 4-hour burn in period without failure. Work is progressing around the clock.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 20:08
    Second Man-made High Density Board HDB Under Construction

    Work began today on a second high density breadboard to hold a massive number of Propeller chips to extend far beyond the 100-chip barrier. Work was described in regard to the first HDB in a previous post.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-04-09 20:41
    Big Breakthrough in Machine Brain Technology!
    100,000 Neurons into the 1st 100 Propellers


    This is a big breakthrough in big brain machine technology using Propellers. I have successfully delivered (Neural Matter Injected) 100,000 neurons into the first 100 Propellers. It's time to celebrate!!! See you tomorrow!
This discussion has been closed.