#robotics Logs
Oct 05 2018
#robotics Calendar
12:01 AM rue_mohr: 8
12:26 AM rue_mohr: ok done
12:26 AM rue_mohr: post the files for ya
12:29 AM rue_mohr: orlock, http://ruemohr.org/~ircjunk/not_public_dont_open/orlocks_parts/
12:29 AM rue_mohr: I think you will find it'll render in about 5 seconds
12:29 AM rue_mohr: there are only 2 csg operations
12:30 AM rue_mohr: you need both files
12:30 AM rue_mohr: I'm going to go all creative liscence on it now
12:30 AM rue_mohr: well not *right* now, but later
12:31 AM rue_mohr: oh and i'll build in support material
08:54 AM rue_mohr: orlock, ?
08:57 AM rue_mohr: orlock!?.... (:/
09:00 AM rue_mohr: hmm, I think I need to know how to use orgnized chaos to train a neural network
09:44 AM amenji: guys, what's the \tau symbol represent in the integral term of a simple PID shown in
09:44 AM amenji: https://www.dozuki.com/Guide/How+to+Link+to+a+Specific+Section+o...
09:45 AM amenji: oops
09:45 AM amenji: https://www.dozuki.com/Guide/How+to+Link+to+a+Specific+Section+o...
09:45 AM amenji: what's the \tau symbol represent in the integram term of a simple PID here: https://en.wikipedia.org/wiki/PID_controller#PID_controller_theory?
06:47 PM Jak_o_Shadows: amenji is gone
06:48 PM Jak_o_Shadows: but the answer is that it doesn't really mean anything in itself, it's an integratin thing
07:19 PM rue_mohr: orlock, ?
07:51 PM Tom_L: dam dude.. does he owe you money??
07:51 PM Tom_L: heh
07:52 PM Tom_L: he has been a little scarce...
08:48 PM zhanx: ok my extruder nozzle came lose some how
08:50 PM Tom_L: rue_mohr, http://tom-itx.no-ip.biz:81/~webpage/rue/encoder/Collet.jpg
08:52 PM zhanx: Tom_L its leaking https://imgur.com/a/3A6ocDp :D
08:52 PM Tom_L: just a little
08:53 PM zhanx: it should be square to the bed, but that tilt lets it leak a tad bit
09:00 PM rue_mohr: zh
09:00 PM rue_mohr: send me picstures
09:00 PM rue_mohr: the seal of the nozzel is made as a flat surface junction by betweent eh nozzel and the stainless peice
09:01 PM rue_mohr: dont crank on it
09:01 PM zhanx: ok i just pulled it off, reheating it now
09:01 PM rue_mohr: make sure that the nozzel is not jammed all the way up against the heater block
09:01 PM rue_mohr: make sure the ... I dont know if I can prevent you breaking it, even if I finish what I'm saying...
09:01 PM zhanx: you on email computer?
09:02 PM rue_mohr: yes
09:02 PM rue_mohr: zhanx, what seals the plastic in the nozzel asembly..
09:03 PM zhanx: nothing i can see, 2 emails on the way while i clean this up
09:03 PM rue_mohr: the first challange is getting them to hear the sound you made
09:03 PM rue_mohr: the second challange is getting them to hear the words you said
09:04 PM rue_mohr: the third challange is getting them to understand the thought you expressed
09:04 PM rue_mohr: communication is hopeless
09:05 PM rue_mohr: zhanx, my goal, is to make an nural network that can understand phonames, and repeat them for me
09:05 PM rue_mohr: so that I can just hand people a "what?" box and they can push the button to have things repeated to them as many times as they want
09:05 PM rue_mohr: at whatever speed they want
09:06 PM rue_mohr: this way, I can talk into it full speed, throw them the box, and leave
09:06 PM rue_mohr: I only know one person who could ever understand me when I'm talking full speed
09:06 PM rue_mohr: most of the time I need to talk so slow I get borred
09:06 PM rue_mohr: usually forgetting what I'm talking about
09:07 PM rue_mohr: and people still need me to repeat
09:07 PM rue_mohr: *sigh*
09:08 PM rue_mohr: zhanx, in the next 12 hours you have a 97% chance of breaking the stainless thermal isolation tube on your printer head
09:08 PM zhanx: nope
09:08 PM rue_mohr: :(
09:08 PM rue_mohr: the nozzel needs to not be tightened right up against the heating block
09:08 PM zhanx: I cleaned up the threads and did not over tighten it
09:09 PM zhanx: its tight but not like i want to break it tight
09:09 PM rue_mohr: if it goes it far enough for the shoulder to thouch the block, then the heater block is sitting too low on the stainless tube
09:10 PM rue_mohr: the seal is between the flat shoulders between the stainless pipe and the nozzel
09:10 PM rue_mohr: with the stainless tube not having any pressure on it, you hold the block with one wrench, and tighten the nozzel up against the stainless with another wrench, firm, making sure no stress is put on that stainless tube
09:11 PM rue_mohr: aaaanyhow
09:11 PM rue_mohr: sorry, suppose I'm depresssed
09:11 PM rue_mohr: tried to get a nap, 4 phonecalls
09:13 PM Tom_L: rue_mohr, http://tom-itx.no-ip.biz:81/~webpage/rue/encoder/Collet.jpg
09:13 PM zhanx: i had 42 phone calls today
09:13 PM Tom_L: http://tom-itx.no-ip.biz:81/~webpage/rue/encoder/Collet_Drawing.jpg
09:14 PM rue_mohr: Tom_L, tahts the spirit
09:14 PM zhanx: i need to setup an server so i don't have to email
09:15 PM _unreal_: wow https://www.youtube.com/watch?v=yWd4mzGqQYo
09:15 PM Tom_L: updated the drawing...
09:15 PM rue_mohr: I need rest
09:16 PM _unreal_: I need..... no more alcohal
09:16 PM rue_mohr: apparently my vow to speed in as long as I need rommorow has been uprooted by a trip to see an old familty friend on the first ferry tommorow mornig
09:17 PM _unreal_: ?
09:17 PM _unreal_: :) typical canadian... always taking the farry
09:17 PM _unreal_: ferry
09:21 PM zhanx: ok rue printing your 100mm test square again
09:22 PM rue_mohr: rememebr to nOT peel it off the bed
09:24 PM zhanx: yep, and i need to slightly adjust the bed
09:24 PM rue_mohr: :)
09:24 PM zhanx: but its looking good
09:24 PM rue_mohr: the L with the verticle right?
09:24 PM Tom_L: added a few more dimensions...
09:25 PM rue_mohr: got it, resaving
09:25 PM rue_mohr: 3/16 right?
09:25 PM Tom_L: the 149.9227 deg should just be 30 deg
09:25 PM Tom_L: .1875, yes
09:26 PM Tom_L: slots stop .2" from each end
09:27 PM Tom_L: now if i had a cnc lathe...
09:28 PM zhanx: 102.39mm on the x axis same on the y
09:28 PM zhanx: to outside edges
09:57 PM rue_mohr: your supposed to check to see if its square
09:57 PM rue_mohr: too
09:57 PM rue_mohr: do not measure the size of it, thats not how ie towrks
09:58 PM rue_mohr: left edge to left edge
09:58 PM rue_mohr: that way the width of the traces cancels out
09:58 PM rue_mohr: I dont know if I'm being heard
10:01 PM * rue_mohr closes orlocks part
10:07 PM rue_mohr: this shoudl not be so hard
10:07 PM rue_mohr: I just cant see it
10:08 PM rue_mohr: nural network -> O <- state machine
10:08 PM rue_mohr: cmon rue
10:13 PM zhanx: rue its square
10:13 PM zhanx: I had to adjust the bed a tad bit for levelness
10:13 PM zhanx: but its .05 out on the last test square all the way around now
10:14 PM zhanx: i got my glass out and its half round for the first layer like you wanted
10:15 PM zhanx: issue now is its getting cold in here and the pla is not sticking right
10:15 PM zhanx: a heated bed would help
10:26 PM rue_mohr: good show
10:26 PM rue_mohr: no wait
10:26 PM rue_mohr: its not sticking becasue of the wax on the tape
10:27 PM rue_mohr: zhanx,
10:27 PM rue_mohr: take some alcohol, put it on a paper towel, give it a REALLY FAST wipe
10:27 PM rue_mohr: the plastic will stick like its part of the atomic matter of the tape
10:27 PM zhanx: k
10:27 PM zhanx: crap its in the work truck
10:28 PM rue_mohr: the alcohol pulls the glue up and removes the release wax
10:28 PM rue_mohr: you can get a feel for how much you need to wipe a table to get what you want
10:28 PM rue_mohr: cause your a nural network
10:29 PM rue_mohr: https://www.youtube.com/watch?v=AuEz4Ul9tHM
10:29 PM zhanx: this is true
10:30 PM rue_mohr: I just did that, its not the one I wanted
10:30 PM rue_mohr: I wanted the one where it has it learn the values
10:30 PM rue_mohr: *sigh*
10:30 PM zhanx: jesus, i just tryed to scroll that spread sheet on youtube
10:30 PM Tom_L: hahahaha
10:30 PM * rue_mohr gets another tea
10:31 PM zhanx: i guess my brain is deep fried from today
10:31 PM zhanx: serial connection on a T-1 was being stubborn
10:32 PM zhanx: turned out the gateway layer provided blocked the router
10:33 PM rue_mohr: I often try to grab and rotate images of 3d models
10:33 PM Tom_L: i do that all the time
10:33 PM Tom_L: and they follow the mouse
10:33 PM rue_mohr: :)
10:34 PM zhanx: lol i have like 20 of them 100mm squares now
10:34 PM Tom_L: i got my ball mouse set to pan, zoom and rotate all with 2 clicks
10:34 PM rue_mohr: zhanx, there is a symantic error in how neural networks are usually done, I'm gonna try to untwist it
10:34 PM rue_mohr: atleast its not much palstic
10:35 PM zhanx: true
10:35 PM Tom_L: https://www.youtube.com/watch?v=kNPGXgzxoHw
10:35 PM rue_mohr: thats part of why I designed it
10:35 PM zhanx: not worried about the plastic, really, I want it right
10:36 PM rue_mohr: so, the synapse is the strength
10:36 PM rue_mohr: activation is its level
10:36 PM rue_mohr: before the threshold
10:36 PM zhanx: hmm
10:37 PM zhanx: I want to take the work but kids are out of school on monday
10:40 PM zhanx: ok rue now its stuck to the tape
10:41 PM rue_mohr: not fast enough, too much alcohol
10:41 PM rue_mohr: " REALLY FAST"
10:41 PM zhanx: learning curves are fun
10:41 PM rue_mohr: mhm
10:41 PM rue_mohr: its pretty damn sensitive
10:42 PM zhanx: while i watch toms video
10:42 PM Jak_o_Shadows: hmm. Now i need to 3d print a bit, make the screen surround smaller, and do some sort of switch so that I don't drain the car battery from the OBD port
10:42 PM zhanx: Jak_o_Shadows, project?
10:43 PM Jak_o_Shadows: reversing camera into the car
10:43 PM zhanx: i heard obd and was like what? where?
10:43 PM Jak_o_Shadows: I can't be bothered taking the dash apart, so I'll just use the OBD for power
10:43 PM zhanx: ok and you can read the gears for reverse with it also
10:43 PM Jak_o_Shadows: Lol, I could.
10:44 PM Jak_o_Shadows: but it's a wireless camera, so I just put the camera end into the reversing light
10:44 PM zhanx: that is the cheating way
10:44 PM Jak_o_Shadows: yup
10:44 PM Jak_o_Shadows: But once I have the OBD connection up, I can always start reading it
10:44 PM Jak_o_Shadows: ive been meaning to log stuff ro ages
10:46 PM rue_mohr: hmm, you know what a nural network kinda looks like?
10:46 PM Tom_L: dna?
10:47 PM rue_mohr: an fft
10:48 PM rue_mohr: hmm
10:49 PM rue_mohr: if the weights for the hidden layers are stored in an array, and you use the same array for storing the values of all the hidden layers
10:50 PM rue_mohr: nobody is going to read the log, right?
10:50 PM zhanx: nope
10:52 PM rue_mohr: ok, I need to learn more, but this might be a good one
10:52 PM rue_mohr: suppose you have never written fft code, and neural network code?
10:53 PM rue_mohr: the difference is that the fft has its weights come from a sine function, and that there is no threshold on each layer of the calculations
10:53 PM rue_mohr: (no nonlinear element)
10:53 PM rue_mohr: which should mean that a neural network, given an array of audio samples, should be able to pick out frequencies
10:54 PM rue_mohr: I'm working at only about 30% brain capacity and its SO frustrating
10:54 PM rue_mohr: this should all be easy
10:55 PM rue_mohr: ok lets make a trainable network
10:55 PM Tom_L: all i can do is click on a mouse button and paste links
10:55 PM rue_mohr: can I turn fft code into a neural network processor by just changing the weight system?
10:56 PM Tom_L: you're seriously not asking me....
10:56 PM rue_mohr: that would be ... hyper amazing
10:56 PM rue_mohr: cause it means that neural network tech goes back farther than thought
10:56 PM rue_mohr: its always been right there
10:56 PM rue_mohr: since 1957
10:58 PM rue_mohr: https://www.youtube.com/watch?v=3993kRqejHc
10:58 PM rue_mohr: ok, if I hold my breath and grit my teeth, I can do this one
10:59 PM rue_mohr: then do it in C
11:00 PM rue_mohr: I'v not sen any of the 20 channels get spammed yet
11:03 PM rue_mohr: so, all the hidden layers can use the same memory
11:03 PM Tom_L: https://www.youtube.com/watch?v=x69YhAapw4k
11:17 PM rue_mohr: >:| NEED TO FIX SWITCH ON MAIN MOUSE BUTTON
11:37 PM rue_mohr: ok I made it
11:37 PM rue_mohr: now to fix somet hings
11:38 PM rue_mohr: you know the neat things is that I USED to be good at flattening itterative solution methods
11:39 PM rue_mohr: the teacher in electroics school said the values for motor startup resistors could not be just calculated, you had to use itteritive solution finding and live with the best answer you got for how much effort you put in
11:40 PM rue_mohr: but I did it, I found a one-pass solution formula
11:40 PM rue_mohr: I think it took me 3 days
11:40 PM rue_mohr: I wonder if I can do that with neural network training
11:41 PM rue_mohr: I should paint the wall so I have something to target me talking-to to.
11:41 PM rue_mohr: but rue, you ahve me
11:41 PM rue_mohr: why the hell you been so quiet tho?
11:41 PM rue_mohr: well, I'm tired
11:41 PM Tom_L: just watching
11:41 PM rue_mohr: ok, and I'm alone here, changing the world
11:41 PM rue_mohr: ;)
11:41 PM zhanx: its like a show
11:42 PM Tom_L: can't contribute so i'll watch
11:42 PM rue_mohr: mhm
11:42 PM rue_mohr: so, lets makes some changes to this excel sheet
11:42 PM zhanx: should i email a photo of this print?
11:44 PM rue_mohr: yes
11:44 PM rue_mohr: send lots of pictures that are not crazy zoomed in
11:45 PM zhanx: its a cell phone
11:45 PM rue_mohr: the more pics you send me, the more problems I can help prevent
11:48 PM rue_mohr: huh, it cant learn xor
11:48 PM rue_mohr: not enough hidden layers
11:48 PM rue_mohr: ?
11:49 PM rue_mohr: interesting property of a neural network
11:49 PM rue_mohr: if there are not enough layers, it cant learn some things
11:53 PM rue_mohr: the trainign is done wrong
11:53 PM rue_mohr: the correction value should half each time
11:55 PM rue_mohr: damnit, my system is acting up