April 2017 - Pharo-users - lists.pharo.org

little "challenge"

by Stephane Ducasse

#(16 17 17 16 18 17 18 19 19 19 18 19 19 20 19 20 19 20 20 20 19 20) -> { #(17 17 16 18 17 18 19) . #(19 19 18 19 19 20 19) . #(20 19 20 20 20 19 20)} any code?

April 25, 2017

Re: [Pharo-users] improving audio/video quality of pharo days recordings

by Stephane Ducasse

Thanks for the help. We will see what we can do but these videos are always a problem to the point I was not sure that we will record them. Last year thibaut did a great job and I will ask him. On Tue, Apr 25, 2017 at 9:04 PM, Peter Uhnak <i.uhnak(a)gmail.com> wrote: > Hi, > > we had a small discussion on discord about improving the audio/video > quality of recordings of past pharo days, where many recordings are really > hard to see (or completely unreadable), or the presenter is really hard to > understand (especially if they speak quietly or have a strong accent). > > Obviously making really good recordings requires several people to focus > on it (which is hard on a small meetup), however two things could be imho > introduced that could help a lot, with little demand on attendees. > > 1) clip-on microphone > * camera's mic will not capture good sound from 10 meters away > over a crowd of people making noises > 2) VGA recorder (which goes between the presenter's computer and projector) > * then we would have crisp video and not relying on the quality of > the camera > > in fact this is very common format where in the final video you see > primarily the VGA output, and not camera capture from a distance (only on > the presenter), e.g. https://www.youtube.com/watch?v=XA2WjJbmmoM > > Unfortunately I am not attending Pharo Days this year (scheduling > conflict), so I can't help on-site, but I could help with post-processing > where possible, e.g. converting the raw audio/video/VGA streams into one > (as in the linked video) or something. > > Peter > >

April 25, 2017

improving audio/video quality of pharo days recordings

by Peter Uhnak

Hi, we had a small discussion on discord about improving the audio/video quality of recordings of past pharo days, where many recordings are really hard to see (or completely unreadable), or the presenter is really hard to understand (especially if they speak quietly or have a strong accent). Obviously making really good recordings requires several people to focus on it (which is hard on a small meetup), however two things could be imho introduced that could help a lot, with little demand on attendees. 1) clip-on microphone * camera's mic will not capture good sound from 10 meters away over a crowd of people making noises 2) VGA recorder (which goes between the presenter's computer and projector) * then we would have crisp video and not relying on the quality of the camera in fact this is very common format where in the final video you see primarily the VGA output, and not camera capture from a distance (only on the presenter), e.g. https://www.youtube.com/watch?v=XA2WjJbmmoM Unfortunately I am not attending Pharo Days this year (scheduling conflict), so I can't help on-site, but I could help with post-processing where possible, e.g. converting the raw audio/video/VGA streams into one (as in the linked video) or something. Peter

April 25, 2017

Re: [Pharo-users] Hot to retrieve values from Nested Dictionaries

by Paul DeBruicker

JsonObject has the doesNotUnderstand: trick too. Esteban A. Maringolo wrote > 2017-04-25 8:49 GMT-03:00 Markus BÃ¶hm < > markus.boehm@ > >: >> Tx, I tried Your proposal. Just as an example, it works: > >> P.S.: Editor in Playground shows NeoJSONObject keys in red color? > > It's highlighted in red because for the playground the keys are > selectors not implemented by any class in the image. > > The key in the Playground is not such, only a trick of NeoJSONObject > using Smalltalk's #doesNotUnderstand: mechanism to map selectors to > keys. > > Regards! > > > Esteban A. Maringolo -- View this message in context: http://forum.world.st/Hot-to-retrieve-values-from-Nested-Dictionaries-tp494… Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com.

April 25, 2017

Re: [Pharo-users] Hot to retrieve values from Nested Dictionaries

by Esteban A. Maringolo

2017-04-25 8:49 GMT-03:00 Markus BÃ¶hm <markus.boehm(a)gmail.com>: > Tx, I tried Your proposal. Just as an example, it works: > P.S.: Editor in Playground shows NeoJSONObject keys in red color? It's highlighted in red because for the playground the keys are selectors not implemented by any class in the image. The key in the Playground is not such, only a trick of NeoJSONObject using Smalltalk's #doesNotUnderstand: mechanism to map selectors to keys. Regards! Esteban A. Maringolo

April 25, 2017

Re: [Pharo-users] Neural Networks in Pharo

by francesco agati

thanks ;-) 2017-04-25 15:09 GMT+02:00 Oleks <olk.zaytsev(a)gmail.com>: > Hello, > > There isn't one yet. But I will try to create it today. I will let you know > > Cheers, > Oleks > > On Apr 25, 2017 16:10, "francescoagati [via Smalltalk]" <[hidden email] > <http:///user/SendEmail.jtp?type=node&node=4944027&i=0>> wrote: > >> Hi Oleks, >> there is a mode for install neural network from metacello? >> >> 2017-04-25 13:00 GMT+02:00 Alexandre Bergel <[hidden email] >> <http:///user/SendEmail.jtp?type=node&node=4944025&i=0>>: >> >>> Continue to push that topic Oleks. You are on the right track! >>> >>> Alexandre >>> >>> > On Apr 24, 2017, at 1:43 AM, Oleks <[hidden email] >>> <http:///user/SendEmail.jtp?type=node&node=4944025&i=1>> wrote: >>> > >>> > Hello, >>> > >>> > Thanks a lot for your advice! It was very helpful and educating (for >>> > example, I thought that we store biases in the weight matrix and >>> prepend 1 >>> > to input to make it faster, but now I see why it's actually slower that >>> > way). >>> > >>> > I've implemented a multi-layer neural network as a linked list of >>> layers >>> > that propagate the input and error from one to another, similar to the >>> Chain >>> > of Responsibility pattern. Also, now I represent biases as separate >>> vectors. >>> > The LearningAlgorithm is a separate class with Backpropagation as its >>> > subclass (though at this point the network can only learn through >>> > backpropagation, but I'm planning to change that). I'm trying to >>> figure out >>> > how the activation and cost functions should be connected. For example, >>> > cross-entropy works best with logistic sigmoid activation etc. I would >>> like >>> > to give the user a freedom to use whatever he wants (plug in whatever >>> you >>> > like and see what happens), but it can be very inefficient (because >>> some >>> > time-consuming parts of activation and cost derivatives cancel out each >>> > other). >>> > >>> > Also, there is an interface for setting the learning rate for the whole >>> > network, which can be used to choose the learning rate prior to >>> learning, as >>> > well as to change the learning rate after each iteration. I am >>> planning to >>> > implement some optimization algorithms that would automize the process >>> of >>> > choosing a learning rate (adagrad for example), but this would require >>> a bit >>> > different design (maybe I will implement the Optimizer, as you >>> suggested). >>> > >>> > I'm attaching two images with UML diagrams, describing my current >>> > implementation. Could you please tell me what you think about this >>> design? >>> > The first image is a class diagram that shows the whole architecture, >>> and >>> > the second one is a sequence diagram of backpropagation. >>> > >>> > mlnn.png <http://forum.world.st/file/n4943698/mlnn.png> >>> > backprop.png <http://forum.world.st/file/n4943698/backprop.png> >>> > >>> > Sincerely yours, >>> > Oleksandr >>> > >>> > >>> > >>> > -- >>> > View this message in context: http://forum.world.st/Neural-N >>> etworks-in-Pharo-tp4941271p4943698.html >>> > Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com. >>> > >>> >>> -- >>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>> Alexandre Bergel http://www.bergel.eu >>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>> >>> >>> >>> >>> >> >> >> ------------------------------ >> If you reply to this email, your message will be added to the discussion >> below: >> http://forum.world.st/Neural-Networks-in-Pharo-tp4941271p4944025.html >> To unsubscribe from Neural Networks in Pharo, click here. >> NAML >> <http://forum.world.st/template/NamlServlet.jtp?macro=macro_viewer&id=instan…> >> > > ------------------------------ > View this message in context: Re: Neural Networks in Pharo > <http://forum.world.st/Neural-Networks-in-Pharo-tp4941271p4944027.html> > > Sent from the Pharo Smalltalk Users mailing list archive > <http://forum.world.st/Pharo-Smalltalk-Users-f1310670.html> at Nabble.com. >

April 25, 2017

Re: [Pharo-users] Neural Networks in Pharo

by francesco agati

Hi Oleks, there is a mode for install neural network from metacello? 2017-04-25 13:00 GMT+02:00 Alexandre Bergel <alexandre.bergel(a)me.com>: > Continue to push that topic Oleks. You are on the right track! > > Alexandre > > > On Apr 24, 2017, at 1:43 AM, Oleks <olk.zaytsev(a)gmail.com> wrote: > > > > Hello, > > > > Thanks a lot for your advice! It was very helpful and educating (for > > example, I thought that we store biases in the weight matrix and prepend > 1 > > to input to make it faster, but now I see why it's actually slower that > > way). > > > > I've implemented a multi-layer neural network as a linked list of layers > > that propagate the input and error from one to another, similar to the > Chain > > of Responsibility pattern. Also, now I represent biases as separate > vectors. > > The LearningAlgorithm is a separate class with Backpropagation as its > > subclass (though at this point the network can only learn through > > backpropagation, but I'm planning to change that). I'm trying to figure > out > > how the activation and cost functions should be connected. For example, > > cross-entropy works best with logistic sigmoid activation etc. I would > like > > to give the user a freedom to use whatever he wants (plug in whatever you > > like and see what happens), but it can be very inefficient (because some > > time-consuming parts of activation and cost derivatives cancel out each > > other). > > > > Also, there is an interface for setting the learning rate for the whole > > network, which can be used to choose the learning rate prior to > learning, as > > well as to change the learning rate after each iteration. I am planning > to > > implement some optimization algorithms that would automize the process of > > choosing a learning rate (adagrad for example), but this would require a > bit > > different design (maybe I will implement the Optimizer, as you > suggested). > > > > I'm attaching two images with UML diagrams, describing my current > > implementation. Could you please tell me what you think about this > design? > > The first image is a class diagram that shows the whole architecture, and > > the second one is a sequence diagram of backpropagation. > > > > mlnn.png <http://forum.world.st/file/n4943698/mlnn.png> > > backprop.png <http://forum.world.st/file/n4943698/backprop.png> > > > > Sincerely yours, > > Oleksandr > > > > > > > > -- > > View this message in context: http://forum.world.st/Neural- > Networks-in-Pharo-tp4941271p4943698.html > > Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com. > > > > -- > _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: > Alexandre Bergel http://www.bergel.eu > ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. > > > > >

April 25, 2017

Re: [Pharo-users] Neural Networks in Pharo

by Oleks

Hello, There isn't one yet. But I will try to create it today. I will let you know Cheers, Oleks On Apr 25, 2017 16:10, "francescoagati [via Smalltalk]" < ml+s1294792n4944025h0(a)n4.nabble.com> wrote: > Hi Oleks, > there is a mode for install neural network from metacello? > > 2017-04-25 13:00 GMT+02:00 Alexandre Bergel <[hidden email] > <http:///user/SendEmail.jtp?type=node&node=4944025&i=0>>: > >> Continue to push that topic Oleks. You are on the right track! >> >> Alexandre >> >> > On Apr 24, 2017, at 1:43 AM, Oleks <[hidden email] >> <http:///user/SendEmail.jtp?type=node&node=4944025&i=1>> wrote: >> > >> > Hello, >> > >> > Thanks a lot for your advice! It was very helpful and educating (for >> > example, I thought that we store biases in the weight matrix and >> prepend 1 >> > to input to make it faster, but now I see why it's actually slower that >> > way). >> > >> > I've implemented a multi-layer neural network as a linked list of layers >> > that propagate the input and error from one to another, similar to the >> Chain >> > of Responsibility pattern. Also, now I represent biases as separate >> vectors. >> > The LearningAlgorithm is a separate class with Backpropagation as its >> > subclass (though at this point the network can only learn through >> > backpropagation, but I'm planning to change that). I'm trying to figure >> out >> > how the activation and cost functions should be connected. For example, >> > cross-entropy works best with logistic sigmoid activation etc. I would >> like >> > to give the user a freedom to use whatever he wants (plug in whatever >> you >> > like and see what happens), but it can be very inefficient (because some >> > time-consuming parts of activation and cost derivatives cancel out each >> > other). >> > >> > Also, there is an interface for setting the learning rate for the whole >> > network, which can be used to choose the learning rate prior to >> learning, as >> > well as to change the learning rate after each iteration. I am planning >> to >> > implement some optimization algorithms that would automize the process >> of >> > choosing a learning rate (adagrad for example), but this would require >> a bit >> > different design (maybe I will implement the Optimizer, as you >> suggested). >> > >> > I'm attaching two images with UML diagrams, describing my current >> > implementation. Could you please tell me what you think about this >> design? >> > The first image is a class diagram that shows the whole architecture, >> and >> > the second one is a sequence diagram of backpropagation. >> > >> > mlnn.png <http://forum.world.st/file/n4943698/mlnn.png> >> > backprop.png <http://forum.world.st/file/n4943698/backprop.png> >> > >> > Sincerely yours, >> > Oleksandr >> > >> > >> > >> > -- >> > View this message in context: http://forum.world.st/Neural-N >> etworks-in-Pharo-tp4941271p4943698.html >> > Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com. >> > >> >> -- >> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >> Alexandre Bergel http://www.bergel.eu >> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >> >> >> >> >> > > > ------------------------------ > If you reply to this email, your message will be added to the discussion > below: > http://forum.world.st/Neural-Networks-in-Pharo-tp4941271p4944025.html > To unsubscribe from Neural Networks in Pharo, click here > <http://forum.world.st/template/NamlServlet.jtp?macro=unsubscribe_by_code&no…> > . > NAML > <http://forum.world.st/template/NamlServlet.jtp?macro=macro_viewer&id=instan…> > -- View this message in context: http://forum.world.st/Neural-Networks-in-Pharo-tp4941271p4944027.html Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com.

April 25, 2017

ss3 and gemsource planned downtime today at 1:30 pm PDT (-0700 UTC)

by Dale Henrichs

The server hosting http://ss3.gemtalksystems.com/ and http://seaside.gemtalksystems.com/ss will be down for a couple of hours today starting at 1:30 pm PDT for maintenance. Dale

April 25, 2017

Re: [Pharo-users] Hot to retrieve values from Nested Dictionaries

by Markus BÃ¶hm

Tx, I tried Your proposal. Just as an example, it works: (ZnClient new url: 'http://api.wunderground.com/api/APIKEY/hourly/q/germany/munich.json '; contentReader: [ :entity | (NeoJSONReader on: entity readStream) mapClass: NeoJSONObject; propertyNamesAsSymbols: true; next ]; get ) hourly_forecast collect: [:e | e FCTTIME hour_padded asInteger ] P.S.: Editor in Playground shows NeoJSONObject keys in red color? Am 24.04.2017 9:00 nachm. schrieb "Udo Schneider" < udo.schneider(a)homeaddress.de>: > If you are using NeoJSON you might want to take a look at the > NeoJSONObject class. It's basically a Dictionary with "JavaScript" behavior. > I.e. you can simply use messages to access keys and non existing keys > return nil. Your example can thus be written as: > > val := dict1 key1 key2 key3. > > When using NeoJSON you can simply configure your Mapper (Reader/Writer) to > use NeoJSONObject using #mapClass: > > CU, > > Udo > > > On 24/04/2017 20:04, Markus BÃ¶hm wrote: > >> Tx for advice. >> >> P.S.: I got nested dictionaries after JSON API call at wunderground >> weather API. To see all the response in Pharo inspector "live" is really >> great :-) >> >> BR Mike >> >> >> On Mon, Apr 24, 2017 at 7:51 PM, Cyril Ferlicot D. < >> cyril.ferlicot(a)gmail.com <mailto:cyril.ferlicot@gmail.com>> wrote: >> >> Le 24/04/2017 Ã 19:42, Markus BÃ¶hm a Ã©crit : >> > May I ask: >> > What's the idiomatic way to retrieve values from nested >> dictionaries? >> > >> > (((dict1 at: 'key1') at: 'key2') at: 'key3') >> > >> > Are all the brackets really necessary? >> > >> > BR Mike >> >> Hi, >> >> To know if brackets are necessary you can use the "Format" option in >> your method. >> >> Right click -> format >> >> or cmd/alt/ctrl + shift + f >> >> In this case you can just write: >> >> ((dict1 at: 'key1') at: 'key2') at: 'key3'. >> >> >> >> -- >> Cyril Ferlicot >> https://ferlicot.fr >> >> http://www.synectique.eu >> 2 rue Jacques PrÃ©vert 01, >> 59650 Villeneuve d'ascq France >> >> >> > > >

April 25, 2017