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

Re: [Pharo-users] Neural Networks in Pharo

by Alexandre Bergel

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] type checking in Smalltalk

by Igor Stasenko

On 31 March 2017 at 21:34, Sven Van Caekenberghe <sven(a)stfx.eu> wrote: > > > On 31 Mar 2017, at 19:38, Dimitris Chloupis <kilon.alios(a)gmail.com> > wrote: > > > > > > On Fri, 31 Mar 2017 at 19:13, Sven Van Caekenberghe <sven(a)stfx.eu> > wrote: > > if you copy/paste something you should give a reference > > > > > > I did not copy paste anything, 100 % mine. What part you think it's copy > ? > > Ah, sorry then, the formatting seemed to suggest it came from somewhere > else. > > BTW, I don't think the hybrid part is a real thing. > > Although I understand that static typed languages like C, C++, Java and so > many others have their place and use, people in those languages spend an > awful amount of time and code dealing with the types they love so much. The > modern static languages with all their magic are even worse. And even in a > nice static typed program that compiles with no warnings, there are still > dynamic errors lurking everywhere. The world cannot be defined in a static > way. > > The dynamic type errors that you get in Pharo during development are 95% > plain logic errors, once you fix those and write some unit tests, a Pharo > program is just as stable and reliable as anything else that is well > written, tested and debugged. > > And I love Igor's "Don't ask, tell" idea - right on target. > > Hey, hey.. These words, originally, definitely are not mine.. It is something i learned, being among wise people of smalltalk family. > Anyway, that is my personal opinion, I don't want to convince you. > > Sven > -- Best regards, Igor Stasenko.

April 25, 2017

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

by Esteban A. Maringolo

2017-04-24 16:10 GMT-03:00 Ramon Leon <ramon.leon(a)allresnet.com>: > On 04/24/2017 11:16 AM, Esteban A. Maringolo wrote:> >> For these use cases it >> would be nice to have some sort of syntax sugar >> as with the cascade operator, but with a chaining operator instead. > This doesn't even require syntactic sugar, just objects, and this is a more > general problem even with a back to back chain of select/detect/reject style > things. I solved this for myself years ago by introducing what I call a pipe > (nod to unix) allowing me to chain calls in a pipeline without all the > parens with each call acting on the result value of the last call. Trivially > implemented using #doesNotUnderstand: and the cascade operator. > > ^ dict1 asPipe at: 'key1'; at: 'key2'; at: 'key3'. > > Anytime I have a bunch of back to back calls that would require a lot of > parens, I just create a pipe... > > maxRoomsAvailable > ^self findBlocks asPipe > select: [ :e | e blockDate between: actualCheckInDate and: > actualCheckInDate + (nights - 1) days ]; > detectMin: [ :e | e quantityAvailable ]; > quantityAvailable Very neat implementation. I think this simple class could come as part of the base image :) Esteban A. Maringolo

April 24, 2017

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

by Ramon Leon

On 04/24/2017 11:16 AM, Esteban A. Maringolo wrote:> For these use cases it would be nice to have some sort of syntax sugar > as with the cascade operator, but with a chaining operator instead. This doesn't even require syntactic sugar, just objects, and this is a more general problem even with a back to back chain of select/detect/reject style things. I solved this for myself years ago by introducing what I call a pipe (nod to unix) allowing me to chain calls in a pipeline without all the parens with each call acting on the result value of the last call. Trivially implemented using #doesNotUnderstand: and the cascade operator. ^ dict1 asPipe at: 'key1'; at: 'key2'; at: 'key3'. Anytime I have a bunch of back to back calls that would require a lot of parens, I just create a pipe... maxRoomsAvailable ^self findBlocks asPipe select: [ :e | e blockDate between: actualCheckInDate and: actualCheckInDate + (nights - 1) days ]; detectMin: [ :e | e quantityAvailable ]; quantityAvailable The simplest implementation is... Object>>asPipe ^ SPipe with: self Object subclass: #SPipe instanceVariableNames: 'value' classVariableNames: '' poolDictionaries: '' category: 'PharoExtensions' privateValue: anObject value := anObject yourself ^ value yourself doesNotUnderstand: aMessage ^ value := value perform: aMessage selector withArguments: aMessage arguments SPipe class>>with: anObject ^ (self new) privateValue: anObject; yourself -- Ramon Leon

April 24, 2017

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

by Udo Schneider

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 24, 2017

Re: [Pharo-users] [Ann] Pharo Sprint App to improve coordination during Pharo Sprints

by K K Subbu

On Monday 24 April 2017 08:00 PM, Juraj Kubelka wrote: > > Please, can you send me a screenshot? Or submit the screenshot > here https://github.com/JurajKubelka/PharoSprint/issues/new Done. I think the error appears because when a login fails, the message causes a horiz scroll bar to be added to the panel pushing the fields beyond the top edge. I stumbled on this error because I tried using my fogbugz id whereas the login field needs the registered email id for login. Regards .. Subbu

April 24, 2017

Sending byte array in a JSON format

by Juraj Kubelka

Hi, I was playing with GitHub Gist API and I have queried the following Gist: https://gist.github.com/mbostock/5503544 <https://gist.github.com/mbostock/5503544> I was interested how the PNG image is returned: https://gist.github.com/mbostock/5503544#file-thumbnail-png <https://gist.github.com/mbostock/5503544#file-thumbnail-png> I can obtain the whole Gist executing: STONJSON fromString: (ZnClient new url: 'https://api.github.com/gists/5503544'; accept: 'application/vnd.github.v3+json'; get). I can get PNG contents executing: pngData := (ZnEasy get: ((((STONJSON fromString: (ZnClient new url: 'https://api.github.com/gists/5503544'; accept: 'application/vnd.github.v3+json'; get)) at: 'files') at: 'thumbnail.png') at: 'raw_url')) contents. PNGReadWriter formFromStream: rawPng readStream. But the PNG image is part of the Gist query and can be retrieved by: pngContent := ((((STONJSON fromString: (ZnClient new url: 'https://api.github.com/gists/5503544'; accept: 'application/vnd.github.v3+json'; get)) at: 'files') at: 'thumbnail.png') at: 'content'). "As pngContent is a WideString, I cannot use:" PNGReadWriter formFromStream: pngContent readStream. How can I read the PNG image from the pngContent? Any idea? And the reverse question: How can I send the PNG bytes using JSON format? Thanks! Juraj

April 24, 2017

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

by Peter Uhnak

Maybe we could introduce `@` message, so one can chain them and then even execute keyword without parenthesis Dictionary>>@ aKey ^ at: aKey dict @ 'key1' @ 'key2' @ 'key3' dict @ 'key1' @ 'key2' @ 'key3' select: [ :each | ... ] I'm always adding this when hacking around, maybe we could have it in Pharo. Peter On Mon, Apr 24, 2017 at 02:48:48PM -0300, Juraj Kubelka wrote: > > > On Apr 24, 2017, at 14:42, Markus BÃ¶hm <markus.boehm(a)gmail.com> wrote: > > > > 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? > > I think so, without the parenthesis, it is interpreted as one message at:at:at: > > If you use it a lot, you may want to convert the dictionary structure into domain objects. > > Juraj > > > > > BR Mike > >

April 24, 2017

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

by Esteban A. Maringolo

For these use cases it would be nice to have some sort of syntax sugar as with the cascade operator, but with a chaining operator instead. The chaining operator would send the message to the result of the previous statement instead of the original receiver as in the semicolon cascade operator (;). E.g. dict1 at: 'key1' : at: 'key2' : at: 'key3' . I selected the colon character in the lack of non-message related characters. It's only a daydream and nothing that I found very often needing, except for cases like the mentioned in this thread, extracting a deeply nested value from a JSON object tree/dictionary. ps: use the extra parenthesis, they're not that many. :) Regards! Esteban A. Maringolo 2017-04-24 14:48 GMT-03:00 Juraj Kubelka <juraj.kubelka(a)icloud.com>: > >> On Apr 24, 2017, at 14:42, Markus BÃ¶hm <markus.boehm(a)gmail.com> wrote: >> >> 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? > > I think so, without the parenthesis, it is interpreted as one message at:at:at: > > If you use it a lot, you may want to convert the dictionary structure into domain objects. > > Juraj > >> >> BR Mike > >

April 24, 2017

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

by Markus BÃ¶hm

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 > 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 24, 2017