Talk in English - US at Dutch PHP Conference 2015
View Slides: http://es.slideshare.net/egulias/ann-dp-con-2015
Short URL: https://joind.in/talk/e8639
(QR-Code (opens in new window))
Artificial Neural Networks on a Tic Tac Toe console application
Comments are closed.
It seemed Murphy got in the room. The theorem of ANN was nice, though the way it was presented could be more entertaining IMnsHO. The technical part is really good though.
Really asome stuff but i think that if you leave the Symfony stuff out of your talk you could tell more about how it really works in symfony. Know you have a lot of code which get confusing if you jump from one part to an other part.
Leave out the symfony, it was not relevant. It will give you more time to focus on the essence. The theory was interesting, but brought a bit too fast / too dry. The technical part was pretty interesting, looking forward to reading more about the topic and playing with the code myself.
I personally missed a bit more in-depth theory on neutral networks, and actual pointers on how to use/implement the Fann library. A simpler demo application (of 1-3 php files) could have made the examples more clear. I would advise to leave out the symfony console part; that time could have been spent on much more interesting content :)
It was a bit chaotic through the fact it was a talk about both neural networks as well as the Symfony console. Would have been better if Symfony was left out. 45 minutes for the subject alone was already too tight.
The presentation was too bland for people who already knew about ANN. People who didn't, did not seem to understand the speaker's explanation. I would advise the speaker to assume the audience has a basic understanding of ANN, or explain how an ANN works more slowly.
Nn's are simple but here explained badly. Would have been nicer to show more of the use cases.seeing the learningprocessas it happens was really nice though.
I was hoping to learn a bit more on how to create an neural network in PHP. Unfortunately the main focus seemed to be the Symfony Console application which was a bit sad. Maybe split the talk into 2, one focusing on the neural network problem, the other focusing on the Symfony Components.
Symfony console is great. But it did not need to be explained for this to work.
I was waiting to find out how or where the network was trained but I missed that. Where and how did the network find out what was a bad decision and a good decision? The demos were too fast. "Player X won" but you never see the last move or end state of the board.
The speaker needs to get his basic story straight. Inputs, outputs, hidden layers; they're not rocket science. However it seemed that he did not prepare the words and right imagery to explain his story. The whole talk was hard to follow and in the end the most interesting parts, like the weight calculation and back propagation, were totally skipped over. All we heard for the last 10 minutes was: "here's a 1.0 and here's a 0.5 and 0.25 is lower than 0.5." Yes we know, but how did those numbers get there?!
Another note: don't waste too much time talking about libfann. It's just a tool. Its applications in the practical world is more interesting.
Remove the whole Symphony helpers part. It was totally irrelevant. Thanks.
Very good presentation with huge potential.
IMHO leave Symfony aside and add more details about the core - weight calculation and back propagation.
In this presentation I was hoping for a practical explanation of how to implement a ANN. Instead we got a lot of explaining about Symfony console and the helpers you build, and litte info about the ANN.
If I could make a suggestion, try to minimize the Symfony part en expand your explanation of FANN and how to use it (from scratch?) in php.
Okay, so we used a library. I still don't know what is happening behind the scenes but we made something.
Perhaps would be nice to try explain more about how neurons "talks" to other neurons forming a network.
I wish you had gone into more detail about ANN, and perhaps a little less about Symfony (but thanks for the link to the BoardHelper library!); for a technical talk there wasn't enough on network weights, how they are calculated, backwards propagation, etc.
However, it was presented well and it certainly captured my interest - because of this talk I will definitely be researching this topic a lot more with a hope of implementing it in a pet-project somewhere!
ANN is a really cool thing, but the talk was a bit dry. would have loved to see the result: a smart AI player after learning.
I agree with others to leave symphony out of it. Symphony console is cool, but i would suggest more focus on the real cool parts: AI
As other people have said i expected a bit more "Artificial Neural Networks" and less "console application". The talk was good none the less but would have been better with a bit of focus change.