59
ng1905
3y

1. Started to program Python 3
2. Thought of doing a little program just for fun
3. Now 1 1/2 years in developing state with nearly 3000 lines of code and 11 modules
4. Python skills :)

Comments
  • 1
    I'm at step one right now. Hope I'll also go till 4. ;)
  • 0
    Fuck me, that's a lot of python. What are you doing with all of that code?
  • 2
    @AlgoRythm I'm testing some neural nets with real life data e.g. face and speech recognition, emotional analysis...
  • 0
    @ng1905 any tutorials to follow just to get to understand the concept?
  • 2
    @gitpush For neural networks, take a look at the tutorials and/or documentation of tensorflow... It's a really great api and you can use it as high level if you want to... I started with a simple neural net for the processing of information. If you don't want to start with a face recognition (but it would be a good start, just to understand how it works), you can use Microsoft's Project Oxford to query images to their servers and then get back the face rectangle and information about the face's person (like age, gender,...). My aim was to program something that can somehow "decide" about what to do with the input of human emotions and circumstances... But emotional analysis in general is something I'm really interested in. I didn't follow any tutorials (I don't like them to be honest) even if I'm taking way longer with reading the documentations :)
  • 0
    @ng1905 so you reliaed on existing solution for making your network learn. What if I want it to learn basic commands recognition that has no images just text commands. I've checked few articles but I'm still confused 😞
  • 2
    @gitpush I see... So you're trying to train it that it can get a command and do something right? So first of all, you could start with writing a NeuralNetwork from the ground up (with numpy)... In the process you will learn how it works and what it requires as input... To explain it roughly: As a neural net relays on numbers, you have to find a way to interpret your command as a number... Let's say: "Do xy" has the number 10... Then you have to think of the outputs you want to get... If you have 5 things your network should do, create 5 output nodes... All of these output nodes will give you a number between 0 and 1. If you want the network to execute the second action, you have to train it so you get the highest value on output node 2 if you pass command 10 through the input. The less error quote your network has, the better it will give you the right output.
  • 1
    @ng1905 that's exactly what I'm trying to achieve. Now the bad part is I don't know python, though I'm planning to use for this purpose as I know it is one of the easy and fast Lang's for computations . Anything else you'd recommend a newbie in python?
  • 2
    @gitpush I think that python is a great way to start with such things... Just learn the syntax, take care of 4 space indentations instead of { and } and start programming. You will get used to it pretty quickly ;)
  • 2
    @gitpush And don't try to use a neural net for yes/no decisions... Look at it more like probabilities of what's the best you could want from the network
  • 1
    @ng1905 thanks bro for your help really appreciate it 😃
    Let's hope I can learn Python without troubles
  • 2
    @gitpush Any time. If there's something I can do, just comment or dm me (is that even possible? :D)
    Good luck :)
  • 1
    @ng1905 thanks bro :)
    By the way nice website (and nice Google chrome that made me understand German :p)
  • 2
    @gitpush I see :D
    Thanks, appreciate it :)
Add Comment