Machine Learning
Im making a neural network that would guess the price of an item. The input is the text converted into a number and the output would be the price that is likely for the item. Of course though I am very pessimistic about my programming abilities and would like to colab with anyone who is willing to help.
Stuff we will be doing:
- Python 3
- Neural Networks
- Dictionaries and Lists
- Lots of Math (mostly multiplication)
If you were invited, here is the link to the project :D
When a person inputs some text and the price then you should store that to use as training data later, maybe using jasonstore.io which you can find tutorials on repl.it, but that is currently down so maybe a text file, also how are you planing on making the nero net, are you going to feed it each work and get a price for each word then add the prices, i could help with that along with gathering training data, i am also trying to learn how to make good nero nets, so i would not be super useful on that, but i would be willing to help as i learn
@BenjaminNamika1 thanks for the recommendation! I will probably be using php to store it into a json file. :D
@ChezCoder Nice! Good idea
@BenjaminNamika1 thank you!
Are you planning on using a framework (e.g. Pytorch, Tensorflow, MxNet)?
@CaptainAnon not really, im pretty knew to deep learning/machine learning so I don't really know any frameworks yet. But I will look into those :)
@ChezCoder Cool. They tend to make stuff run faster. Do you plan on running the stuff on repl.it?
@CaptainAnon In any case, I'd be happy to help.
@CaptainAnon yeah sure, welcome to the crew :D
I can help with three things:
+python 3
+dictionaries and lists
+math
That is what I can help with
@ChezCoder I can help
@Codemonkey51 sure, invited :D
@ChezCoder can I make string to int us ord() it's a built in function of what you want
@Codemonkey51 im online, lets talk in the chat
@ChezCoder hi?
U on now?
@Codemonkey51 im on
I guess I'm outdated, haha. I'll see what I can learn and I'll edit this comment. :D
Edit #1: A good introductory for neural networks would be this link: https://www.kdnuggets.com/2018/10/simple-neural-network-python.html
Edit #2: Haha, so forget about me, lol. Way better at JavaScript in all this than Py3. Wish you luck, and I will attempt this in JavaScript. :P
Edit #3: Couldn't you just source data from research studies, like this? https://repl.it/@JacksonPhilips/Car-prices-on-average-by-group
The same thing, but with an if() statement: https://repl.it/@JacksonPhilips/Car-prices-on-avg-but-if
I don't understand how you could do this without having to source from websites on average banana prices, average apple prices, and other fruity prices.
Also, neural networks need to be trained, and in my opinion, are far too complex for what you're trying to achieve.
Here's a simple neural network: https://repl.it/@JacksonPhilips/LightgreyFruitfulDevice
Hope you find this helpful! :D
@JacksonPhilips thanks :D
use one hot encoding for ur fruits haha. OR USE DUMMY VARIABLES - thats better. If u want to make this neural net for production dont use python like what finlay44111 said . However,If u want to learn and build and develop it then use python
I can help
@generationXcode sureo, invite sent
The input is the text converted into a number
Interesting but understand that the number needs to really define the item. I would highly recommend understanding and experimenting with word embeddings in which similar words have similar numerical representations.
As an example, if you train it on the price of a banana, it should represent apples in a similar fashion as they are both fruits and your own neural network would deduce a price similar to the banana.
@adityakhanna its better to use one hot encoding for these kind of tasks I think
@adityakhanna im thinking that it will learn from the values. The conversion is super simple:
def convert(strToConvert): result = "" chars = [ " ", "a", "b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m", "n", "o", "p", "q", "r", "s", "t", "u", "v", "w", "x", "y", "z", "0", "1", "2", "3", "4", "5", "6", "7", "8", "9" ] for letter in strToConvert: result = result + str(chars.index(letter.lower())) return result
@generationXcode A one-hot encoding is a way of representing text as matrices (for input into neural networks). An embedding can be applied to the one-hot encoding.
@CaptainAnon ik what onehot encoding is lmao
@generationXcode I just wanted to provide extra clarity.
@CaptainAnon oh hehe
I would like to help, but I can’t, because:
- I have no idea what a neutral network is
- I am bad at math (mostly multiplication)
- I am already doing a multiplayer
if you want some tips on making the neural network you can take a look at my repl names 'multi layer perceptron', it is a standard neural network that can use machine learning to optimise a function. you could also take a look at 3blur1brown's video series on it. I would not recommend python for making it, unless you are making a very small one, since python is horrendously slow.
@finlay44111 alright, thanks :D