Anybody code with Python?

**Bettyboo** · 27-02-2019, 09:29 PM

Well???

I bloody well hate coding, programming, maths; all terrible stuff I try to avoid, but I am having to develop a bit of understanding of Python. Was wondering what other folks thought the best route was?

I have Miniconda3 aswellas Python 2.7 and 3.7 through PyCharm - because it seems easy! I prefer to do stuff locally rather than on a cloud, and much as I don't want to am running through some basic NNs with deep learning hidden layers.

Any suggestions/advice?

**britanicus** · 27-02-2019, 09:40 PM

I watched a bit of this 4 hours course, not really interested but it might be of use if you are

**foobar** · 27-02-2019, 10:06 PM

Originally Posted by Bettyboo

Any suggestions/advice?

You could buckle down and start writing code instead of procrastinating on TD, feverishly researching the demise of the EU and the invisible hand of Soros etc.

Start with Hello World!
---

Which other languages do you know and what level?

What will python be used for? ..give examples.

If teaching..what level?

**Bettyboo** · 27-02-2019, 10:47 PM

^^ I'll have a look at that, thanks - can't hurt.

^ assembly, pascal and C - haven't used them in years; Beng in electronic engineering when I was young.

I did explain, deep learning neural networks. I won't be writing the code, I have some already, but I want to understand it and make adaptions/look at alternatives. &, yes, I am getting on with it; wrote several little, very little, bits of code today. I simply wondered if anybody here is experienced in the field, and might offer some suggestions.

**Looper** · 27-02-2019, 10:57 PM

Why don't you do a NN thread boo.

What is the big picture here?

What are you trying to do?

I am reading 'The Age of Em: Work, Love and Life when Robots Rule the Earth'

But it is not as good as I was hoping. Hanson is just reeling of the economic challenges that face human consciousness emulation.

**Neverna** · 27-02-2019, 11:00 PM

Originally Posted by Bettyboo

I simply wondered if anybody here is experienced in the field, and might offer some suggestions.

Troy is probably your man, Betty.

**Bettyboo** · 27-02-2019, 11:06 PM

^ Troy, the German?

Originally Posted by Looper

Why don't you do a NN thread boo.

A thread on coding neural networks - not sure a lot of folks will enjoy that...

My work is on spiking neural networks which are different, but I want to compare SNN performance to NNs with little autonomous AI. Boring stuff really when you get down and start trying to engineer it end-to-end.

**baldrick** · 27-02-2019, 11:12 PM

I watched this the other day - I enoyed it -

code academy has beginner tutorials and you do not even need to install and compile anything - the basics are still the basics in any language

https://www.codecademy.com/learn/learn-python

**Bettyboo** · 27-02-2019, 11:17 PM

^ I'll have a look, thanks.

**foobar** · 27-02-2019, 11:20 PM

Are you already an expert on deep learning neural networks?

If not, and reviewing advanced code in a language you are not familiar, less of a steep learning curve and more a sheer vertical face.

Your previous experience of low level langs will help some, but not much if its a been a while, especially if you don't enjoy writing code.

Keep writing code....but I don't mean just inputting examples.

How about: how many 3 letter words in a block of text or how many seconds have passed between 1970 and the time of this post.

Just start writing code and learn-as-you-go.

I would put some time into the python side of things before moving on to code review.

**Troy** · 28-02-2019, 02:08 AM

Originally Posted by Bettyboo

^ Troy, the German?

555...Not going to turn me into a square head just yet...

I do use Python but only for scripts and basic guis using qt. I don't have it on my super-duper new win10 machine but I might throw a version on some time, especially if you get me interested in some of this AI work.

If I were you I would start with the Neural Network python libraries and find one that you feel comfortable with. Once that is sorted find out which versions of Python it supports and then go for the latest version. When I say a library you feel comfortable with I mean check out the tutorials and see how easy they are to follow. Do you have a particular library in mind or are you still searching for a good one?

This isn't really my forte, I specialise in real-time software however, I did find several libraries that might be of interest:

1. Theano: Apparently the tutorial for this is pretty darn good and takes you from feet firmly on the floor to Rees-Mogg level in lots and lots of steps. The link to the tutorial:

Deep Learning Tutorials ? DeepLearning 0.1 documentation
2. TensorFlow: There are a few youtube videos on this if you prefer to learn that way. I found a tutorial here: https://www.journaldev.com/18175/pyt...rflow-tutorial
3. NeuPy: A tutorial here: Tutorials ? NeuPy

The list is not exhaustive but I'm a bit tired. The first one, Theano, looks like one to take a serious look at.

If you want a basic Python book to get into it then: http://marvin.cs.uidaho.edu/Teaching...onTutorial.pdf should give you a good foundation.

**Bettyboo** · 28-02-2019, 10:53 AM

^ good stuff, Troy.

I've been doing similar to some stuff you mentioned. I've been working through this online book because I'm interested in sensory AI applications, and this is quite easy:

Neural networks and deep learning

This code: https://github.com/mnielsen/neural-n...src/network.py

I will then apply that on a small AI model (not virtual or simulation, but a real little monster) to which I'll add elements and make it autonomous with various sensory inputs, a homeostatic base, motivation drives, hierarchy, etc (that'll take some time, but a lot of the theoretical work has been mapped out already). It'll likely be using SNNs, but NNs are quite straight forward for stuff like vision, so I want to understand that area generally too. My research suggests that many of the underlying models have been adapted over the years inline with conventional maths (particularly graphs and differentiation), not neurobiology, so they can be fundamentally adapted/improved to show significant performance gains (stuff such as compartmental models are based on science from 1952 which has changed a lot, but the NNs/SNNs is still based on incorrect and out-dated underlying science).

**baldrick** · 28-02-2019, 12:02 PM

you might find this article interesting batty - a bit of a play with Generative Pre-trained Transformer-2 (GPT-2), a Franken-creation from researchers at the non-profit research organization OpenAI.

https://arstechnica.com/information-...-fake-text-ai/

**Bettyboo** · 28-02-2019, 12:42 PM

Originally Posted by baldrick

you might find this article interesting batty - a bit of a play with Generative Pre-trained Transformer-2 (GPT-2), a Franken-creation from researchers at the non-profit research organization OpenAI.

https://arstechnica.com/information-...-fake-text-ai/

I liked this bit:

In 2016, researcher Robbie Gallagher presented such a project at Shmoocon, where he used a Markov chain bot based on Reddit personal finance posts to build HoneyPhish—a bot that responded to Nigerian "419" scam emails with statistically constructed replies that made the scammers think they had a real human on the other end.

**Troy** · 07-03-2019, 02:25 PM

Let me know which version of python you download Betty and which library. I mentioned to a colleague who said tensorflow without prompting or suggestion...and he is German.

Thread: Anybody code with Python?

Thread Tools

Display

Anybody code with Python?

Thread Information

Users Browsing this Thread

Posting Permissions