I just started playing around with machine learning in Python today. It's so fucking amazing, man!
All the concepts that come up when you search for tutorials on YouTube (you know, neural networks, SVM, Linear/Logic regression and all that fun stuff) seem overwhelming at first. I must admit, it took me more than 5 hours just to get everything set up the way it should be but, the end result was so satisfying when it finally worked (after ~100 errors).
If any of you guys want to start, I suggest visiting these YouTube channels:
- https://youtube.com/channel/...
- http://youtube.com/playlist/...

In my previous rant about IPv6 (https://devrant.com/rants/2184688 if you're interested) I got a lot of very valuable insights in the comments and I figured that I might as well summarize what I've learned from them.

So, there's 128 bits of IP space to go around in IPv6, where 64 bits are assigned to the internet, and 64 bits to the private network of end users. Private as in, behind a router of some kind, equivalent to the bogon address spaces in IPv4. Which is nice, it ensures that everyone has the same address space to play with.. but it should've been (in my opinion) differently assigned. The internet is orders of magnitude larger than private networks. Most SOHO networks only have a handful of devices in them that need addressing. The internet on the other hand has, well, billions of devices in it. As mentioned before I doubt that this total number will be more than a multiple of the total world population. Not many people or companies use more than a few public IP addresses (again, what's inside the SOHO networks is separate from that). Consider this the equivalent of the amount of public IP's you currently control. In my case that would be 4, one for my home network and 3 for the internet-facing servers I own.

There's various ways in which overall network complexity is reduced in IPv6. This includes IPSec which is now part of the protocol suite and thus no longer an extension. Standardizing this is a good thing, and honestly I'm surprised that this wasn't the case before.

Many people seem to oppose the way IPv6 is presented, hexadecimal is not something many people use every day. Personally I've grown quite fond of the decimal representation of IPv4. Then again, there is a binary conversion involved in classless IPv4. Hexadecimal makes this conversion easier.

There seems to be opposition to memorizing IPv6 addresses, for which DNS can be used. I agree, I use this for my IPv4 network already. Makes life easier when you can just address devices by a domain name. For any developers out there with no experience with administration that think that this is bullshit - imagine having to remember the IP address of Facebook, Google, Stack Overflow and every other website you visit. Add to the list however many devices you want to be present in the imaginary network. For me right now that's between 20 and 30 hosts, and gradually increasing. Scalability can be a bitch.

Any other things.. Oh yeah. The average amount of devices in a SOHO network is not quite 1 anymore - there are currently about half a dozen devices in a home network that need to be addressed. This number increases as more devices become smart devices. That said of course, it's nowhere close to needing 64 bits and will likely never need it. Again, for any devs that think that this is bullshit - prove me wrong. I happen to know in one particular instance that they have centralized all their resources into a single PC. This seems to be common with developers and I think it's normal. But it also reduces the chances to see what networks with many devices in it are like. Again, scalability can be a bitch.

Thanks a lot everyone for your comments on the matter, I've learned a lot and really appreciate it. Do check out the previous rant and particularly the comments on it if you're interested. See ya!

So apparently some major vpn connection providers got compromised some time ago.

https://twitter.com/hexdefined/...

https://twitter.com/cryptostorm_is/...

adding the fact that major enterprise vpn network providers had security flaws earlier this year

https://sdxcentral.com/articles/...

Sums up what was the major topic in security this year.

At the end I see something like cloud act that allows wiretapping anyone.

https://justice.gov/opa/pr/...

And when we multiply this by number of companies that have services in cloud that sums up privacy these days.

Non existent.

!rant

Thanks google for giving me the opportunity to work with neural networks without being an expert about them.
http://automl.github.io/auto-sklear...

To sum it up:
1. Preprocess data
2. Use Automl to train classifier
3. ????
4. Profit

http://ai-junkie.com is a brilliant website - it's finally allowed me to understand neural networks and genetic algorithms properly!

Opening Discussion Here,
I am trying to make a simple zen game that use Neural Network. the game is simple just a square object with a certain Viewing Angle and Viewing Distance with an objective is finding a food in a map with some other non food object as an obstacle.

i am encountered some problem. i am trying to find a way to make "seeing object in certain viewing region" and i am come up with two ideas described in the picture below. the problem is, i don't want to feed the NN with too much information, by that i mean i don't want to tell the AI what object it is, i want it to find out what it is by it self. and i cannot find a way to implement this either because of the framework limitation that i use (p5.js) or simply i cannot find a way how to.

i am on my way there tho, currently here i am (in pseudocode):
https://pastebin.com/7Ae1ZNYa

what do you think ?

Coolest project I'll continually be working on.

http://jimquessenberry.com

Selling my Dad's famous BBQ sauces and rubs has been my hobby and passion for years. I'm lucky that my Dad was a computer enthusiast in the 1980's and also had a knack for marketing himself. All the while also being a somewhat famous character in the pioneering sport of competition BBQ cooking.

My brother and I shared the following machines growing up:

Commodore 64 w/ 2 Disk Drives, VicModem, & Tape Drive

Tandy 1000 Original Radio Shack IBM PC Clone

IBM 5150 w/ 20mb Hard Drive Expansion (Still Have This In Near Mint Condition)

Tandy 1000 RSX 386 with Win 3.11 For Networks

A Homebuilt Pentium 90 MHz Tower with Soundblaster and 16bit onboard video.

All that time on those machines learning various flavors of BASIC and crude graphic design got me where I am today.

That and learning how to BBQ... ;)