Return Styles: Pseud0ch, Terminal, Valhalla, NES, Geocities, Blue Moon.

Pages: 1-4041-

/prog/ challenge

Name: Anonymous 2013-12-24 17:34

Make a captcha-solving program.

Reward: you'll be rich. Not by me. Other people will pay you out the ass though.

Name: Anonymous 2013-12-24 17:43

Step 1: install gentoo

Name: Anonymous 2013-12-24 17:45

Step 2: read SICP

Name: Anonymous 2013-12-24 18:30

Ｓｔｅｐ　３： ？？？

Name: Anonymous 2013-12-25 16:08

Step 4: PROFIT!

Name: Anonymous 2013-12-26 14:13

Step 5: read SICP again

Name: Anonymous 2013-12-26 14:46

Step 6: install Windows and Visual Basic

Name: Anonymous 2013-12-26 14:49

step 7: *grabs dick*

Name: Anonymous 2013-12-26 15:12

>>7
LOL only peopl from 3rd world like Cudeer still use Micro$hit WinDOS in 2013

Name: Anonymous 2013-12-26 15:25

I'm naming mine ``SIX MILLION CAPTCHAS: NEVER AGAIN!'' in honor of Sussman's people.

Name: Anonymous 2013-12-26 15:33

>>10
oy vey :)

Name: Anonymous 2013-12-26 16:55

I'm currently working on machine learning. The only problem about convolutional neural networks for this task is that even it works well against scaling, it doesn't work well with rotation (as seen in Captchas).

Name: Anonymous 2013-12-26 22:22

You need rotation-invariant feature detectors =)

Name: Anonymous 2013-12-27 3:28

ROTATE MY ANUS

Name: Anonymous 2013-12-28 3:53

What's the best trainable neural network? I wouldn't even know where to start writing an OCR (besides loading the image file) but I can solve 1000 Captchas if it means the neural net will be able to learn how to solve the rest.

Name: Anonymous 2013-12-28 9:10

>>15
As I said, convolutional neural networks are the way to go for pictures. But you should add a few rotation-invariant feature detectors as added >>13 .

Name: Anonymous 2013-12-28 13:19

How does Lush sound?
If you do research and development in signal processing, image processing, machine learning, computer vision, bio-informatics, data mining, statistics, simulation, optimization, or artificial intelligence, and feel limited by Matlab and other existing tools, Lush is for you. If you want a simple environment to experiment with graphics, video, and sounds, Lush is for you.
http://lush.sourceforge.net/

Name: Anonymous 2013-12-28 14:13

>>17
Lush is outdated. It was pushed by LeCun and his friends, but now he uses Torch7. I worked with this a bit, it sucked. Hard.
I prefer Octave (Matlab clone), and there's a neat Python framework tightly integrated with Numpy called Theano, that can use the GPU and do dynamic compilation. Check deeplearning.net, there are a few tutorials for logistic regression, Multi Layer Perceptron and finally, CNNs.

Name: Anonymous 2013-12-29 0:38

Implement *Lisp using the EUs on a Haswell and write your own algorithms.

Name: Anonymous 2013-12-29 0:48

>>19
Intel is shit. They would be 6 feet under if IBM chose the 68k for their PC as they originally planned.

Name: Anonymous 2013-12-30 1:25

The hardest part is probably figuring out where the boundaries of the rotated/wavy/distorted characters are.

Name: Anonymous 2013-12-30 1:38

>>20

What's so good about the 68k in comparison with e.g. x86?

Name: Anonymous 2013-12-30 2:48

>>22
Today, not much except for a cleaner and simpler instruction set, but back when the PC came out it was different.
The original 68000 had 32-bit registers (x86 was 16-bit), a flat address space, could address up to 16 MB (24-bit addressing but later 32-bit, like AMD64 is 48-bit/56-bit but extensible to 64-bit, while x86 only allowed 1 MB), a cleaner instruction set, more registers (8 data and 8 address vs 4 data and 4 address), user and supervisor modes (x86 only had real mode), and more assembly addressing modes. 32-bit addressing would have eliminated the need for HIMEM, the 640 K limit, the A20 gate, and DOS extenders.
Before AMD64, x86 had "real", "286 protected", and "386 enhanced" modes but because the 68k designers chose 32-bit pointers it only had one "mode" for its entire lifetime.

Name: Anonymous 2013-12-30 5:44

>>23

Cool. You seem knowledgable about this. There were some obstacles with regards to memory virtualization and the motorolla chips vs intel? What was all that about?

Also why the simultaneous move to PowerPC by Amiga, Apple, IBM and Motorolla. Was that technical at all?

Name: Anonymous 2013-12-30 6:58

>>21
nope, edge detection is easy.
Convolution with a gradient or look the zeros of the Laplacian.

Name: Anonymous 2013-12-30 15:41

>>24
I'm not sure about Amiga, but Apple, IBM, and Motorolla did the simultaneous move because they formed the AIM alliance to develop and maintain PowerPC.

Name: Anonymous 2013-12-31 4:40

>>24
The 68000 had a bus error trap, which had to be used with a second 68k because it was unable to restart the instruction (this was fixed in the 68010, which had full virtual memory). The 8086 didn't have virtual memory at all (its segments were just segment*16+offset). The 286 had virtual memory based on multiple 64K segments and a segment not present exception (for swapping). Segmentation would cut down on buffer overflows, stack smashing, and other vulnerabilities, but isn't compatible with the Unix memory model.

Name: Anonymous 2013-12-31 11:36

I miss my PowerPC Mac running OS9

Name: Anonymous 2013-12-31 15:54

>>28
>le pedophile sage

Name: Anonymous 2013-12-31 16:48

>>29
Great contribution to the thread.

Name: surrealdeal 2013-12-31 17:35

>>27
aka: shit i read about on carolyn meniel's site when I was 12

Name: Anonymous 2013-12-31 18:55

Where's my premade neural net software? It's 2013. Just give us a trainable OCR already.

Name: Anonymous 2013-12-31 18:57

How about a JavaScript userscript captcha solver that automatically works on all sites containing ReCAPTCHA?

Name: surrealdeal 2013-12-31 19:31

>>32
shit never actually works; it's a variation on regressive functions. there's no 'real' 'neural nets'

you'd just come up with a new captcha that uses different font-warp functions, neural net breaks

Name: Anonymous 2013-12-31 19:44

>>34
Then you keep refreshing the CAPTCHA until you get one it can read.

Name: Cudder !MhMRSATORI!fR8duoqGZdD/iE5 2014-01-01 4:35

>>20
And the PC would've likely not sold as well either because it would be far too expensive. No one needed that sort of capability for a PC at the time. CP/M and 8-bitters (6502, Z80) didn't start declining in popularity until the early 90s.

Name: Anonymous 2014-01-01 11:59

And
kill yourself
the
kill yourself
PC
kill yourself
would've
kill yourself
likely
kill yourself
not
kill yourself
sold
kill yourself
as
kill yourself
well
kill yourself
either
kill yourself
because
kill yourself
it
kill yourself
would
kill yourself
be
kill yourself
far
kill yourself
too
kill yourself
expensive.
kill yourself
No
kill yourself
one
kill yourself
needed
kill yourself
that
kill yourself
sort
kill yourself
of
kill yourself
capability
kill yourself
for
kill yourself
a
kill yourself
PC
kill yourself
at
kill yourself
the
kill yourself
time.
kill yourself
CP/M
kill yourself
and
kill yourself
8-bitters
kill yourself
(6502,
kill yourself
Z80)
kill yourself
didn't
kill yourself
start
kill yourself
declining
kill yourself
in
kill yourself
popularity
kill yourself
until
kill yourself
the
kill yourself
early
kill yourself
90s.
kill yourself

Name: Anonymous 2014-01-01 17:07

>>37
upboat

Name: Anonymous 2014-01-01 20:22

>>36
People bought the PC because it had the IBM name on it. They would have bought a turd if it had the IBM brand. DOS 1.0 was one of the worst OSes ever and people still bought it because it was endorsed by IBM! If IBM was as popular today as it was in 1980, they could sell a PC with LoseThos and a 640x480 CRT and get 90% of the business marketshare.

Name: Anonymous 2014-01-01 20:33

>>36-39
http://www.vanwensveen.nl/rants/microsoft/IhateMS_1.html

Name: Cudder !MhMRSATORI!fR8duoqGZdD/iE5 2014-01-02 2:11

>>39
Because IBM meant dependable backwards-compatibility and stability. DOS 1.0 may have been one of the worst but it served the needs of the people at the time. It needn't be any better than that.

One thing that more evangelists and idealots need to understand: people really don't care about innovation, "better design", "elegance", whtaever. No matter how much you try to tell them to care. They just want to get shit done.

Name: Anonymous 2014-01-02 2:16

>>41

I care about understandability and run-time modifiability because I consider this essential to "getting shit done" without it taking forever. Hence my love of Common Lisp, Smalltalk and Forth (more specifically the systems these langauges facilitate(d))

Name: Anonymous 2014-01-02 16:58

>>41
>le kike sage

Name: Anonymous 2014-01-03 7:28

Fuck you >>19, you successfully derailed the thread.

Name: Anonymous 2014-01-04 5:48

>>44
nice dubs
Don't change these.
Name: Email:
Entire Thread Thread List
All trademarks and copyrights on this site are owned by their respective parties. All comments and posts are the responsibility of their own posters.
- Tinychan + 2ch Mode -