If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(NBC News)   17 year old girl builds neural network to analyze breasts   (nbcnews.com) divider line 97
    More: Hero, boobs, minimally invasive procedure, Galapagos Islands, Google Science Fair, cancer types  
•       •       •

18388 clicks; posted to Main » on 21 Aug 2012 at 1:30 PM (2 years ago)   |  Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



97 Comments   (+0 »)
   
View Voting Results: Smartest and Funniest

Archived thread

First | « | 1 | 2 | » | Last | Show all
 
2012-08-21 02:25:04 PM

Yanks_RSJ: Odd Bird: BeerBear: Odd Bird: GIS fail, this is the largest pick I could find

[static2.dmcdn.net image 160x120] 

/would participate in her study

don't think that's her. someone already posted her pic

see post above yours for clue, if that's what floats your schooner

You dumb bastard.


Fantastic exchange. Would lol again.
 
2012-08-21 02:26:04 PM
The drug companies will probably stop this right in its tracks if she really did accomplish it. They don't want a cure. Cures stop the money stream. They want cancer and they want to sell you drugs to treat it.
 
2012-08-21 02:26:44 PM
Where are all the "girls suck at math and science" farkers now?
 
2012-08-21 02:27:20 PM

brap: I tried to pull that stunt at the Science Fair too.  Unfortunately the judges did not award my brilliant invention "The Fantastic Gropinator" with the blue ribbon.  Apparently, medical science was not ready for the technology that is my hands, mouth, and rejected my patent for the "special veiny probinator meat wand cancer detector".


I bet you're republican. Please share with us your knowledge of female anatomy, and why rape is always the woman's fault.

\+1 for you.
 
2012-08-21 02:31:40 PM
Her father, Director of the Neural Net Breast Cancer Investigation Network, commented: "She did it all her self (wink-wink). No help from me or my team of breast cancer neural net researchers (wink-wink)".
 
2012-08-21 02:39:39 PM
As a social conservative, I have to say that women's health is not where this technology should be used. There are too many men's health issues that need attending! Why is this little girl not in the kitchen or in the breeding room?
 
2012-08-21 02:39:43 PM

95629: The drug companies will probably stop this right in its tracks if she really did accomplish it. They don't want a cure. Cures stop the money stream. They want cancer and they want to sell you drugs to treat it.


Detection != Cure
 
2012-08-21 02:42:34 PM
Ppppffft. I've had a breast analyzing neural network for almost 50 years, and MINE does legs and asses too.
 
2012-08-21 02:43:09 PM

Moonfisher: Where are all the "girls suck at math and science" farkers now?


Every couple of centuries, a statistically insignificant exception to the rule comes along. For example: Marie Curie, Sophie Germain, Ada Lovelace, Admiral Grace Hopper, and... er... well that's about it.
 
2012-08-21 02:58:15 PM

WhippingBoy: Moonfisher: Where are all the "girls suck at math and science" farkers now?

Every couple of centuries, a statistically insignificant exception to the rule comes along. For example: Marie Curie, Sophie Germain, Ada Lovelace, Admiral Grace Hopper, and... er... well that's about it.


pythagorus' sex cult derived all their mathematics knowledge from the women concubines. you see, the women developed a complex skin art that involved the precise laying of triangles, what they called a tri-force. through the systematic design and re-design of the tri-force, they unlocked the fundamental nature of the triangle, and in it, all of math. however, women were forbidden from teaching in the agora, so pythagorus would take their theorems and profess them as if they were his own.
 
2012-08-21 03:09:53 PM

This text is now purple: "she claims"

This is why science has peer-review and evidence-based results.


She's 17. There probably are errors in her work. As we type this, she's probably trying to fix those errors and we're posting snark on fark.
 
2012-08-21 03:10:19 PM

FlyingV936: 95629: The drug companies will probably stop this right in its tracks if she really did accomplish it. They don't want a cure. Cures stop the money stream. They want cancer and they want to sell you drugs to treat it.

Detection != Cure


What the drug companies really want is a cancer detecting drug that produces a lot of false positives for a type of cancer that can only be treated by taking another drug once a day everyday for the rest of your life.
 
2012-08-21 03:15:32 PM
This text is now purple:

1. What did she do about the local maxima problem?

Dunno.

2. She tuned her algorithm to her data set. Of course it did well on that set. How does it do on other data sets?

Her slides say that she did leave-out validation. Slides 7 and 8 suggest that she did validation on several different data sets, but don't give enough information to determine if those were truly withheld or if she did something different (e.g., new leave-out tests on each set).
 
2012-08-21 03:15:33 PM
 
2012-08-21 03:16:57 PM

This text is now purple: Nemnoch: This text is now purple: GAT_00: She built a training neural network? Fun stuff. I want to see the output SOM, assuming she used one.

Two questions:

1. What did she do about the local maxima problem?
2. She tuned her algorithm to her data set. Of course it did well on that set. How does it do on other data sets?

Per the article she claims that it would do well on other cancers with minor tweaking. If so, it would be amazing.

Per the article she claims that it would do well on other cancers with minor tweaking. If so, it would be amazing.

"she claims"

This is why science has peer-review and evidence-based results.


Hence the "if so" qualification. Sheesh, lighten up.
 
2012-08-21 03:22:25 PM
Let's see the Muslims do that.
 
2012-08-21 03:29:38 PM
Phhbbt, big deal. A neural network is actually surprisingly easy to build, if you have a good stats software and know the basics. Weka, an open source software, does the job surprisingly well.

A model can look intimidating at first, but makes perfect sense when you understand each node and value.

Yes, I need to get out more.
 
2012-08-21 03:31:06 PM
Get her together with that kid who invented the early test for pancreatic cancer.
 
2012-08-21 03:33:48 PM
We already have a neural network that analyzes breasts; it's called a 17 year old boys.
 
2012-08-21 03:47:06 PM

This text is now purple: GAT_00: She built a training neural network? Fun stuff. I want to see the output SOM, assuming she used one.

Two questions:

1. What did she do about the local maxima problem?
2. She tuned her algorithm to her data set. Of course it did well on that set. How does it do on other data sets?


Assuming it is a detection algorithm, she's built a set of likely values and ranges and when you add the new data point, it can tell if it is within the cancer or not cancer part of the network. That's at least the way a training SOM should work.
 
2012-08-21 03:51:52 PM
But all it does is tell her she will die in 15 years instead of finding it later and telling her she will die in 7 years.
 
2012-08-21 03:53:36 PM

limboslam: Let's see the Muslims

Institute for Creation Research do that.

FTFY.

Some of the best programmers I know are muslims.
 
2012-08-21 03:55:58 PM
What do you need to BUILD one for? I've already got a neural network that is happy to analyze breasts.

/It always seems to output "I like!", so maybe it's not as discriminating as she wants.
 
2012-08-21 03:58:36 PM

Ambitwistor: This text is now purple:

1. What did she do about the local maxima problem?

Dunno.

2. She tuned her algorithm to her data set. Of course it did well on that set. How does it do on other data sets?

Her slides say that she did leave-out validation. Slides 7 and 8 suggest that she did validation on several different data sets, but don't give enough information to determine if those were truly withheld or if she did something different (e.g., new leave-out tests on each set).


??????????

It was my understanding there would be no math.

youbentmywookie.com
 
2012-08-21 04:03:59 PM

Banned on the Run: It was my understanding there would be no math.

[youbentmywookie.com image 700x438]


Only need to count to two!

i45.tinypic.com
 
2012-08-21 04:07:46 PM

I May Be Crazy But...: Banned on the Run: It was my understanding there would be no math.

[youbentmywookie.com image 700x438]

Only need to count to two!

[i45.tinypic.com image 453x589]


Or maybe a smaller, NSFW two.
 
2012-08-21 04:13:12 PM
That's funny. The byline is John Roach, which was the name of my college AI professor. Who, oddly enough, was quite the skeptic about neural nets -- he was more of a theorem-prover and expert-systems guy.
 
2012-08-21 04:34:04 PM

I May Be Crazy But...: I May Be Crazy But...: Banned on the Run: It was my understanding there would be no math.

[youbentmywookie.com image 700x438]

Only need to count to two!

[i45.tinypic.com image 453x589]

Or maybe a smaller, NSFW two.


Oh Sweet Lord. Those are perfect.
 
2012-08-21 04:35:53 PM

McGrits: I May Be Crazy But...: I May Be Crazy But...: Banned on the Run: It was my understanding there would be no math.

[youbentmywookie.com image 700x438]

Only need to count to two!

[i45.tinypic.com image 453x589]

Or maybe a smaller, NSFW two.

Oh Sweet Lord. Those are perfect.


I saw her first!
 
2012-08-21 04:38:34 PM

I May Be Crazy But...: I May Be Crazy But...: Banned on the Run: It was my understanding there would be no math.

[youbentmywookie.com image 700x438]

Only need to count to two!

[i45.tinypic.com image 453x589]

Or maybe a smaller, NSFW two.


I'll be back for this.
 
2012-08-21 04:51:31 PM
I wanted to leave some time for folks to do other stuff if they're wanting to talk about neural networks, but since it's pretty quiet...

i49.tinypic.com
 
2012-08-21 04:55:08 PM

I May Be Crazy But...: I wanted to leave some time for folks to do other stuff if they're wanting to talk about neural networks, but since it's pretty quiet...

[i49.tinypic.com image 431x600]


Can never have too many hot chicks in any thread, please continue!
 
2012-08-21 04:55:30 PM
i50.tinypic.com
 
2012-08-21 04:58:38 PM

Nemnoch: I May Be Crazy But...: I wanted to leave some time for folks to do other stuff if they're wanting to talk about neural networks, but since it's pretty quiet...

[i49.tinypic.com image 431x600]

Can never have too many hot chicks in any thread, please continue!


No, I refuse.

i48.tinypic.com

Okay, I lied.
 
2012-08-21 04:59:55 PM
Another NSFW for those who can view 'em.
 
2012-08-21 05:07:31 PM
i49.tinypic.com

And I'll let y'all get back to the computer science.
 
2012-08-21 05:14:09 PM
As a neural network expert, I think it's time to face the truth. My carrier took a wrong turn somewhere in the past.
 
2012-08-21 05:30:10 PM

I May Be Crazy But...: Banned on the Run: It was my understanding there would be no math.

[youbentmywookie.com image 700x438]

Only need to count to two!

[i45.tinypic.com image 453x589]


www.wizards.com

Two. Not more than two...
 
2012-08-21 06:34:28 PM

Ambitwistor: This text is now purple: Nemnoch: This text is now purple: GAT_00: She built a training neural network? Fun stuff. I want to see the output SOM, assuming she used one.

Two questions:

1. What did she do about the local maxima problem?
2. She tuned her algorithm to her data set. Of course it did well on that set. How does it do on other data sets?

Per the article she claims that it would do well on other cancers with minor tweaking. If so, it would be amazing.

Per the article she claims that it would do well on other cancers with minor tweaking. If so, it would be amazing.

"she claims"

This is why science has peer-review and evidence-based results.

Hence the "if so" qualification. Sheesh, lighten up.


I just run into a lot of overly-optimistic claims in the press that usually fall all over themselves in the lit or once thumbs get taken off scales.

Most new ideas are crap. Most of the crappy solutions we currently have are crappy because the problem is hard, not being of avarice or because they don't care. Most promising new therapies are just as bad or worse as the old crappy therapies. 

I'm not saying she didn't do something great. I'm just saying it's unlikely.
 
2012-08-21 07:23:22 PM
When does it become self aware and decide our fate in an instant?
 
2012-08-21 07:38:15 PM

Torhathel: When does it become self aware and decide our fate in an instant?


That's the beauty of the design: the innate fascination that all intelligent beings have for breasts will serve to blunt
the formation of any evil ambitions in the AI construct.
 
2012-08-21 07:58:35 PM
17 years old, female, and from Florida...way to kill some stereotypes, Brittany!
 
2012-08-21 10:17:45 PM

95629: The drug companies will probably stop this right in its tracks if she really did accomplish it. They don't want a cure. Cures stop the money stream. They want cancer and they want to sell you drugs to treat it.


This this thissy this this.

/this
 
2012-08-22 02:54:09 AM

Liz Lemon: 95629: The drug companies will probably stop this right in its tracks if she really did accomplish it. They don't want a cure. Cures stop the money stream. They want cancer and they want to sell you drugs to treat it.

This this thissy this this.

/this


Data set is here, the diagnostic method is about 20 yrs old. I worked with this data while taking doing my girlfriend's homework for a statistical learning class. The data set is very friendly, we achieved at least 98% accuracy with various classification models and I think the original authors did a little better with a different method. Then they moved the decision boundary over closer to the non-cancer samples to reduce the false negative rate.

The overall scope and polish of the project seem excellent for a high school student. If I understand correctly, she has written and trained her own neural network, tested it against commercial products, and packaged the method for use by other diagnosticians. So you've got at least a little biology, some programming and IT stuff, and some salesmanship.

So it is a diagnostic tool, not anything like a cure.
 
2012-08-22 03:46:08 AM

RatMaster999: I May Be Crazy But...: Banned on the Run: It was my understanding there would be no math.

[youbentmywookie.com image 700x438]

Only need to count to two!

[i45.tinypic.com image 453x589]

[www.wizards.com image 200x327]

Two. Not more than two...


Bupu lives!!
 
2012-08-22 09:06:17 AM
"You didnt do that"
 
2012-08-22 11:40:51 AM
FTA: Brittany Wenger: "I'm just seventeen." Her father replied, "She's too young." "I'm old enough," said Brittany.

1.bp.blogspot.com
 
Displayed 47 of 97 comments

First | « | 1 | 2 | » | Last | Show all

View Voting Results: Smartest and Funniest


This thread is archived, and closed to new comments.

Continue Farking
Submit a Link »






Report