That's a bit rubbish. I'm not questioning the quality of the research; just expressing disappointment at how poor the discrimination is in this day and age. (My MSc thesis TWENTY-PLUS YEARS AGO was on measuring brain volumes.)
Can machine learning correctly predict whether someone is male or female, just with data from brain imaging? Yes, ~80% of the time. New paper in NeuroImage: https://www.sciencedirect.com/science/article/pii/S105381191830065X …pic.twitter.com/q40lgrBSyP
-
-
-
Relatively small training sample tho. I’m going to do something similar in the much bigger UK Biobank dataset.
-
pah, why faff around with shitty brain images when the question can be easily answered with a dna sample :P
-
Or just by asking them. But that’s not QUITE the point...
-
The question I asked in my MSc could have been answered much more cheaply and accurately by not wasting money on MRI machines and simply removing the subjects' brains and dissecting them.
-
Bloody ethics committees though. Amirite?
Final de la conversación
Conversación nueva -
-
-
Gotta say, it's blowing my mind a little to see "Gaussian curvature" and "mean curvature" appearing in a neuroscience paper. Do you know what they are measuring the curvature of, and why?
-
The cortex?
-
*shrug* These are examples of math that was once declared "never to be of any use in the real world", before finding use in general relativity. I love finding new places differential geometry pop up. I'd love to know more about why these curvatures are considered interesting here
Final de la conversación
Conversación nueva -
-
-
can you do using trans ppl to see if their behavior matches their identity? maybe could become an objective test for true trans genderism
Gracias. Twitter usará esto para mejorar tu cronología. DeshacerDeshacer
-
-
-
So...even a machine can tell the difference.
Gracias. Twitter usará esto para mejorar tu cronología. DeshacerDeshacer
-
-
-
how old are subjects? does paper account for possible changes in brain morphology from social conditioning ie neuroplasticity
Gracias. Twitter usará esto para mejorar tu cronología. DeshacerDeshacer
-
-
-
“The algorithm was not able to distinguish more than two sexes, male and female. Sorry, freaks, you don’t have a poo poo platter of genders to choose from. You’re either man, woman, or mentally ill.”https://heartiste.wordpress.com/2018/03/12/machine-learning-correctly-predicts-sex-from-brain-imaging-feminists-btfo-again/ …
Gracias. Twitter usará esto para mejorar tu cronología. DeshacerDeshacer
-
-
-
Only 80%?
Gracias. Twitter usará esto para mejorar tu cronología. DeshacerDeshacer
-
-
-
looks like simple volume is main predictor. What has this got to do with machine learning?
-
Teaching machines to be sexist, gotta start somewhere.
-
I guess you think even researching gender differences is sexist Mario?
-
Facts are sexist so we don't need facts.
Final de la conversación
Conversación nueva -
-
-
Releasing the raw imaging data to the ML community would produce more powerful predictions via better features and algorithms. Sample is likely mixed race. That info could boost accuracy.
Gracias. Twitter usará esto para mejorar tu cronología. DeshacerDeshacer
-
Parece que el contenido está tardando un poco en cargarse.
Puede que Twitter esté saturado o experimentando un problema momentáneo. Inténtalo de nuevo o visita el Estado de Twitter para más información.