Deep learning thread.
What deep learning related projects are you working on anon?
Sexy ladies.
>>54309496
>Teaching a neural net to rate sexy pics
Titties and hips.
I'm sorry can you repeat the question
She has a funny looking face.
http://playground.tensorflow.org
dataset: spiral
noise : 0
What's the best you can do /g/?
>>54309455
I'm helping Google identify certain images. On a daily basis. I didn't get a choice however.
>>54311948
I wonder if they'll release their dataset someday
>/g/ will never like deep learning
>>54310906
>funny looking
I think you mistyped "absolutely gorgeous"
>>54311245
Learning rate: 1
Activation: Tanh
Regularization: 0.3
Classification
4 layers: 6/6/4/2
Inputs: x1, x2, sin(x1), sin(x2)
Data: 80%
Noise: 0
Batch Size: 30
After 500 iterations:
OUTPUT
Test loss 0.025
Training loss 0.003
http://playground.tensorflow.org/#activation=tanh&batchSize=30&dataset=spiral®Dataset=reg-plane&learningRate=1®ularizationRate=0.3&noise=0&networkShape=6,6,4,2&seed=0.03362&showTestData=false&discretize=false&percTrainData=80&x=true&y=true&xTimesY=false&xSquared=false&ySquared=false&cosX=false&sinX=true&cosY=false&sinY=true&collectStats=false&problem=classification
Beat me.
>>54314069
This one reaches 0.001 on the spiral in less than 200 iterations, and beats the other datasets in less than 20 iterations.
http://playground.tensorflow.org/#activation=tanh®ularization=L1&batchSize=30&dataset=spiral®Dataset=reg-plane&learningRate=1®ularizationRate=0.01&noise=0&networkShape=6,6,4,2&seed=0.63205&showTestData=false&discretize=false&percTrainData=80&x=true&y=true&xTimesY=false&xSquared=false&ySquared=false&cosX=false&sinX=true&cosY=false&sinY=true&collectStats=false&problem=classification
>>54314155
>Iterations 000,200
>Training loss 0.016
dude
>>54314052
Some of these kids aren't used to seeing beautiful girls with noses on their face.