Dan Ryan - user overview - MetaOptimize Q+A

Dan Ryan

406

karma

Registered user
real name	Dan Ryan
member since	Feb 11 '13 at 17:32
last seen	Apr 04 '14 at 12:26
location	United States

I have a Ph.D. in mathematics and am currently a post-doctoral researcher at the National Institute for Mathematical and Biological Synthesis (NIMBioS) at the University of Tennessee.

overview recent activity karma history favorites

7 Questions

votes

answers

956

views

A good linux (Ubuntu) program for making drawings of neural network diagrams for publications

May 01 '14 at 11:49 Jose Tagarro1 (asked by Dan Ryan406●7●11●16 )

diagrams

votes

answers

820

views

Stopping and evaluation criteria for Autoencoders that will be used for supervised learning

Nov 08 '13 at 16:22 Dan Ryan406●7●11●16

votes

answers

585

views

Early Stopping with penalty terms

Oct 27 '13 at 20:35 Ben Gimpert61●4 (asked by Dan Ryan406●7●11●16 )

early-stopping regularization neural-networks

votes

answers

757

views

Initializing a dropout net with a deep Boltzmann mahcine

Apr 17 '13 at 15:50 Dan Ryan406●7●11●16

dropout dbm

votes

answers

views

Using multiple GPUs to train RBM/DBM models

Mar 12 '13 at 22:21 Rakesh Chalasani264●1●2●10 (asked by Dan Ryan406●7●11●16 )

gpu rbm dbm

votes

answers

683

views

Generative fine-tuning and partition function estimation of a DBM with missing data

Feb 22 '13 at 14:48 Dan Ryan406●7●11●16

rbm dbm fine-tuning ais

votes

answers

views

Mean field activation of NReLU units

Feb 11 '13 at 17:32 Dan Ryan406●7●11●16

nrelu rbm dbm mean-field

21 Answers

Activation function that depends on whole layer?

Dropout implementation in Convolutional Architecture

maxout network + autoencoder?

Estimating confidence of a prediction

Deep learning with hand-engineered features

removal of irrelevant features

convolutional network code

Training for Convolutional Neural Network

How to derive an objective function for estimating multinomial parameters

What types of unsupervised feature extraction are out there?

How does one reverse the max-pooling step for a convolutional auto-encoder ?

Mini-batches too big for memory (Theano)

Dropout: how to scale up weights after pretraining when the related 2 layers has different dropout probability?

best cuda machine learning library

Weight updates for contrastive divergence learned RBM with gauss visible units

Can I estimate the Hessian or curvature of an unknown (almost) quadratic function without calculating any gradients?

Value of deep -vs- shallow NNs

Deep neural network with unbalanced classes

Energy function of an RBM with noisy rectified linear units (NReLUs)

Using Visible Rectified Linear Units with RBM

5 Votes

convolutional × 2
convolution × 2
theano × 2
datamining × 2
diagrams × 2
priors × 2
nrelu × 2
cnn × 2
feature-extraction × 2
crbm × 2

gpu × 2
nn × 2
step-size × 1
rectified-linear-units × 1
maxout × 1
rectified × 1
regression × 1
gaussian × 1
divisive × 1
pretraining × 1

ais × 1
regularization × 1
restricted × 1
feature-selection × 1
memory × 1
prediction × 1
deep-belief-networks × 1
boltzmann × 1
energy × 1
early-stopping × 1

linear-regression × 1
scaling × 1
probability × 1
confidence × 1
contrastive-divergence × 1
fine-tuning × 1
linear × 1
sgd × 1
units × 1
normalization × 1

11 Kudos

● Prominent Question × 7
● Well-Known Question × 7
● Conspicuous Question × 7
● Appreciated Answer × 5
● Cognoscenti × 2
● Revisionist × 1

● Advocate × 1
● Blue Pencil × 1
● Mentor × 1
● Esteemed Answer × 1
● Guide × 1

Registered user

7 Questions

A good linux (Ubuntu) program for making drawings of neural network diagrams for publications

Stopping and evaluation criteria for Autoencoders that will be used for supervised learning

Early Stopping with penalty terms

Initializing a dropout net with a deep Boltzmann mahcine

Using multiple GPUs to train RBM/DBM models

Generative fine-tuning and partition function estimation of a DBM with missing data

Mean field activation of NReLU units

21 Answers

5 Votes

50 Tags

11 Kudos