machine-learning


What is happen when the number of file smaller than batch in HDF5


I have a hdf5 layer that read the information from the list.txt as
layer {
name: "data"
type: "HDF5Data"
top: "data"
top: "label"
include {
phase: TEST
}
hdf5_data_param {
source: "./list.txt"
batch_size: 4
shuffle: true
}
}
where list.txt contains two path files
/home/user/file1.h5
/home/user/file2.h5
while the batch size is 4. What is happen with above code? Can the data choose 4 files to feed the network?
You have two hdf5 files, but each file may contain more than a single training example. Thus, effectively, you may have far more than batch_size: 4 examples.
Caffe does not really cares about the actual number of training examples: when it finishes to process all the examples (aka "epoch") it simply starts over reading the samples again. Caffe cycles through all the samples until number of training/testing iteration is reached.

Related Links

How does it work Q-learning+NN
What is the best approach to detect Spatio-Temporal anomalies in a Dynamic Bayesian Network?
Why cost function of linear regression is bowl shaped?
Neural Network outputs the same classifier for all inputs
Is it possible to create a garden that sustains itself using Machine learning?
How to give a constant input to keras
Error: import graphlab
Why Element of Statistical Learning [Hastie, et al] 2e gives a different kind of explaination of Model Stacking
Why is the gradient descent of logistic regression like this?
Decision boundary calculation in SVM
What should be the “learning multiple” of the up-sampling layers in the Fully Convolutional Network?
Monte Carlo RL - why is the computational expense of estimating the value of a single state is independent of the number of states?
Printing from Custom Layers in Keras
Create identification-verification combined train theano
How to show epoch time from caffe log file?
Does inference time depend on number of parameter? [closed]

Categories

HOME
google-analytics
loops
instagram
maven-lifecycle
etcd
webserver
windows-forms-designer
contiki
ais
opentsdb
purescript
memcached
zk
ninja
row-number
jlabel
amazon-route53
google-fusion-tables
web-component
odbc
react-boilerplate
libcurl
azure-storage-blobs
xlwings
cider
packet-capture
preview
ctype
vowpalwabbit
uilabel
vim-syntax-highlighting
neoscms
jung
opensaml
webex
dreamhost
azure-sql-server
gold-parser
yield
getelementsbyclassname
react-d3
xquery-sql
virsh
run-configuration
subtitle
fluentbootstrap
heyzap
lab-color-space
zoo
on-screen-keyboard
prime-factoring
angularjs-service
news-feed
mockery
swig-template
ubuntu-15.10
qtquick2
idfa
grape
irs
velocity.js
sqlsrv
busy-waiting
circular-buffer
build-system
zenoss
virtual-address-space
buttonclick
apiaxle
kendo-tooltip
panning
payflowlink
mathtype
java-service-wrapper
mysql-error-1052
grit
memory-mapped-files
struts-config
xemacs
twisted.internet
memory-profiling
mono-service
soaplite
cgpoint
cocos2d-python
incremental-linking
astoria

Resources

Encrypt Message



code
soft
python
ios
c
html
jquery
cloud
mobile