machine learning - Low accuracy with change to TensorFlow Cifar10 example -


i trying modify network structure provided cifar10 in tensorflow. typically, added convolution layer (conv12) after first convolution layer (conv1). no matter how set filter (i tried 1x1, 3x3, 5x5) , whether using weight decay or not, having new layer decrease accuracy below 10%. equivalent random guess in cifar10 since there 10 classes.

the code structure following, don't modify other part of cifar except setting size of input image 48x48 (instead of 24x24). guess input size should not matter.

note conv12 depthwise convolution layer because want add linear layer after conv1 layer in order minimize change original code. doing expected accuracy should similar original version, decreases around 10%. (i tried normal convolution layer didn't work also.)

  tf.variable_scope('conv1') scope:     kernel1 = _variable_with_weight_decay('weights', shape=[5, 5, 3, 64],                                          stddev=1e-4, wd=0.0)     conv_1 = tf.nn.conv2d(images, kernel1, [1, 1, 1, 1], padding='same')     biases1 = _variable_on_cpu('biases', [64], tf.constant_initializer(0.0))     bias1 = tf.nn.bias_add(conv_1, biases1)     conv1 = tf.nn.relu(bias1, name=scope.name)     _activation_summary(conv1)     tf.variable_scope('conv12') scope:     kernel12 = _variable_with_weight_decay('weights', shape=[1, 1, 64, 1],                                          stddev=1e-4, wd=0.0)     #conv_12 = tf.nn.conv2d(conv1, kernel12, [1, 1, 1, 1], padding='same')     conv_12 = tf.nn.depthwise_conv2d(conv1, kernel12, [1, 1, 1, 1], padding='same')     biases12 = _variable_on_cpu('biases', [64], tf.constant_initializer(0.0))     bias12 = tf.nn.bias_add(conv_12, biases12)             conv12 = tf.nn.relu(bias12)     _activation_summary(conv12)    pool1 = tf.nn.max_pool(conv12, ksize=[1, 3, 3, 1], strides=[1, 2, 2, 1],                      padding='same', name='pool1') .....

could please tell me wrong code?

your second convolution:

kernel12 = _variable_with_weight_decay('weights', shape=[1, 1, 64, 1]

is taking depth-64 output of previous layer , squeezing down depth-1 output. doesn't seem match whichever code have following (if it's conv2 cifar example tensorflow, isn't going work well, because 1 expects depth-64 input.)

perhaps wanted shape=[1, 1, 64, 64], add "inception-style" 1x1 convolutional layer model?


Comments

Post a Comment

Popular posts from this blog

javascript - jQuery: Add class depending on URL in the best way -

i have html progress meter: <div class="col-md-3" style="margin-left: -20px;"> <div class="progress-pos active" id="progess-1"> <div class="progress-pos-inner"> login </div> </div> </div> <div class="col-md-3 progress-pos-container"> <div class="progress-pos" id="progess-2"> <div class="progress-pos-inner"> scan items </div> </div> </div> <div class="col-md-3 progress-pos-container"> <div class="progress-pos" id="progess-3"> <div class="progress-pos-inner"> confirm address </div> </div> </div> <div class="col-md-3 progress-pos-container"> <div class="progress-pos" id="progess-4"> <div class="progress-pos-in...
Read more

caching - How to check if a url path exists in the service worker cache -

i need check if particular url path exists in service worker cache. example, suppose url is: /myserviceworker/service?a=110&b=70 this url exists in cache, there many of them different values of , b. now, suppose want refresh of these urls, how can that? i want know how access key values service worker cache. if know key, plan follows: var url = new url(key); if(url.pathname === "\/myserviceworker/service") refetch key but not sure how cache key , in format is. mean, string or url? cache api has match() method returns promise resolving in response object if match or undefined if no match exists. second parameter object can specify ignoresearch not take account url parameters. the ignoresearch option supported firefox ( chrome status here ). in other hand, retrieve cache entries, can use keys() method.
Read more

Redirect to a HTTPS version using .htaccess -

i need modify .htaccess file redirect urls https version without "www". http://example.com --> https://example.com http://www.example.com --> https://example.com https://www.example.com --> https://example.com this how .htaccess file looks like: rewriteengine on rewritecond %{request_filename} !-d rewritecond %{request_filename} !-f rewriterule ^(.*)$ redirect.php?id=$1&page=$2 [qsa,l] what modifications need make .htaccess in order forward https without "www"? it depends on doing current code. looks file called redirect trying display php page seo friendly url. think might confusing people. anyway, need , force https without www, need 1 more rule above. also second rule needs two capture groups in rewriterule test string because wanting 2 different values using $1 , $2 references. rewriteengine on #rediect www non www and/or http https --- combinations. rewritecond %{http_host} !^example\.com [nc,or] rewritecond %{https} !^...
Read more