Styjun

What are modes in real world?

Coworkers accusing me of "cheating" for working more efficiently

A pencil in a beaker of water

Avoid showing cancel button on dialog

LeetCode 65: Valid Number (Python)

Why is Robin Hood French in Shrek?

Did the Allies reverse the threads on secret microfilm-hiding buttons to thwart the Germans?

Why does hashing public keys not actually provide any quantum resistance?

What can I use for input conversion instead of scanf?

As tourist in China do I have to fear consequences for having publicly liked South Park?

What could an alternative human-powered transportation method look like?

Python Printable Characters

How do journals gain and lose reputation?

How do the other crew members know the xenomorph is "big", if they haven't seen it?

Fiducial placement

What do I get by paying more for a bicycle?

Would there be a difference between boiling whole black peppercorns or fine ground black pepp in a stew?

Why are one-word titles so dominant in books, film, and games?

What’s the difference between 实在 and 确实?

Company asks (more than once) if I can involve family members in project

Rational Number RNG

Is it safe to plug one travel adapter into another?

Covering an 8x8 grid with X pentominoes

Define a range using a formula

Is the calculated loss associated with all samples or not?

how to read batches in one hdf5 data file for training?mini-batch gradient descent implementation in tensorflowHow to compute a per-class parameter from a mini-batch in TensorFlow?How to update model parameters with accumulated gradients?How to use different loss function in tensorflow, DNNLinearCombinedClassifierTensorflow optimizers: loss sum vs meantensorflow RNN loss not decreasingThe loss of full connected neural network doesn't fall training with tensorflowtensorflow custom loss that are not in the form of sum of single sample errorsLoss not changing no matter the learning rate

.everyoneloves__top-leaderboard:empty,.everyoneloves__mid-leaderboard:empty,.everyoneloves__bot-mid-leaderboard:empty
margin-bottom:0;

3

I am new to deep learning and Tensorflow. I got some basic questions with the sample code below :

for j in range(no_of_training_sample):
 ...
 ...
 _, _loss = sess.run([train_step, loss], X: x, Y: y)

1. For the value of _loss returned, is it the loss for each data sample or the sum from data sample 0 up to data sample j ?




2. When will the parameter w, h and b be updated, after each sess.run()?




3. How can I change the code to turn it into mini-batch ?

I tried to search the internet, I but cannot get quick answers from it.

deep-learning tensorflow

share|improve this question

edited Mar 28 at 21:17

nbro

6,3601010 gold badges6060 silver badges106106 bronze badges

asked Mar 28 at 13:14

MluiMlui

5411 bronze badge

migrated from ai.stackexchange.com Mar 28 at 20:55

This question came from our site for people interested in conceptual questions about life and challenges in a world where "cognitive" functions can be mimicked in purely digital environment.

add a comment
 | 

3

I am new to deep learning and Tensorflow. I got some basic questions with the sample code below :

for j in range(no_of_training_sample):
 ...
 ...
 _, _loss = sess.run([train_step, loss], X: x, Y: y)

1. For the value of _loss returned, is it the loss for each data sample or the sum from data sample 0 up to data sample j ?




2. When will the parameter w, h and b be updated, after each sess.run()?




3. How can I change the code to turn it into mini-batch ?

I tried to search the internet, I but cannot get quick answers from it.

deep-learning tensorflow

share|improve this question

edited Mar 28 at 21:17

nbro

6,3601010 gold badges6060 silver badges106106 bronze badges

asked Mar 28 at 13:14

MluiMlui

5411 bronze badge

migrated from ai.stackexchange.com Mar 28 at 20:55

This question came from our site for people interested in conceptual questions about life and challenges in a world where "cognitive" functions can be mimicked in purely digital environment.

add a comment
 | 

I am new to deep learning and Tensorflow. I got some basic questions with the sample code below :

for j in range(no_of_training_sample):
 ...
 ...
 _, _loss = sess.run([train_step, loss], X: x, Y: y)

1. For the value of _loss returned, is it the loss for each data sample or the sum from data sample 0 up to data sample j ?




2. When will the parameter w, h and b be updated, after each sess.run()?




3. How can I change the code to turn it into mini-batch ?

I tried to search the internet, I but cannot get quick answers from it.

deep-learning tensorflow

share|improve this question

edited Mar 28 at 21:17

nbro

6,3601010 gold badges6060 silver badges106106 bronze badges

asked Mar 28 at 13:14

MluiMlui

5411 bronze badge

for j in range(no_of_training_sample):
 ...
 ...
 _, _loss = sess.run([train_step, loss], X: x, Y: y)

share|improve this question

edited Mar 28 at 21:17

nbro

6,3601010 gold badges6060 silver badges106106 bronze badges

asked Mar 28 at 13:14

MluiMlui

5411 bronze badge

share|improve this question

edited Mar 28 at 21:17

nbro

6,3601010 gold badges6060 silver badges106106 bronze badges

asked Mar 28 at 13:14

MluiMlui

5411 bronze badge

edited Mar 28 at 21:17

nbro

6,3601010 gold badges6060 silver badges106106 bronze badges

edited Mar 28 at 21:17

nbro

6,3601010 gold badges6060 silver badges106106 bronze badges

asked Mar 28 at 13:14

MluiMlui

5411 bronze badge

asked Mar 28 at 13:14

MluiMlui

5411 bronze badge

1 Answer
1

active

oldest

votes

0

_loss is the returned value of loss (the second element in the list that you pass as first argument to the run function), after having executed a step of the computation graph (TF is based on "static" executions of computation graphs which represent the operations that need to be run).

The loss can be defined in different ways (e.g. you can define it as the cross entropy between the predicted values and the target values), so the result you will obtain in _loss will depend not only on the data you pass to the session as second argument (in your case X: x, Y: y), but also on the way you compute the loss.

The parameters of your network are updated when, in the computation graph, e.g., the function minimize is called.

To change the code and make it use mini-batches (rather than the full dataset at once), instead of passing x and y to the placeholders X and Y in the expression X: x, Y: y (when you call , _loss = sess.run([train_step, loss], X: x, Y: y)), you will have to create an "iterator" which gives you a subset of x (and the corresponding y) that you can pass to X (and Y respectively). You can actually use one of the new TF APIs which facilitate this task (instead of creating the iterator from scratch).

(If you had just asked one question, my answer could have been more detailed).

share|improve this answer

answered Mar 28 at 21:39

nbronbro

6,3601010 gold badges6060 silver badges106106 bronze badges

add a comment
 | 

Your Answer

StackExchange.ifUsing("editor", function ()
StackExchange.using("externalEditor", function ()
StackExchange.using("snippets", function ()
StackExchange.snippets.init();
);
);
, "code-snippets");

StackExchange.ready(function()
var channelOptions =
tags: "".split(" "),
id: "1"
;
initTagRenderer("".split(" "), "".split(" "), channelOptions);

StackExchange.using("externalEditor", function()
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled)
StackExchange.using("snippets", function()
createEditor();
);

else
createEditor();

);

function createEditor()
StackExchange.prepareEditor(
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: true,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: 10,
bindNavPrevention: true,
postfix: "",
imageUploader:
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/4.0/"u003ecc by-sa 4.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
,
onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
);



);

draft saved

draft discarded

Sign up or log in

StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);

Sign up using Google

Sign up using Facebook

Sign up using Email and Password

Post as a guest

Name

Email

Required, but never shown

StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f55406724%2fis-the-calculated-loss-associated-with-all-samples-or-not%23new-answer', 'question_page');

);

Post as a guest

Name

Email

Required, but never shown

0

_loss is the returned value of loss (the second element in the list that you pass as first argument to the run function), after having executed a step of the computation graph (TF is based on "static" executions of computation graphs which represent the operations that need to be run).

The loss can be defined in different ways (e.g. you can define it as the cross entropy between the predicted values and the target values), so the result you will obtain in _loss will depend not only on the data you pass to the session as second argument (in your case X: x, Y: y), but also on the way you compute the loss.

The parameters of your network are updated when, in the computation graph, e.g., the function minimize is called.

To change the code and make it use mini-batches (rather than the full dataset at once), instead of passing x and y to the placeholders X and Y in the expression X: x, Y: y (when you call , _loss = sess.run([train_step, loss], X: x, Y: y)), you will have to create an "iterator" which gives you a subset of x (and the corresponding y) that you can pass to X (and Y respectively). You can actually use one of the new TF APIs which facilitate this task (instead of creating the iterator from scratch).

(If you had just asked one question, my answer could have been more detailed).

share|improve this answer

answered Mar 28 at 21:39

nbronbro

6,3601010 gold badges6060 silver badges106106 bronze badges

add a comment
 | 

0

_loss is the returned value of loss (the second element in the list that you pass as first argument to the run function), after having executed a step of the computation graph (TF is based on "static" executions of computation graphs which represent the operations that need to be run).

The loss can be defined in different ways (e.g. you can define it as the cross entropy between the predicted values and the target values), so the result you will obtain in _loss will depend not only on the data you pass to the session as second argument (in your case X: x, Y: y), but also on the way you compute the loss.

The parameters of your network are updated when, in the computation graph, e.g., the function minimize is called.

To change the code and make it use mini-batches (rather than the full dataset at once), instead of passing x and y to the placeholders X and Y in the expression X: x, Y: y (when you call , _loss = sess.run([train_step, loss], X: x, Y: y)), you will have to create an "iterator" which gives you a subset of x (and the corresponding y) that you can pass to X (and Y respectively). You can actually use one of the new TF APIs which facilitate this task (instead of creating the iterator from scratch).

(If you had just asked one question, my answer could have been more detailed).

share|improve this answer

answered Mar 28 at 21:39

nbronbro

6,3601010 gold badges6060 silver badges106106 bronze badges

add a comment
 | 

_loss is the returned value of loss (the second element in the list that you pass as first argument to the run function), after having executed a step of the computation graph (TF is based on "static" executions of computation graphs which represent the operations that need to be run).

The loss can be defined in different ways (e.g. you can define it as the cross entropy between the predicted values and the target values), so the result you will obtain in _loss will depend not only on the data you pass to the session as second argument (in your case X: x, Y: y), but also on the way you compute the loss.

The parameters of your network are updated when, in the computation graph, e.g., the function minimize is called.

To change the code and make it use mini-batches (rather than the full dataset at once), instead of passing x and y to the placeholders X and Y in the expression X: x, Y: y (when you call , _loss = sess.run([train_step, loss], X: x, Y: y)), you will have to create an "iterator" which gives you a subset of x (and the corresponding y) that you can pass to X (and Y respectively). You can actually use one of the new TF APIs which facilitate this task (instead of creating the iterator from scratch).

(If you had just asked one question, my answer could have been more detailed).

share|improve this answer

answered Mar 28 at 21:39

nbronbro

6,3601010 gold badges6060 silver badges106106 bronze badges

share|improve this answer

answered Mar 28 at 21:39

nbronbro

6,3601010 gold badges6060 silver badges106106 bronze badges

answered Mar 28 at 21:39

nbronbro

6,3601010 gold badges6060 silver badges106106 bronze badges

answered Mar 28 at 21:39

nbronbro

6,3601010 gold badges6060 silver badges106106 bronze badges