Gsyfuy

Cold is to Refrigerator as warm is to?

Why does tar appear to skip file contents when output file is /dev/null?

How can I make names more distinctive without making them longer?

Is there a documented rationale why the House Ways and Means chairman can demand tax info?

Is above average number of years spent on PhD considered a red flag in future academia or industry positions?

Can a monk deflect thrown melee weapons?

How is simplicity better than precision and clarity in prose?

Need a suitable toxic chemical for a murder plot in my novel

3 doors, three guards, one stone

What was the last x86 CPU that did not have the x87 floating-point unit built in?

What to do with post with dry rot?

Determine whether f is a function, an injection, a surjection

Can I add database to AWS RDS MySQL without creating new instance?

What do I do if technical issues prevent me from filing my return on time?

Why does this iterative way of solving of equation work?

Estimate capacitor parameters

If I can make up priors, why can't I make up posteriors?

What computer would be fastest for Mathematica Home Edition?

Are my PIs rude or am I just being too sensitive?

Can the prologue be the backstory of your main character?

How do I automatically answer y in bash script?

Can smartphones with the same camera sensor have different image quality?

Was credit for the black hole image misattributed?

Jazz greats knew nothing of modes. Why are they used to improvise on standards?

What loss function to use when labels are probabilities?

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

2

$begingroup$

What loss function is most appropriate when training a model with target values that are probabilities? For example, I have a 3-output model with x=[some features] and y=[0.2, 0.3, 0.5].

It seems like something like cross-entropy doesn't make sense here since it assumes that a single target is the correct label.

Would something like MSE (after applying softmax) make sense, or is there a better loss function?

neural-networks loss-functions probability-distribution

share|improve this question

asked 7 hours ago

Thomas JohnsonThomas Johnson

1133

New contributor

Thomas Johnson is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.

$endgroup$

add a comment | 

2

$begingroup$

What loss function is most appropriate when training a model with target values that are probabilities? For example, I have a 3-output model with x=[some features] and y=[0.2, 0.3, 0.5].

It seems like something like cross-entropy doesn't make sense here since it assumes that a single target is the correct label.

Would something like MSE (after applying softmax) make sense, or is there a better loss function?

neural-networks loss-functions probability-distribution

share|improve this question

asked 7 hours ago

Thomas JohnsonThomas Johnson

1133

New contributor

Thomas Johnson is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.

$endgroup$

add a comment | 

$begingroup$

What loss function is most appropriate when training a model with target values that are probabilities? For example, I have a 3-output model with x=[some features] and y=[0.2, 0.3, 0.5].

It seems like something like cross-entropy doesn't make sense here since it assumes that a single target is the correct label.

Would something like MSE (after applying softmax) make sense, or is there a better loss function?

neural-networks loss-functions probability-distribution

share|improve this question

asked 7 hours ago

Thomas JohnsonThomas Johnson

1133

New contributor

Thomas Johnson is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.

$endgroup$

share|improve this question

asked 7 hours ago

Thomas JohnsonThomas Johnson

1133

New contributor

Thomas Johnson is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.

share|improve this question

asked 7 hours ago

Thomas JohnsonThomas Johnson

1133

New contributor

Thomas Johnson is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.

asked 7 hours ago

Thomas JohnsonThomas Johnson

1133

New contributor

Thomas Johnson is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.

asked 7 hours ago

Thomas JohnsonThomas Johnson

1133

1 Answer
1

active

oldest

votes

3

$begingroup$

Actually, the cross-entropy loss function would be appropriate here, since it measures the "distance" between a distribution $q$ and the "true" distribution $p$.

You are right, though, that using a loss function called "cross_entropy" in many APIs would be a mistake. This is because these functions, as you said, assume a one-hot label. You would need to use the general cross-entropy function,

$$H(p,q)=-sum_{xin X} p(x) log q(x).$$
$ $

Note that one-hot labels would mean that
$$
p(x) =
begin{cases}
1 & text{if }x text{ is the true label}\
0 & text{otherwise}
end{cases}$$

which causes the cross-entropy $H(p,q)$ to reduce to the form you're familiar with:

$$H(p,q) = -log q(x_{label})$$

share|improve this answer

answered 6 hours ago

Philip RaeisghasemPhilip Raeisghasem

988119

$endgroup$

add a comment | 

Your Answer

StackExchange.ready(function() {
var channelOptions = {
tags: "".split(" "),
id: "658"
};
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: false,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: null,
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/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
},
noCode: true, onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
});


}
});

Thomas Johnson is a new contributor. Be nice, and check out our Code of Conduct.

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%2fai.stackexchange.com%2fquestions%2f11816%2fwhat-loss-function-to-use-when-labels-are-probabilities%23new-answer', 'question_page');
}
);

Post as a guest

Name

Email

Required, but never shown

3

$begingroup$

Actually, the cross-entropy loss function would be appropriate here, since it measures the "distance" between a distribution $q$ and the "true" distribution $p$.

You are right, though, that using a loss function called "cross_entropy" in many APIs would be a mistake. This is because these functions, as you said, assume a one-hot label. You would need to use the general cross-entropy function,

$$H(p,q)=-sum_{xin X} p(x) log q(x).$$
$ $

Note that one-hot labels would mean that
$$
p(x) =
begin{cases}
1 & text{if }x text{ is the true label}\
0 & text{otherwise}
end{cases}$$

which causes the cross-entropy $H(p,q)$ to reduce to the form you're familiar with:

$$H(p,q) = -log q(x_{label})$$

share|improve this answer

answered 6 hours ago

Philip RaeisghasemPhilip Raeisghasem

988119

$endgroup$

add a comment | 

3

$begingroup$

Actually, the cross-entropy loss function would be appropriate here, since it measures the "distance" between a distribution $q$ and the "true" distribution $p$.

You are right, though, that using a loss function called "cross_entropy" in many APIs would be a mistake. This is because these functions, as you said, assume a one-hot label. You would need to use the general cross-entropy function,

$$H(p,q)=-sum_{xin X} p(x) log q(x).$$
$ $

Note that one-hot labels would mean that
$$
p(x) =
begin{cases}
1 & text{if }x text{ is the true label}\
0 & text{otherwise}
end{cases}$$

which causes the cross-entropy $H(p,q)$ to reduce to the form you're familiar with:

$$H(p,q) = -log q(x_{label})$$

share|improve this answer

answered 6 hours ago

Philip RaeisghasemPhilip Raeisghasem

988119

$endgroup$

add a comment | 

$begingroup$

Actually, the cross-entropy loss function would be appropriate here, since it measures the "distance" between a distribution $q$ and the "true" distribution $p$.

You are right, though, that using a loss function called "cross_entropy" in many APIs would be a mistake. This is because these functions, as you said, assume a one-hot label. You would need to use the general cross-entropy function,

$$H(p,q)=-sum_{xin X} p(x) log q(x).$$
$ $

Note that one-hot labels would mean that
$$
p(x) =
begin{cases}
1 & text{if }x text{ is the true label}\
0 & text{otherwise}
end{cases}$$

which causes the cross-entropy $H(p,q)$ to reduce to the form you're familiar with:

$$H(p,q) = -log q(x_{label})$$

share|improve this answer

answered 6 hours ago

Philip RaeisghasemPhilip Raeisghasem

988119

$endgroup$

share|improve this answer

answered 6 hours ago

Philip RaeisghasemPhilip Raeisghasem

988119

answered 6 hours ago

Philip RaeisghasemPhilip Raeisghasem

988119

answered 6 hours ago

Philip RaeisghasemPhilip Raeisghasem

988119