Trjtdtk

The IT department bottlenecks progress. How should I handle this?

Is there a working SACD iso player for Ubuntu?

Open a doc from terminal, but not by its name

Which one is correct as adjective “protruding” or “protruded”?

Has any country ever had 2 former presidents in jail simultaneously?

How to implement a feedback to keep the DC gain at zero for this conceptual passive filter?

What prevents the use of a multi-segment ILS for non-straight approaches?

When a Cleric spontaneously casts a Cure Light Wounds spell, will a Pearl of Power recover the original spell or Cure Light Wounds?

On a tidally locked planet, would time be quantized?

What does "landing" mean in this context?

Count the occurrence of each unique word in the file

What is Cash Advance APR?

What is this called? Old film camera viewer?

Can I sign legal documents with a smiley face?

Loading commands from file

Closed-form expression for certain product

Problem with TransformedDistribution

Is it improper etiquette to ask your opponent what his/her rating is before the game?

Travelling outside the UK without a passport

Why did the EU agree to delay the Brexit deadline?

Longest common substring in linear time

Writing bit difficult equation in latex

When were female captains banned from Starfleet?

Is there a single word describing earning money through any means?

What activation function should I use for a specific regression problem?

How flexible is the link between objective function and output layer activation function?Technology stack for linear regression on (not so) large datasetParameterization regression of rotation angleAlternatives to linear activation function in regression tasks to limit the outputWhat is the Time Complexity of Linear Regression?Activation function vs Squashing functionUsing Keras to Predict a Function Following a Normal Distributionbest activation function for ensemble?Regression with -1,1 target range - Should we use a tanh activation in the last 1 unit dense layer?Images Score Regression only regresses to the average of the target values

2

$begingroup$

Which is better for regression problems create a neural net with tanh/sigmoid and exp(like) activations or ReLU and linear? Standard is to use ReLU but it's brute force solution that requires certain net size and I would like to avoid creating a very big net, also sigmoid is much more prefered but in my case regression will output values from range (0, 1e7)... maybe also sigmoid net with linear head will work? I am curious about your take on the subject.

machine-learning neural-network deep-learning regression activation-function

share|improve this question

edited 2 days ago

Media

7,38762161

asked 2 days ago

questerquester

363

$endgroup$

add a comment | 

2

$begingroup$

Which is better for regression problems create a neural net with tanh/sigmoid and exp(like) activations or ReLU and linear? Standard is to use ReLU but it's brute force solution that requires certain net size and I would like to avoid creating a very big net, also sigmoid is much more prefered but in my case regression will output values from range (0, 1e7)... maybe also sigmoid net with linear head will work? I am curious about your take on the subject.

machine-learning neural-network deep-learning regression activation-function

share|improve this question

edited 2 days ago

Media

7,38762161

asked 2 days ago

questerquester

363

$endgroup$

add a comment | 

$begingroup$

Which is better for regression problems create a neural net with tanh/sigmoid and exp(like) activations or ReLU and linear? Standard is to use ReLU but it's brute force solution that requires certain net size and I would like to avoid creating a very big net, also sigmoid is much more prefered but in my case regression will output values from range (0, 1e7)... maybe also sigmoid net with linear head will work? I am curious about your take on the subject.

machine-learning neural-network deep-learning regression activation-function

share|improve this question

edited 2 days ago

Media

7,38762161

asked 2 days ago

questerquester

363

$endgroup$

share|improve this question

edited 2 days ago

Media

7,38762161

asked 2 days ago

questerquester

363

share|improve this question

edited 2 days ago

Media

7,38762161

asked 2 days ago

questerquester

363

edited 2 days ago

Media

7,38762161

edited 2 days ago

Media

7,38762161

asked 2 days ago

questerquester

363

asked 2 days ago

questerquester

363

2 Answers
2

active

oldest

votes

2

$begingroup$

The issue with sigmoid and tanh activations is that their gradients saturate for extreme values of their arguments. This may occur if you do not normalize your inputs. If the learned weights of the unit are such that the gradient of its activation is close to zero, it will take longer for any updates to be reflected in the unit's weights. A final layer with no non-linearity will help you scale up your hidden layers' outputs.

In the end, the performance is application specific. You should try out both kinds of activations on a subset of your data and see which performs better.

share|improve this answer

answered 2 days ago

hazrmardhazrmard

1364

$endgroup$

add a comment | 

1

$begingroup$

There are two points that have to be considered.

1. Take care of the output of your network. If that's a Real number and can take any value, you have to use linear activation as the output.


2. The inner activations highly depend on your task and the size of the network that you use. What I'm going to tell you is based on experience. If you don't have a very deep network, $tanh$ and $ReLU$ may not differ very much in convergence time. If you're using very deep networks, don't use $tahn$ at all. $ReLU$ is also not recommended in some contexts. You can employ $PReLU$ in very deep networks. It does not add too many parameters to learn. You can also use $leaky-ReLU$ in order to avoid dying ReLU problem which may occur.

Finally, about the other nonlinearity that you've referred; try not to use $Sigmoid$ due to that fact that it's expected value is not equal to zero but half. It's a bit statistical stuff, but you can consider it's roughly hard for a network to learn shifted weights.

share|improve this answer

edited 2 days ago

answered 2 days ago

MediaMedia

7,38762161

$endgroup$

add a comment | 

Your Answer

StackExchange.ifUsing("editor", function ()
return StackExchange.using("mathjaxEditing", function ()
StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix)
StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["$", "$"], ["\\(","\\)"]]);
);
);
, "mathjax-editing");

StackExchange.ready(function()
var channelOptions =
tags: "".split(" "),
id: "557"
;
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
,
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%2fdatascience.stackexchange.com%2fquestions%2f47751%2fwhat-activation-function-should-i-use-for-a-specific-regression-problem%23new-answer', 'question_page');

);

Post as a guest

Name

Email

Required, but never shown

2

$begingroup$

The issue with sigmoid and tanh activations is that their gradients saturate for extreme values of their arguments. This may occur if you do not normalize your inputs. If the learned weights of the unit are such that the gradient of its activation is close to zero, it will take longer for any updates to be reflected in the unit's weights. A final layer with no non-linearity will help you scale up your hidden layers' outputs.

In the end, the performance is application specific. You should try out both kinds of activations on a subset of your data and see which performs better.

share|improve this answer

answered 2 days ago

hazrmardhazrmard

1364

$endgroup$

add a comment | 

2

$begingroup$

The issue with sigmoid and tanh activations is that their gradients saturate for extreme values of their arguments. This may occur if you do not normalize your inputs. If the learned weights of the unit are such that the gradient of its activation is close to zero, it will take longer for any updates to be reflected in the unit's weights. A final layer with no non-linearity will help you scale up your hidden layers' outputs.

In the end, the performance is application specific. You should try out both kinds of activations on a subset of your data and see which performs better.

share|improve this answer

answered 2 days ago

hazrmardhazrmard

1364

$endgroup$

add a comment | 

$begingroup$

The issue with sigmoid and tanh activations is that their gradients saturate for extreme values of their arguments. This may occur if you do not normalize your inputs. If the learned weights of the unit are such that the gradient of its activation is close to zero, it will take longer for any updates to be reflected in the unit's weights. A final layer with no non-linearity will help you scale up your hidden layers' outputs.

In the end, the performance is application specific. You should try out both kinds of activations on a subset of your data and see which performs better.

share|improve this answer

answered 2 days ago

hazrmardhazrmard

1364

$endgroup$

share|improve this answer

answered 2 days ago

hazrmardhazrmard

1364

answered 2 days ago

hazrmardhazrmard

1364

answered 2 days ago

hazrmardhazrmard

1364

1

$begingroup$

There are two points that have to be considered.

1. Take care of the output of your network. If that's a Real number and can take any value, you have to use linear activation as the output.


2. The inner activations highly depend on your task and the size of the network that you use. What I'm going to tell you is based on experience. If you don't have a very deep network, $tanh$ and $ReLU$ may not differ very much in convergence time. If you're using very deep networks, don't use $tahn$ at all. $ReLU$ is also not recommended in some contexts. You can employ $PReLU$ in very deep networks. It does not add too many parameters to learn. You can also use $leaky-ReLU$ in order to avoid dying ReLU problem which may occur.

Finally, about the other nonlinearity that you've referred; try not to use $Sigmoid$ due to that fact that it's expected value is not equal to zero but half. It's a bit statistical stuff, but you can consider it's roughly hard for a network to learn shifted weights.

share|improve this answer

edited 2 days ago

answered 2 days ago

MediaMedia

7,38762161

$endgroup$

add a comment | 

1

$begingroup$

There are two points that have to be considered.

1. Take care of the output of your network. If that's a Real number and can take any value, you have to use linear activation as the output.


2. The inner activations highly depend on your task and the size of the network that you use. What I'm going to tell you is based on experience. If you don't have a very deep network, $tanh$ and $ReLU$ may not differ very much in convergence time. If you're using very deep networks, don't use $tahn$ at all. $ReLU$ is also not recommended in some contexts. You can employ $PReLU$ in very deep networks. It does not add too many parameters to learn. You can also use $leaky-ReLU$ in order to avoid dying ReLU problem which may occur.

Finally, about the other nonlinearity that you've referred; try not to use $Sigmoid$ due to that fact that it's expected value is not equal to zero but half. It's a bit statistical stuff, but you can consider it's roughly hard for a network to learn shifted weights.

share|improve this answer

edited 2 days ago

answered 2 days ago

MediaMedia

7,38762161

$endgroup$

add a comment | 

$begingroup$

There are two points that have to be considered.

1. Take care of the output of your network. If that's a Real number and can take any value, you have to use linear activation as the output.


2. The inner activations highly depend on your task and the size of the network that you use. What I'm going to tell you is based on experience. If you don't have a very deep network, $tanh$ and $ReLU$ may not differ very much in convergence time. If you're using very deep networks, don't use $tahn$ at all. $ReLU$ is also not recommended in some contexts. You can employ $PReLU$ in very deep networks. It does not add too many parameters to learn. You can also use $leaky-ReLU$ in order to avoid dying ReLU problem which may occur.

Finally, about the other nonlinearity that you've referred; try not to use $Sigmoid$ due to that fact that it's expected value is not equal to zero but half. It's a bit statistical stuff, but you can consider it's roughly hard for a network to learn shifted weights.

share|improve this answer

edited 2 days ago

answered 2 days ago

MediaMedia

7,38762161

$endgroup$

share|improve this answer

edited 2 days ago

answered 2 days ago

MediaMedia

7,38762161

answered 2 days ago

MediaMedia

7,38762161

answered 2 days ago

MediaMedia

7,38762161