Ygthkb

Do the speed limit reductions due to pollution also apply to electric cars in France?

How to know if I am a 'Real Developer'

What happened to Hermione’s clothing and other possessions after the memory wipe in Deathly Hallows?

Other than edits for international editions, did Harry Potter and the Philosopher's Stone receive errata?

How to get a 2D Plot from a 3D Listplot?

Does rolling friction increase speed of a wheel?

How can I deduce the power of a capacitor from its datasheet?

What can I do to encourage my players to use their consumables?

Lubuntu 18.10 File Manager: How to view directory tree structure?

what is the required document in Agile

How do I avoid the "chosen hero" feeling?

Is it possible to detect 100% of SQLi with a simple regex?

How do I add a strong "onion flavor" to the biryani (in restaurant style)?

Was there a pre-determined arrangement for the division of Germany in case it surrendered before any Soviet forces entered its territory?

Renting a 2CV in France

Why is Shelob considered evil?

Solving the linear first order differential equaition?

What would be some possible ways of escaping higher gravity planets?

Taking an academic pseudonym?

Crack the bank account's password!

How to draw a node with two options using TikZ graphs in LaTeX

What is an explicit bijection in combinatorics?

I have trouble understanding this fallacy: "If A, then B. Therefore if not-B, then not-A."

Explicit Riemann Hilbert correspondence

Why do neural networks need so many examples to perform?

Examples of drastic improvements when using deep neural networksDoes Neural Networks based classification need a dimension reduction?Do Neural Networks need “compound” features?Overtraining Neural Networks?Why use gradient descent with neural networks?To what extent are convolutional neural networks inspired by biology?Deep networks vs shallow networks: why do we need depth?One big neural network or many small neural networks?Why do neural networks need feature selection / engineering?Deep neural networks versus tall neural networks

1

$begingroup$

So we are talking artificial intelligence.

I think human child at 2 needs like 5 instances of a car to be able to identify it with reasonable accuracy regardless of color, make, etc. When my son was 2, he was able to identify trams and trains, even though he had seen just few. He was usually confusing one with each other, apparently his NN was not trained enough, but still.

To paraphrase myself, I could ask what ANN’s miss to be able to learn way quicker? Is transfer learning an answer?

neural-networks neuroscience

share|cite|improve this question

asked 4 hours ago

MarcinMarcin

1717

$endgroup$

add a comment | 

1

$begingroup$

So we are talking artificial intelligence.

I think human child at 2 needs like 5 instances of a car to be able to identify it with reasonable accuracy regardless of color, make, etc. When my son was 2, he was able to identify trams and trains, even though he had seen just few. He was usually confusing one with each other, apparently his NN was not trained enough, but still.

To paraphrase myself, I could ask what ANN’s miss to be able to learn way quicker? Is transfer learning an answer?

neural-networks neuroscience

share|cite|improve this question

asked 4 hours ago

MarcinMarcin

1717

$endgroup$

add a comment | 

$begingroup$

So we are talking artificial intelligence.

I think human child at 2 needs like 5 instances of a car to be able to identify it with reasonable accuracy regardless of color, make, etc. When my son was 2, he was able to identify trams and trains, even though he had seen just few. He was usually confusing one with each other, apparently his NN was not trained enough, but still.

To paraphrase myself, I could ask what ANN’s miss to be able to learn way quicker? Is transfer learning an answer?

neural-networks neuroscience

share|cite|improve this question

asked 4 hours ago

MarcinMarcin

1717

$endgroup$

share|cite|improve this question

asked 4 hours ago

MarcinMarcin

1717

share|cite|improve this question

asked 4 hours ago

MarcinMarcin

1717

asked 4 hours ago

MarcinMarcin

1717

asked 4 hours ago

MarcinMarcin

1717

3 Answers
3

active

oldest

votes

2

$begingroup$

First of all, at age two, child knows a lot about world and actively applies this knowledge. Child does a lot of "transfer learning" by applying this knowledge to new concepts.

Second, before seeing those five "labeled" examples of car, child sees a lot of cars on the street, on TV, toy cars etc., so also a lot of "unsupervised learning" happens beforehand.

Finally, neural networks have almost nothing in common with human brain, so there's not much point in comparing them. Also notice that there are algorithms for one shot learning, and pretty much research on it currently happens.

share|cite|improve this answer

answered 3 hours ago

Tim♦Tim

58k9127219

$endgroup$

add a comment | 

2

$begingroup$

There's a kind of goalpost moving that underlies this question. It used to be that NNs weren't very good at image recognition, so no one compared them to humans. Now that NNs are good at image tasks, suddenly it's the fault of NNs that they require a lot of training data to be comparable to children.

You can also turn this logic on its head. Suppose a child sees a number of cars the day that it's born. I wouldn't expect the child to be able to pick out a car the next day, or the next week even though it's seen so many examples. Why are newborns so slow to learn? Because it takes a lot of exposure to the real world and time to change the child's neural pathways ("training data").

share|cite|improve this answer

answered 2 hours ago

SycoraxSycorax

40.4k12103203

$endgroup$

add a comment | 

0

$begingroup$

Please remember that neural networks are very complex structures with lots of parametres to tune, therefore you should not expect it to "learn" without huge numbers of examples.

This is the reason why simpler models are often preferred when samples are not huge.

share|cite|improve this answer

answered 1 hour ago

asdfasdf

413

New contributor

asdf 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 | 

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: "65"
};
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%2fstats.stackexchange.com%2fquestions%2f394118%2fwhy-do-neural-networks-need-so-many-examples-to-perform%23new-answer', 'question_page');
}
);

Post as a guest

Name

Email

Required, but never shown

2

$begingroup$

First of all, at age two, child knows a lot about world and actively applies this knowledge. Child does a lot of "transfer learning" by applying this knowledge to new concepts.

Second, before seeing those five "labeled" examples of car, child sees a lot of cars on the street, on TV, toy cars etc., so also a lot of "unsupervised learning" happens beforehand.

Finally, neural networks have almost nothing in common with human brain, so there's not much point in comparing them. Also notice that there are algorithms for one shot learning, and pretty much research on it currently happens.

share|cite|improve this answer

answered 3 hours ago

Tim♦Tim

58k9127219

$endgroup$

add a comment | 

2

$begingroup$

First of all, at age two, child knows a lot about world and actively applies this knowledge. Child does a lot of "transfer learning" by applying this knowledge to new concepts.

Second, before seeing those five "labeled" examples of car, child sees a lot of cars on the street, on TV, toy cars etc., so also a lot of "unsupervised learning" happens beforehand.

Finally, neural networks have almost nothing in common with human brain, so there's not much point in comparing them. Also notice that there are algorithms for one shot learning, and pretty much research on it currently happens.

share|cite|improve this answer

answered 3 hours ago

Tim♦Tim

58k9127219

$endgroup$

add a comment | 

$begingroup$

First of all, at age two, child knows a lot about world and actively applies this knowledge. Child does a lot of "transfer learning" by applying this knowledge to new concepts.

Second, before seeing those five "labeled" examples of car, child sees a lot of cars on the street, on TV, toy cars etc., so also a lot of "unsupervised learning" happens beforehand.

Finally, neural networks have almost nothing in common with human brain, so there's not much point in comparing them. Also notice that there are algorithms for one shot learning, and pretty much research on it currently happens.

share|cite|improve this answer

answered 3 hours ago

Tim♦Tim

58k9127219

$endgroup$

share|cite|improve this answer

answered 3 hours ago

Tim♦Tim

58k9127219

answered 3 hours ago

Tim♦Tim

58k9127219

answered 3 hours ago

Tim♦Tim

58k9127219

2

$begingroup$

There's a kind of goalpost moving that underlies this question. It used to be that NNs weren't very good at image recognition, so no one compared them to humans. Now that NNs are good at image tasks, suddenly it's the fault of NNs that they require a lot of training data to be comparable to children.

You can also turn this logic on its head. Suppose a child sees a number of cars the day that it's born. I wouldn't expect the child to be able to pick out a car the next day, or the next week even though it's seen so many examples. Why are newborns so slow to learn? Because it takes a lot of exposure to the real world and time to change the child's neural pathways ("training data").

share|cite|improve this answer

answered 2 hours ago

SycoraxSycorax

40.4k12103203

$endgroup$

add a comment | 

2

$begingroup$

There's a kind of goalpost moving that underlies this question. It used to be that NNs weren't very good at image recognition, so no one compared them to humans. Now that NNs are good at image tasks, suddenly it's the fault of NNs that they require a lot of training data to be comparable to children.

You can also turn this logic on its head. Suppose a child sees a number of cars the day that it's born. I wouldn't expect the child to be able to pick out a car the next day, or the next week even though it's seen so many examples. Why are newborns so slow to learn? Because it takes a lot of exposure to the real world and time to change the child's neural pathways ("training data").

share|cite|improve this answer

answered 2 hours ago

SycoraxSycorax

40.4k12103203

$endgroup$

add a comment | 

$begingroup$

There's a kind of goalpost moving that underlies this question. It used to be that NNs weren't very good at image recognition, so no one compared them to humans. Now that NNs are good at image tasks, suddenly it's the fault of NNs that they require a lot of training data to be comparable to children.

You can also turn this logic on its head. Suppose a child sees a number of cars the day that it's born. I wouldn't expect the child to be able to pick out a car the next day, or the next week even though it's seen so many examples. Why are newborns so slow to learn? Because it takes a lot of exposure to the real world and time to change the child's neural pathways ("training data").

share|cite|improve this answer

answered 2 hours ago

SycoraxSycorax

40.4k12103203

$endgroup$

share|cite|improve this answer

answered 2 hours ago

SycoraxSycorax

40.4k12103203

answered 2 hours ago

SycoraxSycorax

40.4k12103203

answered 2 hours ago

SycoraxSycorax

40.4k12103203

0

$begingroup$

Please remember that neural networks are very complex structures with lots of parametres to tune, therefore you should not expect it to "learn" without huge numbers of examples.

This is the reason why simpler models are often preferred when samples are not huge.

share|cite|improve this answer

answered 1 hour ago

asdfasdf

413

New contributor

asdf 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 | 

0

$begingroup$

Please remember that neural networks are very complex structures with lots of parametres to tune, therefore you should not expect it to "learn" without huge numbers of examples.

This is the reason why simpler models are often preferred when samples are not huge.

share|cite|improve this answer

answered 1 hour ago

asdfasdf

413

New contributor

asdf 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 | 

$begingroup$

Please remember that neural networks are very complex structures with lots of parametres to tune, therefore you should not expect it to "learn" without huge numbers of examples.

This is the reason why simpler models are often preferred when samples are not huge.

share|cite|improve this answer

answered 1 hour ago

asdfasdf

413

New contributor

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

$endgroup$

share|cite|improve this answer

answered 1 hour ago

asdfasdf

413

New contributor

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

answered 1 hour ago

asdfasdf

413

New contributor

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

answered 1 hour ago

asdfasdf

413