All language subtitles for 1. What is Deep Learning

af Afrikaans
ak Akan
sq Albanian
am Amharic
ar Arabic
hy Armenian
az Azerbaijani
eu Basque
be Belarusian
bem Bemba
bn Bengali
bh Bihari
bs Bosnian
br Breton
bg Bulgarian
km Cambodian
ca Catalan
chr Cherokee
ny Chichewa
zh-CN Chinese (Simplified)
zh-TW Chinese (Traditional)
co Corsican
hr Croatian
cs Czech
da Danish
nl Dutch
en English
eo Esperanto
et Estonian
ee Ewe
fo Faroese
tl Filipino
fi Finnish
fr French Download
fy Frisian
gaa Ga
gl Galician
ka Georgian
de German
el Greek
gn Guarani
gu Gujarati
ht Haitian Creole
ha Hausa
haw Hawaiian
iw Hebrew
hi Hindi
hu Hungarian
is Icelandic
ig Igbo
id Indonesian
ia Interlingua
ga Irish
it Italian
ja Japanese
jw Javanese
kn Kannada
kk Kazakh
rw Kinyarwanda
rn Kirundi
kg Kongo
ko Korean
kri Krio (Sierra Leone)
ku Kurdish
ckb Kurdish (Soranî)
ky Kyrgyz
lo Laothian
la Latin
lv Latvian
ln Lingala
lt Lithuanian
loz Lozi
lg Luganda
ach Luo
mk Macedonian
mg Malagasy
ms Malay
ml Malayalam
mt Maltese
mi Maori
mr Marathi
mfe Mauritian Creole
mo Moldavian
mn Mongolian
sr-ME Montenegrin
ne Nepali
pcm Nigerian Pidgin
nso Northern Sotho
no Norwegian
nn Norwegian (Nynorsk)
oc Occitan
or Oriya
om Oromo
ps Pashto
fa Persian
pl Polish
pt-BR Portuguese (Brazil)
pt-PT Portuguese (Portugal)
pa Punjabi
qu Quechua
ro Romanian
rm Romansh
nyn Runyakitara
ru Russian
gd Scots Gaelic
sr Serbian
sh Serbo-Croatian
st Sesotho
tn Setswana
crs Seychellois Creole
sn Shona
sd Sindhi
si Sinhalese
sk Slovak
sl Slovenian
so Somali
es Spanish
es-419 Spanish (Latin American)
su Sundanese
sw Swahili
sv Swedish
tg Tajik
ta Tamil
tt Tatar
te Telugu
th Thai
ti Tigrinya
to Tonga
lua Tshiluba
tum Tumbuka
tr Turkish
tk Turkmen
tw Twi
ug Uighur
uk Ukrainian
ur Urdu
uz Uzbek
vi Vietnamese
cy Welsh
wo Wolof
xh Xhosa
yi Yiddish
yo Yoruba
zu Zulu
Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 0 1 00:00:00,300 --> 00:00:01,610 What Alison should know-- 1 2 00:00:01,610 --> 00:00:04,400 What is the Internet, anyway? 2 3 00:00:04,400 --> 00:00:07,870 Internet is that massive computer network, 3 4 00:00:07,870 --> 00:00:10,900 the one that's becoming really big now. 4 5 00:00:10,900 --> 00:00:11,920 What do you mean that's big? 5 6 00:00:11,920 --> 00:00:14,320 How does (mumbles), you write to it, like mail? 6 7 00:00:14,320 --> 00:00:16,290 No, a lot of people use it to communicate with, 7 8 00:00:16,290 --> 00:00:17,950 I guess they can communicate with NBC, 8 9 00:00:17,950 --> 00:00:18,840 writers and producers. 9 10 00:00:18,840 --> 00:00:21,308 Alison, can you explain what Internet is? 10 11 00:00:21,308 --> 00:00:24,058 (dramatic music) 11 12 00:00:29,060 --> 00:00:30,550 How amazing is that? 12 13 00:00:30,550 --> 00:00:31,900 Just over 20 years ago, 13 14 00:00:31,900 --> 00:00:34,380 people didn't even know what the Internet was 14 15 00:00:34,380 --> 00:00:37,210 and today we can't even imagine our lives without it. 15 16 00:00:37,210 --> 00:00:39,460 Welcome to the Deep Learning A-Z course. 16 17 00:00:39,460 --> 00:00:40,620 My name is Kirill Eremenko, 17 18 00:00:40,620 --> 00:00:43,140 and along with the co-instructor Hadelin de Ponteves, 18 19 00:00:43,140 --> 00:00:45,370 we're super excited to have you on board. 19 20 00:00:45,370 --> 00:00:47,500 And today we're going to give you a quick overview 20 21 00:00:47,500 --> 00:00:52,180 of what deep learning is and why it's picking up right now. 21 22 00:00:52,180 --> 00:00:53,520 So, let's get started. 22 23 00:00:53,520 --> 00:00:55,290 Why did we have a look at that clip? 23 24 00:00:55,290 --> 00:00:57,570 And what is this photo over here? 24 25 00:00:57,570 --> 00:01:00,190 Well, that clip was from 1994, 25 26 00:01:00,190 --> 00:01:03,110 this is a photo of a computer from 1980. 26 27 00:01:03,110 --> 00:01:05,990 And the reason why we're kinda delving into history 27 28 00:01:05,990 --> 00:01:08,980 a little bit is because neural networks, 28 29 00:01:08,980 --> 00:01:09,970 along with deep learning, 29 30 00:01:09,970 --> 00:01:12,060 have been around for quite some time, 30 31 00:01:12,060 --> 00:01:14,830 and they've only started picking up now 31 32 00:01:14,830 --> 00:01:16,660 and impacting the world right now. 32 33 00:01:16,660 --> 00:01:19,180 But if you look back at the '80s you'll see that, 33 34 00:01:19,180 --> 00:01:22,160 even though they were invented in the '60s and '70s, 34 35 00:01:22,160 --> 00:01:27,160 they really caught onto a trend or caught wind in the '80s. 35 36 00:01:28,410 --> 00:01:30,700 People started talking about them a lot, 36 37 00:01:30,700 --> 00:01:32,780 there was a lot of research in that area, 37 38 00:01:32,780 --> 00:01:35,810 and everybody thought that deep learning or neural networks 38 39 00:01:35,810 --> 00:01:39,890 were this new thing that is going to impact the world, 39 40 00:01:39,890 --> 00:01:41,230 is going to change everything, 40 41 00:01:41,230 --> 00:01:42,910 is gonna solve all the world problems, 41 42 00:01:42,910 --> 00:01:46,080 and then it kind of slowly died off over the next decade. 42 43 00:01:46,080 --> 00:01:46,913 So what happened? 43 44 00:01:46,913 --> 00:01:49,810 Why did the neural networks not survive 44 45 00:01:49,810 --> 00:01:51,790 and not change the world? 45 46 00:01:51,790 --> 00:01:53,950 The reason for that, that they were just not good enough, 46 47 00:01:53,950 --> 00:01:57,160 that they're not that good at predicting things 47 48 00:01:57,160 --> 00:01:58,570 and not that good at modeling 48 49 00:01:58,570 --> 00:02:02,270 and, basically, just not a good invention, 49 50 00:02:02,270 --> 00:02:03,330 or is there another reason? 50 51 00:02:03,330 --> 00:02:05,060 Well, actually, there is another reason. 51 52 00:02:05,060 --> 00:02:06,660 And the reason is in front of us, 52 53 00:02:06,660 --> 00:02:08,800 it's the fact that technology back then 53 54 00:02:08,800 --> 00:02:11,560 was not up to the right standard 54 55 00:02:11,560 --> 00:02:13,660 to facilitate neural networks. 55 56 00:02:13,660 --> 00:02:16,310 In order for neural networks and deep learning 56 57 00:02:16,310 --> 00:02:17,800 to work properly, you need two things. 57 58 00:02:17,800 --> 00:02:20,130 You need data, and you need a lot of data, 58 59 00:02:20,130 --> 00:02:21,530 and you need processing power, 59 60 00:02:21,530 --> 00:02:24,030 you need strong computers to process that data 60 61 00:02:24,030 --> 00:02:25,900 and facilitate the neural networks. 61 62 00:02:25,900 --> 00:02:29,750 So let's have a look at how data, 62 63 00:02:29,750 --> 00:02:32,240 or storage of data, has evolved over the years, 63 64 00:02:32,240 --> 00:02:34,760 and then we'll look at how technology has evolved. 64 65 00:02:34,760 --> 00:02:36,090 So, here we've got three years. 65 66 00:02:36,090 --> 00:02:38,763 1956, 1980, 2017. 66 67 00:02:39,750 --> 00:02:43,160 How did storage look back in 1956? 67 68 00:02:43,160 --> 00:02:45,150 Well, there is a hard drive, 68 69 00:02:45,150 --> 00:02:48,380 and that hard drive is only a 5, wait for it, 69 70 00:02:48,380 --> 00:02:49,680 megabyte hard drive. 70 71 00:02:49,680 --> 00:02:53,570 That's 5 megabytes right there on the forklift, 71 72 00:02:53,570 --> 00:02:56,510 the size of a small room, that's a hard drive 72 73 00:02:56,510 --> 00:03:00,873 being transported to another location on a plane. 73 74 00:03:02,450 --> 00:03:05,470 That is what storage looked like in 1956. 74 75 00:03:05,470 --> 00:03:08,930 You had to pay, a company had to pay $2,500, 75 76 00:03:08,930 --> 00:03:12,700 of those days dollars, to rent that hard drive. 76 77 00:03:12,700 --> 00:03:15,333 To rent it, not buy it, to rent it for one month. 77 78 00:03:16,280 --> 00:03:18,710 In 1980, the situation improved a little bit. 78 79 00:03:18,710 --> 00:03:22,730 So here we got a 10 megabytes hard drive for $3,500. 79 80 00:03:22,730 --> 00:03:25,070 It's still very expensive and only 10 megabytes, 80 81 00:03:25,070 --> 00:03:27,170 so that's like one photo these days. 81 82 00:03:27,170 --> 00:03:32,170 And today, in 2017, we've got a 256 gigabyte SSD card 82 83 00:03:32,780 --> 00:03:36,980 for $150 which can fit on your finger. 83 84 00:03:36,980 --> 00:03:39,510 And if you're watching this video, 84 85 00:03:39,510 --> 00:03:42,540 a year later or, like, in 2019 or 2025, 85 86 00:03:42,540 --> 00:03:43,760 you're probably laughing to yourself 86 87 00:03:43,760 --> 00:03:47,160 because by then you have even stronger storage capacity. 87 88 00:03:47,160 --> 00:03:49,060 But, nevertheless, the point stands. 88 89 00:03:49,060 --> 00:03:51,170 If we compare these across the board 89 90 00:03:51,170 --> 00:03:53,920 and, without even taking price and size into consideration, 90 91 00:03:53,920 --> 00:03:58,280 just the capacity of whatever was trending at the time, 91 92 00:03:58,280 --> 00:04:03,280 so from 1956 to 1980 capacity increased about double 92 93 00:04:04,130 --> 00:04:09,110 and then it increased about 25,600 times. 93 94 00:04:09,110 --> 00:04:13,410 And the length of the period is not that different, 94 95 00:04:13,410 --> 00:04:16,130 from 1956 to 1980 is 24 years, 95 96 00:04:16,130 --> 00:04:18,710 from 1980 to 2017, 37 years, 96 97 00:04:18,710 --> 00:04:21,680 so not that much of an increase in time 97 98 00:04:21,680 --> 00:04:24,800 but a huge jump in technological progress, 98 99 00:04:24,800 --> 00:04:28,140 and that stands to show that this is not a linear trend, 99 100 00:04:28,140 --> 00:04:30,540 this is an exponential growth in technology. 100 101 00:04:30,540 --> 00:04:33,750 And if we take into account price and size, 101 102 00:04:33,750 --> 00:04:37,170 it'll be in the millions of increase. 102 103 00:04:37,170 --> 00:04:40,540 And here we actually have a chart on a logarithmic scale, 103 104 00:04:40,540 --> 00:04:44,420 so if we plot the hard drive cost per gigabyte 104 105 00:04:44,420 --> 00:04:46,300 you'll see that it looks something like this. 105 106 00:04:46,300 --> 00:04:49,830 We're very quickly approaching zero. 106 107 00:04:49,830 --> 00:04:52,880 Right now you can get storage on DropBox and Google Drive, 107 108 00:04:52,880 --> 00:04:55,671 which doesn't cost you anything, Cloud storage, 108 109 00:04:55,671 --> 00:04:57,250 and that's going to continue. 109 110 00:04:57,250 --> 00:04:58,890 And, in fact, over the years, 110 111 00:04:58,890 --> 00:05:01,180 this is going to go even further. 111 112 00:05:01,180 --> 00:05:03,820 Right now, scientists are even looking into 112 113 00:05:03,820 --> 00:05:05,925 using DNA for storage. 113 114 00:05:05,925 --> 00:05:07,640 Right now it's quite expensive, 114 115 00:05:07,640 --> 00:05:11,950 it costs $7,000 to synthesize 2 megabytes of data 115 116 00:05:12,920 --> 00:05:15,140 and then another $2,000 to read it, 116 117 00:05:15,140 --> 00:05:17,130 but that kinda reminds you of this whole situation 117 118 00:05:17,130 --> 00:05:18,900 of the hard drive and the plane, 118 119 00:05:18,900 --> 00:05:21,420 that this is gonna be mitigated very, very quickly 119 120 00:05:21,420 --> 00:05:23,220 with this exponential growth. 120 121 00:05:23,220 --> 00:05:25,210 10 years from now, 20 years from now, 121 122 00:05:25,210 --> 00:05:26,960 everybody's gonna be using DNA storage 122 123 00:05:26,960 --> 00:05:28,600 if we go down this direction. 123 124 00:05:28,600 --> 00:05:30,990 Here are some stats on all that. 124 125 00:05:30,990 --> 00:05:33,550 You can explore this further, maybe pause the video 125 126 00:05:33,550 --> 00:05:35,440 if you want to read a bit more about this. 126 127 00:05:35,440 --> 00:05:36,910 This is from nature.com. 127 128 00:05:36,910 --> 00:05:40,430 And, basically, you can store all of the world's data 128 129 00:05:40,430 --> 00:05:44,610 in just 1 kilogram of DNA storage. 129 130 00:05:44,610 --> 00:05:47,610 Or you can store about one billion terabytes of data 130 131 00:05:47,610 --> 00:05:49,300 in one gram of DNA storage. 131 132 00:05:49,300 --> 00:05:52,070 So, that's just something to show 132 133 00:05:52,070 --> 00:05:53,650 how quickly we're progressing 133 134 00:05:53,650 --> 00:05:56,700 and that this is why deep learning is picking up 134 135 00:05:56,700 --> 00:05:58,890 now that we are finally at the stage 135 136 00:05:58,890 --> 00:06:00,990 where we have enough data to train 136 137 00:06:02,283 --> 00:06:04,200 super cool, super sophisticated models. 137 138 00:06:04,200 --> 00:06:06,368 Back then, in the '80s when it was first initiated, 138 139 00:06:06,368 --> 00:06:08,610 it just wasn't the case. 139 140 00:06:08,610 --> 00:06:12,680 And the second thing we talked about is processing capacity. 140 141 00:06:12,680 --> 00:06:15,240 So here we've got an exponential curve, 141 142 00:06:15,240 --> 00:06:17,760 again, on a log scale. 142 143 00:06:17,760 --> 00:06:20,140 It's not ideally portrayed here 143 144 00:06:20,140 --> 00:06:21,910 but on the right you can see it's a log scale. 144 145 00:06:21,910 --> 00:06:24,310 And this is how computers have been evolving. 145 146 00:06:24,310 --> 00:06:26,280 So, again, feel free to pause this slide. 146 147 00:06:26,280 --> 00:06:28,800 This is called Moore's law, you've probably heard of it, 147 148 00:06:28,800 --> 00:06:32,180 how quickly the processing capacity of computers 148 149 00:06:32,180 --> 00:06:34,120 has been evolving. 149 150 00:06:34,120 --> 00:06:35,840 Right now we're somewhere over here, 150 151 00:06:35,840 --> 00:06:39,080 where an average computer you can buy for 1,000 bucks 151 152 00:06:39,080 --> 00:06:43,600 thinks at the speed of the brain of a rat. 152 153 00:06:43,600 --> 00:06:47,520 And by 2025 it'll be the speed of a human, or 2023. 153 154 00:06:47,520 --> 00:06:50,970 And then by 2050 or 2045, 154 155 00:06:50,970 --> 00:06:54,650 we'll surpass all of the humans combined. 155 156 00:06:54,650 --> 00:06:58,250 So, basically, we are entering the era of computers 156 157 00:06:58,250 --> 00:07:00,250 that are extremely powerful, 157 158 00:07:00,250 --> 00:07:05,250 that can process things way faster than we can imagine, 158 159 00:07:05,620 --> 00:07:08,490 and that is what is facilitating deep learning. 159 160 00:07:08,490 --> 00:07:10,710 So, all this brings us to the question, 160 161 00:07:10,710 --> 00:07:12,160 what is deep learning? 161 162 00:07:12,160 --> 00:07:15,350 What is this whole neural network situation? 162 163 00:07:15,350 --> 00:07:18,110 What is going on, what are we even talking about here? 163 164 00:07:18,110 --> 00:07:20,460 And you've probably seen a picture of something like this, 164 165 00:07:20,460 --> 00:07:21,500 so let's dive into it. 165 166 00:07:21,500 --> 00:07:23,390 What is deep learning? 166 167 00:07:23,390 --> 00:07:25,510 This gentleman over here, Geoffrey Hinton, 167 168 00:07:25,510 --> 00:07:28,963 is known as the godfather of deep learning. 168 169 00:07:29,850 --> 00:07:33,400 He did research on deep learning in the '80s 169 170 00:07:33,400 --> 00:07:35,740 and he's done lots and lots of work, 170 171 00:07:35,740 --> 00:07:40,740 lots of research papers he's published in deep learning. 171 172 00:07:40,970 --> 00:07:42,910 Right now he works at Google. 172 173 00:07:42,910 --> 00:07:45,160 So, a lot of the things that we're gonna be talking about 173 174 00:07:45,160 --> 00:07:47,550 actually come from Geoffrey Hinton. 174 175 00:07:47,550 --> 00:07:49,730 You can see he's got a quite a few Youtube videos, 175 176 00:07:49,730 --> 00:07:52,310 he explains things really well, so, 176 177 00:07:52,310 --> 00:07:53,600 highly recommend checking them out. 177 178 00:07:53,600 --> 00:07:56,510 So, the idea behind deep learning is to 178 179 00:07:57,610 --> 00:07:59,197 look at the human brain, 179 180 00:07:59,197 --> 00:08:02,340 and there's gonna be quite a bit of neuroscience coming up 180 181 00:08:02,340 --> 00:08:03,230 in these tutorials, 181 182 00:08:03,230 --> 00:08:05,250 and what we're trying to do here 182 183 00:08:05,250 --> 00:08:09,810 is to mimic how the human brain operates. 183 184 00:08:09,810 --> 00:08:10,950 We don't know that much, 184 185 00:08:10,950 --> 00:08:12,240 we don't know everything about the human brain, 185 186 00:08:12,240 --> 00:08:14,110 but that little amount that we know, 186 187 00:08:14,110 --> 00:08:16,640 we want to mimic it and recreate it. 187 188 00:08:16,640 --> 00:08:17,473 And why is that? 188 189 00:08:17,473 --> 00:08:18,920 Well, because the human brain seems to be 189 190 00:08:18,920 --> 00:08:22,170 one of the most powerful tools on this planet for learning, 190 191 00:08:22,170 --> 00:08:25,500 for learning adapting skills and then applying them. 191 192 00:08:25,500 --> 00:08:29,510 If computers could copy that, then we could just leverage 192 193 00:08:29,510 --> 00:08:32,900 what natural selection has already decided for us, 193 194 00:08:32,900 --> 00:08:36,770 all of those algorithms that it has decided are the best, 194 195 00:08:36,770 --> 00:08:38,600 we're just gonna leverage that, (mumbles). 195 196 00:08:39,780 --> 00:08:41,640 So, let's see how this works. 196 197 00:08:41,640 --> 00:08:44,960 Here we've got some neurons. 197 198 00:08:44,960 --> 00:08:48,900 These are neurons which have been smeared onto glass 198 199 00:08:48,900 --> 00:08:51,220 and then have been looked at under a microscope 199 200 00:08:51,220 --> 00:08:52,130 with some coloring, 200 201 00:08:52,130 --> 00:08:54,350 and you can see what they look like. 201 202 00:08:54,350 --> 00:08:56,810 They have like a body, they have these branches, 202 203 00:08:56,810 --> 00:08:58,529 and they have like tails, and so on, 203 204 00:08:58,529 --> 00:09:01,443 you can see then they have a nucleus inside in the middle. 204 205 00:09:02,670 --> 00:09:05,210 That's basically what a neuron looks like. 205 206 00:09:05,210 --> 00:09:08,020 In the human brain there is approximately 206 207 00:09:08,020 --> 00:09:10,630 a 100 billion neurons altogether. 207 208 00:09:10,630 --> 00:09:11,680 These are individual neurons, 208 209 00:09:11,680 --> 00:09:13,640 these are actually motor neurons 209 210 00:09:13,640 --> 00:09:15,220 because they're bigger, they're easier to see. 210 211 00:09:15,220 --> 00:09:18,310 But nevertheless there's a 100 billion neurons 211 212 00:09:18,310 --> 00:09:19,870 in the human brain, 212 213 00:09:19,870 --> 00:09:21,430 and each neuron is connected 213 214 00:09:21,430 --> 00:09:23,880 to as many as about a 1,000 of its neighbors. 214 215 00:09:23,880 --> 00:09:26,510 So, to give you a picture, this is what it looks like. 215 216 00:09:26,510 --> 00:09:31,510 This is an actual section of the human brain, 216 217 00:09:32,580 --> 00:09:37,580 this is the cerebellum, which is this part of your brain. 217 218 00:09:38,080 --> 00:09:38,913 At the back. 218 219 00:09:38,913 --> 00:09:40,910 It's responsible for 219 220 00:09:42,210 --> 00:09:45,720 motorics and for keeping your balance 220 221 00:09:45,720 --> 00:09:48,520 and some language capabilities and stuff like that. 221 222 00:09:48,520 --> 00:09:53,520 This is just to show how vast, how many neurons there are. 222 223 00:09:54,380 --> 00:09:57,900 There are billions and billions and billions of neurons 223 224 00:09:57,900 --> 00:09:58,820 all connecting your brain. 224 225 00:09:58,820 --> 00:10:02,040 It's not like we're talking about five or 500 or a 1,000 225 226 00:10:02,040 --> 00:10:05,310 or a million, there's billions of neurons in there. 226 227 00:10:05,310 --> 00:10:08,260 Yeah, so, that's what we're gonna be trying to recreate. 227 228 00:10:08,260 --> 00:10:11,800 So, how do we recreate this in a computer? 228 229 00:10:11,800 --> 00:10:14,440 Well, we create an artificial structure, 229 230 00:10:14,440 --> 00:10:16,320 called an artificial neural net, 230 231 00:10:16,320 --> 00:10:20,480 where we have nodes, or neurons. 231 232 00:10:20,480 --> 00:10:23,360 And we're gonna have some neurons for input value, 232 233 00:10:23,360 --> 00:10:27,530 so these are values that you know about a certain situation. 233 234 00:10:27,530 --> 00:10:29,490 For instance, if you're modeling something, 234 235 00:10:29,490 --> 00:10:30,705 you want to predict something, 235 236 00:10:30,705 --> 00:10:32,020 you're always gonna have some input, 236 237 00:10:32,020 --> 00:10:35,228 something to start your predictions off. 237 238 00:10:35,228 --> 00:10:36,690 That's called the input layer. 238 239 00:10:36,690 --> 00:10:38,010 Then you have the output, 239 240 00:10:38,010 --> 00:10:40,130 so that's a value that you want to predict, 240 241 00:10:40,130 --> 00:10:41,060 whether it's a price, 241 242 00:10:41,060 --> 00:10:44,340 whether it's is somebody going to leave the bank 242 243 00:10:44,340 --> 00:10:45,410 or stay in the bank, 243 244 00:10:45,410 --> 00:10:47,860 is this a fraudulent transaction, 244 245 00:10:47,860 --> 00:10:50,810 is this a real transaction, and so on. 245 246 00:10:50,810 --> 00:10:52,440 So that's gonna be output layer. 246 247 00:10:52,440 --> 00:10:55,403 And in between, we're going to have a hidden layer. 247 248 00:10:56,510 --> 00:10:59,690 As you can see, in your brain you have so many neurons, 248 249 00:10:59,690 --> 00:11:01,300 so some information is coming in 249 250 00:11:01,300 --> 00:11:05,020 through your eyes, ears, nose, basically your senses, 250 251 00:11:05,020 --> 00:11:08,670 and then it's not just going right away to the output 251 252 00:11:08,670 --> 00:11:09,630 where you have the result, 252 253 00:11:09,630 --> 00:11:11,600 it's going through all of these billions and billions 253 254 00:11:11,600 --> 00:11:14,440 and billions of neurons before it gets to the output. 254 255 00:11:14,440 --> 00:11:15,760 And this is the whole concept behind it 255 256 00:11:15,760 --> 00:11:16,960 that we're going to model the brain, 256 257 00:11:16,960 --> 00:11:18,550 so we need these hidden layers 257 258 00:11:18,550 --> 00:11:20,510 that are there before the output. 258 259 00:11:20,510 --> 00:11:23,120 So, the input layers neurons are connected 259 260 00:11:23,120 --> 00:11:24,300 to the hidden layer neurons, 260 261 00:11:24,300 --> 00:11:26,350 the hidden layer neurons are connected to output value. 261 262 00:11:26,350 --> 00:11:30,490 So this is pretty cool but what is this all about? 262 263 00:11:30,490 --> 00:11:32,020 Where is the deep learning here? 263 264 00:11:32,020 --> 00:11:32,890 Why is it called deep learning? 264 265 00:11:32,890 --> 00:11:34,000 There's nothing deep in here. 265 266 00:11:34,000 --> 00:11:36,970 Well, this is kind of like an option 266 267 00:11:36,970 --> 00:11:39,390 which one might call shallow learning, 267 268 00:11:39,390 --> 00:11:41,800 where there isn't much indeed going on. 268 269 00:11:41,800 --> 00:11:43,340 But, why is it called deep learning? 269 270 00:11:43,340 --> 00:11:46,090 Well, because then we take this to the next level. 270 271 00:11:46,090 --> 00:11:48,180 We separate it even further 271 272 00:11:48,180 --> 00:11:50,940 and we have not just one hidden layer, 272 273 00:11:50,940 --> 00:11:55,470 we have lots and lots and lots of hidden layers, 273 274 00:11:55,470 --> 00:11:57,710 and then we connect everything, 274 275 00:11:57,710 --> 00:11:59,110 just like in the human brain. 275 276 00:11:59,110 --> 00:12:01,850 Connect everything, interconnect everything, 276 277 00:12:01,850 --> 00:12:06,110 and that's how the input values are processed 277 278 00:12:06,110 --> 00:12:07,380 through all these hidden layers, 278 279 00:12:07,380 --> 00:12:10,150 just like in the human brain, then we have an output value. 279 280 00:12:10,150 --> 00:12:12,360 And now we're talking deep learning. 280 281 00:12:12,360 --> 00:12:14,290 So that's what deep learning is all about 281 282 00:12:14,290 --> 00:12:15,800 on a very abstract level. 282 283 00:12:15,800 --> 00:12:18,260 In the further tutorials we're going to dissect 283 284 00:12:18,260 --> 00:12:20,400 and dive deep into deep learning, 284 285 00:12:20,400 --> 00:12:21,840 and by the end of it you will know 285 286 00:12:21,840 --> 00:12:23,370 what deep learning is all about 286 287 00:12:23,370 --> 00:12:26,350 and you'll know how to apply it in your projects. 287 288 00:12:26,350 --> 00:12:29,680 Super excited about this, can't wait to get started, 288 289 00:12:29,680 --> 00:12:31,840 and I look forward to seeing you on the next tutorial. 289 290 00:12:31,840 --> 00:12:33,803 Until then, enjoy deep learning. 24436

Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.