id
stringlengths
66
88
text
stringlengths
46
1.7k
duration
float64
2.32
30
dnsmos
float64
2.8
3.65
sampling_rate
int64
24k
24k
path
stringlengths
79
101
audio_content_token_indices
listlengths
58
749
audio_global_embedding
listlengths
128
128
audio_token_len
int64
58
749
audio_duration
float64
2.32
30
bilibili_data_13232720_BV11t411A7Ym_p2_BV11t411A7Ym_p2_m4-dialogue_0616556
[S1] Now, the nice thing is, this course, not only is this course free, but these books are free as well. The elements of statistical learning has been free and, and the PDF's available on our websites. This new book is going to be free beginning of January when the course begins. And, uh, and that's a, with agreement with the, um, uh, with the publishers. [S2] But if you want to buy the book, that's okay too. It's nice having the hard copy. But if you want, the PDF is available. [S1] So, uh, um, we hope you enjoy the rest of the class.
28.28
3.365393
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p2_BV11t411A7Ym_p2_m4-dialogue_0616556.mp3
[ 11984, 9265, 1030, 8597, 12206, 10732, 8053, 10550, 4780, 4649, 3609, 3284, 12590, 10316, 3007, 5941, 11688, 12114, 6056, 912, 6050, 11757, 12437, 12638, 12191, 11678, 8462, 2960, 280, 3681, 3096, 731, 11125, 3987, 4560, 6762, 11493, 3898, ...
[ 1.444374918937683, -4.927915096282959, -0.2330906242132187, 0.3604261875152588, 0.32190006971359253, 0.20705515146255493, 0.3966147005558014, -0.14030104875564575, 0.20731276273727417, -3.9635539054870605, 0.8538199663162231, 0.4653603434562683, -0.5182810425758362, -0.24750030040740967, ...
707
28.28
bilibili_data_13232720_BV11t411A7Ym_p32_BV11t411A7Ym_p32_m4-dialogue_0890550
[S1] So, welcome back. Today we're going to cover model selection and regularization. But we have a special guest, my former graduate student, Daniella Witten. [S2] Hi. [S1] Welcome, Daniella. [S2] Thank you. [S1] Daniella is now at University of Washington, but maybe you want to tell students a bit about yourself and how you got to be here?
16.24
3.064741
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p32_BV11t411A7Ym_p32_m4-dialogue_0890550.mp3
[ 8210, 10003, 857, 3802, 8940, 11442, 5806, 607, 555, 3310, 5951, 6382, 12514, 3886, 7719, 9826, 3070, 12214, 11187, 8122, 10035, 11354, 4906, 10768, 6201, 8577, 9842, 2132, 4195, 3933, 12134, 2917, 6875, 11882, 12187, 1124, 8289, 7873, 850,...
[ 0.7156381607055664, -5.670505046844482, -0.05756833404302597, 0.17718605697155, 0.2410196214914322, -0.10684072226285934, 0.32507991790771484, -0.1263495683670044, 0.17672660946846008, -3.9543442726135254, 0.5978732109069824, 0.11596927046775818, -0.191175639629364, -0.08356188982725143, ...
406
16.24
bilibili_data_13232720_BV11t411A7Ym_p32_BV11t411A7Ym_p32_m4-dialogue_0890551
[S1] Yeah, well, I, um, in college, I studied math and biology, and when I was graduating, I knew I wanted to, to go to grad school in something, but I couldn't really decide on one particular thing that I wanted to study for the rest of my life. And so I ended up doing a PhD in statistics, and I was lucky enough to do it at Stanford with Rob. [S2] Mm-hmm. [S1] And, um, I graduated in 2010, moved up to Seattle, and I'm now an assistant professor at the University of Washington in the Biostat Department there.
27.56
3.311084
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p32_BV11t411A7Ym_p32_m4-dialogue_0890551.mp3
[ 11930, 9355, 3291, 9555, 10576, 10536, 5920, 8488, 10666, 8183, 878, 733, 3310, 10667, 10666, 10546, 10547, 5974, 12533, 8189, 10749, 10742, 11253, 11748, 11660, 11147, 11734, 9894, 9653, 11645, 11126, 11190, 11430, 10909, 11990, 9430, 4820, ...
[ 0.8274993300437927, -5.3887786865234375, 0.17948229610919952, 0.15098589658737183, 0.4068223237991333, -0.26640984416007996, 0.20888838171958923, 0.05062808841466904, 0.11955271661281586, -3.953413963317871, 0.3289985656738281, -0.2734428644180298, -0.016154779121279716, -0.034436218440532...
689
27.56
bilibili_data_13232720_BV11t411A7Ym_p33_BV11t411A7Ym_p33_m4-dialogue_0302721
[S1] Um, and we, we'd want to use best subset, maybe not beyond 10 or 20, say. [S2] Yeah, I don't think I would probably even use it for 20. I mean, I think I would use best subset if I've got a handful of predictors, and if I've got 10, I wouldn't be using that anymore, most likely. [S1] Yeah, so the point is, it's not always best to do a full search, even when you can do it, because you pay a price in variance.
17.72
3.130265
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p33_BV11t411A7Ym_p33_m4-dialogue_0302721.mp3
[ 3611, 11482, 9396, 11126, 11573, 9006, 9070, 8494, 3438, 6510, 11567, 9502, 3091, 549, 6804, 9910, 6583, 5660, 2804, 7981, 11614, 6740, 1569, 2741, 12502, 7811, 2669, 8566, 9365, 4949, 8517, 3286, 3457, 5802, 7599, 12662, 9525, 8160, 5512, ...
[ 1.0265231132507324, -5.6075944900512695, -0.07703837752342224, 0.22960197925567627, 0.44639724493026733, -0.15851090848445892, 0.3125098943710327, -0.060289524495601654, 0.31436100602149963, -3.994500160217285, 0.4564926028251648, -0.10985749214887619, -0.11825940012931824, -0.142358884215...
443
17.72
bilibili_data_13232720_BV11t411A7Ym_p35_BV11t411A7Ym_p35_m4-dialogue_0407423
[S1] That's right, yeah. [S2] Unlike the RSS curve. [S1] Exactly. It's a little hard to see. [S2] Yeah. [S1] But this is slightly increasing. It's smallest- [S2] Yeah. [S1] ... with four predictors, and then it goes up a little bit. But, you know, I don't really think that there's compelling evidence here that, that four is really better than three or better than five. So if it were me, you know, I think simpler's always better, so I'd probably choose a model with, you know, three predictors, maximum four predictors. [S2] Yeah, I agree.
23.08
3.15811
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p35_BV11t411A7Ym_p35_m4-dialogue_0407423.mp3
[ 6682, 1026, 6645, 11131, 11582, 8401, 968, 3308, 10709, 6133, 2438, 6355, 12757, 11246, 4093, 6958, 3491, 5923, 5484, 11099, 10533, 4373, 4371, 6006, 2852, 388, 2927, 3868, 8073, 467, 1774, 3319, 4057, 4816, 11932, 1884, 11565, 9551, 8988, ...
[ 0.8697289228439331, -5.464321136474609, -0.18287968635559082, 0.24159753322601318, 0.5142613649368286, -0.0579272136092186, -0.10593994706869125, -0.06580329686403275, 0.028913481160998344, -3.8501651287078857, 0.3796170651912689, -0.036516424268484116, 0.049748778343200684, 0.011965651996...
577
23.08
bilibili_data_13232720_BV11t411A7Ym_p35_BV11t411A7Ym_p35_m4-dialogue_0407424
[S1] So, already we see that, that CP is restricted to cases where you've got, uh, N bigger than P. [S2] That's right. [S1] And, and even if P is close to N, you're going to have a problem because you're asking if sigma squared might be, uh, far too low.
12.52
3.129652
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p35_BV11t411A7Ym_p35_m4-dialogue_0407424.mp3
[ 7259, 12408, 1738, 6048, 1419, 9007, 5887, 2663, 3223, 3803, 12452, 7270, 4198, 9844, 3236, 12507, 12450, 10833, 6390, 5493, 3349, 2924, 10740, 8963, 11749, 9420, 2870, 3912, 7136, 12685, 3716, 3509, 2132, 3922, 721, 5776, 3446, 4309, 3664,...
[ 1.157393455505371, -5.388333320617676, -0.14060303568840027, 0.36507806181907654, 0.2256491482257843, -0.1651933640241623, 0.36256328225135803, -0.08857374638319016, 0.21384841203689575, -3.9808998107910156, 0.5977161526679993, 0.22678588330745697, -0.21055187284946442, -0.1685938835144043...
313
12.52
bilibili_data_13232720_BV11t411A7Ym_p35_BV11t411A7Ym_p35_m4-dialogue_0407425
[S1] So, um, we want a large value of adjusted R squared. And, um, and so the adjusted R squared, in practice, people really like it. It tends to work really well. So some statisticians don't like it as much as, um, C-P, AIC, and BIC. [S2] [LAUGHS] [S1] And the reason is because it sort of works well empirically, but some statisticians feel that it doesn't kind of have the theoretical backing of some other approaches. [S2] Yeah. [S1] What do you think of that, Rob?
23.76
3.20294
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p35_BV11t411A7Ym_p35_m4-dialogue_0407425.mp3
[ 6184, 592, 5993, 2832, 9069, 11116, 10605, 10606, 8038, 10454, 7894, 4709, 9021, 8062, 5999, 4094, 9815, 3110, 9358, 7254, 8977, 9545, 8211, 7279, 9902, 5301, 10038, 9910, 5805, 6742, 4299, 9500, 3673, 9265, 5146, 4114, 564, 11598, 11589, ...
[ 0.8037304282188416, -5.331024169921875, -0.05024796724319458, 0.17459483444690704, 0.488565593957901, -0.18870411813259125, 0.052660197019577026, 0.05128660425543785, -0.020246436819434166, -3.9270424842834473, 0.2867027819156647, -0.10891751199960709, -0.07495246827602386, -0.011812456883...
594
23.76
bilibili_data_13232720_BV11t411A7Ym_p35_BV11t411A7Ym_p35_m4-dialogue_0407426
[S1] But one nice thing about adjusted R-squared is like if you're working with, um, someone who's not a statistician, like scientists who aren't statisticians are really familiar with R-squared. And so from un- when to understand R-squared, adjusting R-squared is just a really small one-off and it's kind of easier to explain to someone in a certain sense than AIC, CP, or BIC. And so that's one really nice thing about it. But adjusted R-squared, you can't really generalize to other types of models. [S2] Right. [S1] So if you have like logistic regression, you can't do this.
28.08
2.846638
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p35_BV11t411A7Ym_p35_m4-dialogue_0407426.mp3
[ 6240, 5416, 10604, 10907, 2675, 10484, 8121, 11929, 10818, 10681, 11616, 10066, 2905, 3657, 8692, 12129, 11924, 11756, 6371, 11260, 11706, 9421, 8687, 9931, 9163, 11062, 6428, 2904, 2834, 12454, 5516, 8054, 8151, 1825, 2896, 9835, 302, 12278,...
[ 0.8285036087036133, -5.301857948303223, 0.00397920049726963, 0.08842071145772934, 0.5458993315696716, -0.23126758635044098, -0.005226026754826307, 0.03896922990679741, -0.08638325333595276, -3.9726176261901855, 0.4223346710205078, -0.14990025758743286, -0.04102684557437897, 0.0379043780267...
702
28.08
bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616557
[S1] And I remember when I was a student seeing this figure again and again. Actually, it was in a class that Rob was teaching and being totally confused for like three lectures. [S2] Okay. [S1] So I just want to spare everyone this confusion in case anyone shares the confusion I had. So like, if we look here, this red line indicates the spot at which lambda equals 100, and at that spot, the income coefficient takes on a value of like negative 100. These other six coefficients here are all around zero. Student takes on a value of around 100.
29.08
3.077746
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616557.mp3
[ 7268, 8873, 4323, 12083, 8532, 11194, 11754, 5465, 8457, 7715, 11193, 11497, 8738, 8114, 10584, 8009, 2667, 8052, 11925, 8110, 11702, 9706, 8476, 7988, 9475, 338, 4317, 3280, 6408, 5973, 10517, 6787, 2693, 10758, 12362, 12418, 8347, 12483, ...
[ 0.5943166017532349, -5.433432579040527, 0.08936835825443268, 0.24421943724155426, 0.32945799827575684, -0.0792020708322525, -0.06830614805221558, -0.011309017427265644, 0.029618989676237106, -3.9217355251312256, 0.3810000419616699, -0.0908973440527916, -0.05757802724838257, 0.1056529358029...
727
29.08
bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616558
[S1] ... limit and rating take on values of around 250. And so the point is, what we're plotting here is a ton of different models for a huge grid of lambda values, and you just need to choose a value of lambda and then look at that vertical cross-section. [S2] Good. And is, so as Danielle said, if we chose the value of lambda about, uh, was that 100? [S1] Mm-hmm. [S2] Then it would, it seems like it chooses about three non-zero coefficients, the black, the blue, and the red. Oh, maybe four. And then these guys here are basically, essentially zero, the gray ones. [S1] So they're not quite zero, but they're small. [S2] Yeah.
29.96
3.281651
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616558.mp3
[ 6240, 8864, 3163, 5397, 11756, 12061, 10909, 11748, 9421, 11253, 11853, 3429, 10709, 12764, 9411, 12254, 12501, 12426, 4075, 11692, 9941, 6572, 8061, 11862, 3939, 11766, 8574, 8037, 12758, 1794, 3025, 3300, 3948, 8070, 9062, 6014, 6295, 9232,...
[ 0.8529624938964844, -5.3825812339782715, 0.0042066797614097595, 0.2805165648460388, 0.41650646924972534, -0.04292459413409233, 0.13634099066257477, -0.013555671088397503, 0.17098668217658997, -3.9168529510498047, 0.5250973105430603, 0.12169390916824341, -0.09848470985889435, -0.04105494916...
749
29.96
bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616559
[S1] Right, exactly. So, an equivalent picture on the right now, we've plotted the, um, the standardized coefficients as a function of the, uh, of the, the L2 norm, the sums of the squares, the square of the sum of the squares of the coefficients divided by that, the L2 norm of the fully-squared coefficients. [S2] So, Rob, what's the L2 norm? [S1] Okay. Um, the L2 norm, so the L2 norm of the beta of the vector beta 1 through beta p,
27
3.127779
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616559.mp3
[ 6824, 4689, 3358, 3045, 2331, 9746, 4885, 1282, 9015, 7100, 2358, 11434, 683, 1301, 4810, 9386, 9300, 7342, 12597, 8820, 4701, 4709, 7269, 4843, 11978, 6872, 448, 11116, 7988, 7911, 5983, 4835, 9314, 9388, 6669, 9110, 4124, 9011, 2922, 10...
[ 0.9555280804634094, -5.2461442947387695, -0.07873421162366867, 0.4024467468261719, 0.17684432864189148, -0.11435998976230621, 0.24998413026332855, -0.13612917065620422, 0.3276103436946869, -3.97979736328125, 0.6668581962585449, 0.31916147470474243, -0.23013561964035034, -0.1181829571723938...
675
27
bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616560
[S1] And in between we get again a shrunken coefficient. So these two pictures are really the same, but they've been flipped from left to right. [S2] And the, the x-axis are parameterized in a different way. [S1] Right. [S2] So Rob, why does, um, this x-axis on the right-hand side go from zero to one? Why does it end at one? [S1] Oh, because we just, we plotted the stan- it ends at one because we're plotting as a function of this standardized L2 norm. So at the right-hand side, um, we'll get, we have the, basically the full least squares estimates, so these numerator and denominator are the same.
28.84
3.055839
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616560.mp3
[ 6688, 4112, 4838, 8837, 12702, 5661, 8490, 2263, 10944, 9193, 2747, 7859, 11608, 11657, 11275, 8388, 2795, 9542, 9426, 9695, 9079, 6338, 5999, 9819, 5013, 10095, 9319, 8206, 6221, 9829, 6762, 12338, 10281, 4112, 4685, 7278, 2460, 12761, 764...
[ 1.0008597373962402, -5.280812740325928, -0.13141092658042908, 0.4208798408508301, 0.33485931158065796, -0.08657211065292358, 0.1765514612197876, -0.06266967952251434, 0.18662907183170319, -3.9099972248077393, 0.5598000288009644, 0.29784896969795227, -0.1789337694644928, -0.0940639749169349...
721
28.84
bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616561
[S1] Right. So on the right-hand side here, on this right-hand plot, lambda is zero. [S2] Right. [S1] And so your ridge regression estimate is the same as your least squares estimate, and so that ratio is just one. [S2] Exactly. Okay. Um, I think we've actually just said all this. Thanks for the questions from Daniela. Um, right. So there's the L2 norm defined. I wrote it in the previous slide, and that's what was used for the plotting axis.
26.44
3.069681
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616561.mp3
[ 4780, 8809, 5226, 5470, 10660, 10222, 1803, 3648, 466, 7918, 6070, 12462, 5843, 11260, 5853, 2908, 10724, 11752, 11666, 5899, 10098, 10674, 9522, 10765, 3346, 353, 11582, 11643, 11672, 12490, 3403, 9954, 11667, 9076, 3534, 5582, 8614, 6582, ...
[ 0.8287877440452576, -5.5099310874938965, -0.120933897793293, 0.3744635283946991, 0.38778772950172424, -0.07762701064348221, 0.10446706414222717, -0.10528765618801117, 0.12750114500522614, -3.8109652996063232, 0.5163467526435852, 0.05080293491482735, -0.076676145195961, -0.04782728478312492...
661
26.44
bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616562
[S1] Um, by a certain amount. And we see the same thing on the, in this picture. [S2] And actually, this U-shaped curve that we see for the, um, the mean squared error in this figure in purple comes up again and again. [S1] Yeah. [S2] Where when we're considering, um, a bunch of different models that sort of have different levels of flexibility or complexity, there's usually some sweet spot in the middle that has the smallest test error, and that's really what we're going for. So that's what's marked as an X in these two figures.
23.64
3.149284
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p37_BV11t411A7Ym_p37_m4-dialogue_0616562.mp3
[ 9240, 6185, 6925, 6765, 3958, 3822, 3822, 3814, 2581, 8295, 9758, 7340, 4715, 3691, 8352, 10777, 9753, 8714, 6185, 7245, 4657, 7978, 423, 4965, 3997, 5976, 9169, 2861, 5461, 1747, 3596, 5991, 3606, 8357, 3966, 4349, 622, 1562, 9353, 4787,...
[ 0.8057138323783875, -5.487063884735107, -0.10343439131975174, 0.35644036531448364, 0.5546794533729553, -0.12895993888378143, 0.016712171956896782, -0.022691018879413605, -0.04211200028657913, -3.8723504543304443, 0.47238510847091675, 0.04560260474681854, 0.043120790272951126, -0.0060628987...
591
23.64
bilibili_data_13232720_BV11t411A7Ym_p38_BV11t411A7Ym_p38_m4-dialogue_0580206
[S1] And by the way, if that looks like a total mystery, if you can like reach back to your, your distant or not so distant past, if you ever took AP calculus and you saw Lagrange multipliers in high school, this is really something that you, you might have truly seen in, in high school calculus a long time ago, but for simpler types of problems. [S2] Right. [S1] And that's just a, this is just a more complex application of that same idea.
19.56
3.239547
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p38_BV11t411A7Ym_p38_m4-dialogue_0580206.mp3
[ 9753, 5648, 9589, 11420, 8930, 8189, 11770, 4299, 6068, 5355, 11771, 10118, 6623, 1313, 3513, 12276, 10746, 6820, 5339, 8052, 11947, 4513, 2851, 11123, 12442, 9527, 8002, 5097, 9589, 10477, 7974, 8979, 10430, 7957, 5662, 11566, 11757, 11612, ...
[ 0.6753225922584534, -5.58550500869751, -0.1550883948802948, 0.4027999937534332, 0.4647235572338104, -0.13653916120529175, -0.09144949167966843, -0.03366998955607414, -0.12198065966367722, -3.8995230197906494, 0.47028201818466187, 0.06477968394756317, 0.040841586887836456, 0.031235344707965...
489
19.56
bilibili_data_13232720_BV11t411A7Ym_p38_BV11t411A7Ym_p38_m4-dialogue_0580207
[S1] And one thing we should mention is the, on this right-hand panel, the, the x-axis is something we haven't seen before, which is the R-squared on the training data. [S2] Right. [S1] And the reason we have that x-axis is because in this figure on the right-hand side, we're plotting both ridge regression and the lasso, so it wouldn't make sense to,
15.56
2.809801
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p38_BV11t411A7Ym_p38_m4-dialogue_0580207.mp3
[ 6882, 11699, 11986, 3309, 8050, 11436, 6729, 8121, 12056, 8794, 8050, 12053, 12689, 8666, 9949, 3148, 11125, 11757, 11861, 12681, 6610, 11627, 7749, 11254, 11559, 3907, 402, 1108, 1563, 4234, 937, 9062, 6479, 7429, 4047, 2390, 9324, 9470, 3...
[ 0.9506460428237915, -5.564889430999756, -0.1046319231390953, 0.3419365882873535, 0.36691591143608093, -0.17475426197052002, 0.11855567991733551, -0.03503547981381416, 0.14169928431510925, -3.915398120880127, 0.4850658178329468, -0.017291756346821785, -0.12671725451946259, -0.01472470629960...
389
15.56
bilibili_data_13232720_BV11t411A7Ym_p38_BV11t411A7Ym_p38_m4-dialogue_0580208
[S1] Okay, I would have noticed that detail otherwise. [S2] [LAUGHS] [S1] Okay. So now, now here's a situation where we do, we do perform better with the LASSO, and this is a case where now in the population, only two of the predictors have non-zero coefficients. So, the previous situation was, was, was dense, or non-sparse. This situation is, is sparse. We've only, there's only two predictors in the true model that are non-zero coefficients. Now we can see what happens. Well,
25.44
3.260515
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p38_BV11t411A7Ym_p38_m4-dialogue_0580208.mp3
[ 12394, 10080, 1821, 2398, 1916, 4886, 51, 11566, 8451, 4335, 540, 6301, 6327, 6374, 6366, 3728, 2897, 4415, 2285, 24, 6466, 12613, 7178, 8072, 2027, 2343, 1653, 2742, 2052, 1452, 1373, 164, 3903, 4979, 7447, 1818, 2512, 4787, 4715, 4706, ...
[ 0.9831793308258057, -5.3513946533203125, -0.13978995382785797, 0.35919639468193054, 0.2855237126350403, -0.04689915105700493, 0.27534809708595276, -0.11513731628656387, 0.15971125662326813, -3.9372613430023193, 0.5364754796028137, 0.33437666296958923, -0.191239595413208, -0.139385864138603...
636
25.44
bilibili_data_13232720_BV11t411A7Ym_p40_BV11t411A7Ym_p40_m4-dialogue_0688277
[S1] So in a way it's sort of similar to Ridge and Lazo, right? It's still least squares, it's still a linear model in all the variables, but there's a, there's a constraint on the coefficients. [S2] That's exactly right. [S1] Right. [S2] But we're getting a constraint in a different way. We're not getting a constraint like in the Ridge case by saying, okay, my sum of squared betas needs to be small. Instead we're saying my betas need to take this really funny form if you look at it. But it's got a simple interpretation in terms of least squares on a new set of features.
24.08
3.052532
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p40_BV11t411A7Ym_p40_m4-dialogue_0688277.mp3
[ 3225, 912, 8358, 12006, 10900, 11511, 107, 11125, 11617, 9109, 9895, 273, 1496, 3442, 11102, 2901, 3302, 4521, 3480, 11629, 8797, 3246, 8322, 1375, 661, 226, 11150, 12502, 5825, 4994, 3863, 8268, 679, 3942, 12084, 10611, 7027, 8152, 979, ...
[ 0.8007446527481079, -5.57078218460083, -0.0268155038356781, 0.07066665589809418, 0.5879855155944824, -0.11061541736125946, 0.1462741494178772, -0.0164572112262249, 0.1575763076543808, -3.9024763107299805, 0.31023216247558594, -0.03029124066233635, -0.019615795463323593, 0.02638418041169643...
602
24.08
bilibili_data_13232720_BV11t411A7Ym_p46_BV11t411A7Ym_p46_m4-dialogue_0302722
[S1] Oh, yeah. And, you know, Rob, it just occurred to me that that paper we wrote 30 years ago. [S2] Still good. Still good. [S1] Still a good paper. [S2] Okay. [S1] Okay. So, here we go. Well, the first fact is the truth is never linear.
16.92
3.237565
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p46_BV11t411A7Ym_p46_m4-dialogue_0302722.mp3
[ 9745, 6161, 2198, 7478, 5814, 2718, 3822, 9925, 4367, 4911, 6887, 4331, 4338, 9817, 3234, 4637, 9974, 10215, 5031, 6519, 9814, 2566, 11781, 4186, 6930, 4707, 4301, 11883, 4826, 10778, 9427, 11937, 9812, 7262, 4637, 9830, 9894, 9902, 9894, ...
[ 0.8123587965965271, -5.522218227386475, -0.24454671144485474, 0.12839828431606293, 0.5725323557853699, 0.10476960241794586, 0.546552300453186, -0.12224993109703064, 0.024325767531991005, -3.9908344745635986, 0.27020782232284546, 0.1207561045885086, -0.10590682923793793, -0.1827950775623321...
423
16.92
bilibili_data_13232720_BV11t411A7Ym_p46_BV11t411A7Ym_p46_m4-dialogue_0302723
[S1] Well, when I first saw that, I thought it was crazy, but then I realized the vertical scale is pretty stretched, right? So it's only ranging up to 0.2. [S2] Oh, good point, Rob. Good point. [S1] If you saw that from zero to one, it would look a lot more narrow.
13
3.263926
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p46_BV11t411A7Ym_p46_m4-dialogue_0302723.mp3
[ 6680, 6391, 11422, 2978, 11519, 6518, 12502, 8330, 4317, 12479, 9134, 3292, 9014, 11341, 11710, 4002, 6649, 8099, 8028, 1954, 5912, 2954, 8055, 5430, 8123, 9309, 2987, 11647, 3933, 1406, 6872, 1457, 12022, 8045, 2214, 172, 1237, 1984, 12387...
[ 0.92169189453125, -5.468759059906006, -0.17120563983917236, 0.1640731543302536, 0.14702920615673065, -0.09070394933223724, 0.529656171798706, -0.1372547447681427, 0.1134907454252243, -4.015805244445801, 0.30519986152648926, 0.14765068888664246, -0.023294828832149506, -0.10182292014360428, ...
325
13
bilibili_data_13232720_BV11t411A7Ym_p46_BV11t411A7Ym_p46_m4-dialogue_0302724
[S1] You can already see the advantages it has over, over polynomials, right? Th- th- this is local. Remember, with polynomials, it's a single function for the whole range of, of the, of the x variable. So, for example, if I change a, a point on the left side, it can potentially change the fit on the right side quite a bit for polynomials. [S2] That's a good point, Robert, which I forgot to say. Yeah. [S1] But it, for, uh, for step functions, uh, a fit, uh, a point only affects the, the, uh, fit in, in the partition it's sitting in and not the other partitions.
27.52
3.403745
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p46_BV11t411A7Ym_p46_m4-dialogue_0302724.mp3
[ 2561, 8990, 9357, 12129, 4838, 8533, 7996, 5979, 12123, 3360, 6992, 9555, 11596, 73, 7582, 6934, 296, 5031, 7542, 12589, 12212, 11934, 6980, 3905, 11542, 4387, 840, 6963, 12727, 9143, 12591, 320, 3375, 10414, 542, 2931, 8599, 333, 2708, 3...
[ 0.6293604373931885, -5.451765537261963, -0.014845962636172771, -0.04919625073671341, 0.30194997787475586, 0.016297107562422752, 0.35580089688301086, -0.11264771968126297, 0.10587655752897263, -3.977541923522949, 0.38652530312538147, 0.11371953785419464, -0.023772994056344032, 0.02694218605...
688
27.52
bilibili_data_13232720_BV11t411A7Ym_p47_BV11t411A7Ym_p47_m4-dialogue_0370746
[S1] There's a piecewise polynomial. It's a cubic polynomial. There's the knot at 50. And it's, it's a cubic to the left and a cubic to the right. They're two different cubic polynomials. And they just fit to the data. This one's fit to the data on the left. This one's fit to the data on the right. What do you think of that, Rob? [S2] Uh, there's a break in the middle. It's not- [S1] Pretty ugly, isn't it? It's pretty ugly. So, but that is a piecewise cubic polynomial.
29.36
3.245777
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p47_BV11t411A7Ym_p47_m4-dialogue_0370746.mp3
[ 2600, 4106, 852, 12774, 11740, 3907, 3029, 11567, 2601, 11185, 7509, 12613, 5004, 8080, 1322, 5430, 3959, 7094, 4949, 6016, 3617, 9201, 6397, 317, 3220, 4071, 4382, 3084, 3935, 1853, 1310, 4622, 10917, 11650, 10418, 5699, 3998, 8800, 385, ...
[ 1.1083437204360962, -5.271738052368164, -0.25480887293815613, -0.03627234324812889, 0.5323042869567871, 0.11379264295101166, 0.4887445569038391, -0.10484600812196732, 0.20500919222831726, -4.053943634033203, 0.7315164804458618, 0.14453192055225372, -0.18895778059959412, -0.1957659274339676...
734
29.36
bilibili_data_13232720_BV11t411A7Ym_p47_BV11t411A7Ym_p47_m4-dialogue_0370747
[S1] Okay, so cubic splines, um, are used all over the place and, and you'll see we, we even got fancier versions. One fancier version is, is very handy and it's the one I tend to use all the time. It's called a natural cubic spline. If you had the choice of a cubic spline and a natural cubic spline, which one would you use, Rob? [S2] Uh, natural, I guess. [S1] [LAUGHS] I guess it's a, it's a good choice of names.
24.32
3.397053
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p47_BV11t411A7Ym_p47_m4-dialogue_0370747.mp3
[ 4781, 11424, 2206, 10620, 10530, 4187, 11538, 8693, 5101, 2450, 1692, 3360, 848, 8062, 10965, 8218, 11361, 11819, 6939, 8753, 8908, 7849, 8761, 9306, 9939, 7274, 11347, 10209, 10649, 10522, 8417, 11756, 9877, 10155, 8656, 4761, 4019, 8030, ...
[ 1.1594465970993042, -5.25751256942749, -0.27160704135894775, 0.1212061271071434, 0.443095326423645, 0.13110941648483276, 0.4614598751068115, -0.1112879067659378, 0.19686554372310638, -3.9486300945281982, 0.478599488735199, 0.17327216267585754, -0.26453134417533875, -0.19445952773094177, ...
608
24.32
bilibili_data_13232720_BV11t411A7Ym_p47_BV11t411A7Ym_p47_m4-dialogue_0370748
[S1] The standard errors you see are much wider in places for the cubic spline, especially at the boundary, which is where the act, the, the effect of the natural spline is, is taking place. And the standard errors for the natural spline are, are better there. [S2] I'm actually working on something called organic cubic splines, which will be even better. [S1] [LAUGHS] [S2] I can't give you details yet. [S1] You called me to, but that one should catch on, I think. [S2] Especially here in California, especially. [S1] Especially in California. Very good.
25.64
3.377219
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p47_BV11t411A7Ym_p47_m4-dialogue_0370748.mp3
[ 7208, 6186, 4224, 3374, 6056, 8088, 4297, 6043, 12146, 12705, 11696, 11283, 5587, 12007, 3330, 12590, 12258, 10601, 5402, 8147, 10669, 5974, 833, 1930, 10053, 3351, 2530, 3344, 5075, 12614, 3439, 3319, 814, 539, 10998, 12222, 4643, 11073, 7...
[ 0.9603205323219299, -5.244527816772461, -0.2456013411283493, 0.03258294612169266, 0.5357315540313721, -0.00217336043715477, 0.408257395029068, -0.09063873440027237, 0.19463786482810974, -4.030846118927002, 0.4022175669670105, 0.1525331735610962, -0.14140886068344116, -0.2442539483308792, ...
641
25.64
bilibili_data_13232720_BV11t411A7Ym_p47_BV11t411A7Ym_p47_m4-dialogue_0370749
[S1] Okay. Fitting splines in R is easy. Um, there's a function BS, um, which takes a variable X and has some arguments. You can give it the knots, for example, and, uh, or you, uh, there's other ways of specifying the flexibility, and it'll just do the work. You, you put that in a formula and you can put one of those for what, whichever function you want to be, uh, non-linear, and it'll just do the work for you. [S2] Did you think of the name of that function? [S1] [LAUGHS]
26
3.330631
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p47_BV11t411A7Ym_p47_m4-dialogue_0370749.mp3
[ 12377, 11915, 12690, 12148, 8114, 8349, 11609, 11253, 9709, 2372, 3672, 8112, 8179, 12490, 8081, 11080, 9804, 5901, 8168, 4041, 8936, 3805, 10605, 11444, 11699, 12257, 9877, 11460, 466, 12237, 9806, 5390, 7351, 3326, 6391, 12158, 5468, 467, ...
[ 0.9613686800003052, -5.3872575759887695, -0.21408329904079437, 0.008114282041788101, 0.4259323477745056, 0.06577388942241669, 0.49167531728744507, -0.11778409779071808, 0.1867152899503708, -3.9896702766418457, 0.6089780926704407, 0.13625968992710114, -0.2704838812351227, -0.098556034266948...
650
26
bilibili_data_13232720_BV11t411A7Ym_p47_BV11t411A7Ym_p47_m4-dialogue_0370750
[S1] I can see you don't like the name. [S2] No. [S1] No, don't get rude, Rob. Um, and that's for cubic splines and for natural splines, NS. And here we see a function of age. Here's a natural cubic spline. You can see it's got three knots, interior knots, um, and is a function of age and gives you a very nice fit, standard aerosol included.
23.2
3.098788
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p47_BV11t411A7Ym_p47_m4-dialogue_0370750.mp3
[ 6873, 8493, 10541, 4386, 9379, 10076, 6947, 9516, 12778, 12651, 9577, 9070, 10004, 12642, 9890, 8337, 7372, 6948, 3390, 7581, 9954, 12185, 9701, 11853, 3548, 5520, 8600, 12115, 11458, 9988, 11231, 8386, 11054, 8116, 9060, 5276, 8029, 10622, ...
[ 0.8558416962623596, -5.514735698699951, -0.19655582308769226, 0.0025233286432921886, 0.36884763836860657, 0.07600739598274231, 0.34734106063842773, -0.13923180103302002, 0.149865984916687, -3.9047107696533203, 0.5132555961608887, 0.0006697375793009996, -0.23148629069328308, -0.132230713963...
580
23.2
bilibili_data_13232720_BV11t411A7Ym_p48_BV11t411A7Ym_p48_m4-dialogue_0440467
[S1] It's not as, as crazy a function as you like because the roughness penalty controls how wiggly the function can be. [S2] It's pretty amazing that this has a solution at all, right? It has a nice simple solution. [S1] Exactly. [S2] And I guess the, a clue as to why it's a cubic spline is that the second derivative is in the function. [S1] Yeah. [S2] Right? And there's actually a whole family of splines, whereas if they, if you have like what, a third, how, how does this work, Trevor? If there's a, if it's a k-th derivative, then it's this, 2k plus first spline. [S1] Yeah, that's right. You get, there's a, there's a
29.44
3.170768
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p48_BV11t411A7Ym_p48_m4-dialogue_0440467.mp3
[ 2608, 3192, 9245, 6751, 7824, 848, 3092, 10316, 10612, 8059, 6813, 10840, 979, 4196, 9773, 11119, 3395, 1344, 3843, 3730, 5730, 7206, 11191, 3871, 8216, 3200, 4179, 7205, 12022, 5991, 776, 793, 1736, 6265, 12716, 475, 10740, 12251, 12741, ...
[ 0.7114732265472412, -5.306936264038086, -0.054565027356147766, 0.044458065181970596, 0.42408066987991333, 0.07584529370069504, 0.39516976475715637, -0.0958346575498581, 0.07260607928037643, -4.042118072509766, 0.4775441586971283, 0.030638597905635834, -0.1769314557313919, -0.01154092792421...
736
29.44
bilibili_data_13232720_BV11t411A7Ym_p48_BV11t411A7Ym_p48_m4-dialogue_0440468
[S1] So if you like these kind of mathematical things, it's a, it's a fun proof to go through. [S2] Yeah. And it might show up in one of the quizzes. [S1] [LAUGHS] [S2] No. Okay. A few details. As I said, they, smoothine splines avoid the knot selection issues and, and we have a single lambda that needs to be chosen. That's quite attractive, rather than having to pick a whole bunch of different knots.
24.8
3.042668
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p48_BV11t411A7Ym_p48_m4-dialogue_0440468.mp3
[ 7396, 9776, 1306, 6024, 286, 5743, 5935, 3432, 4955, 9542, 2780, 5404, 8126, 11765, 9989, 11931, 12699, 4450, 12523, 12003, 12002, 9362, 859, 12614, 1473, 9435, 4083, 7038, 8925, 4399, 8294, 10167, 12222, 7030, 8616, 4405, 8222, 12279, 6590...
[ 0.9377713799476624, -5.335350036621094, -0.21882851421833038, 0.13145607709884644, 0.5046132802963257, 0.10946130007505417, 0.41601186990737915, -0.09548170864582062, 0.1705842912197113, -3.9749999046325684, 0.41485410928726196, 0.09007421880960464, -0.19586233794689178, -0.172756105661392...
620
24.8
bilibili_data_13232720_BV11t411A7Ym_p49_BV11t411A7Ym_p49_m4-dialogue_1034577
[S1] Okay. Well, so we already got a, a long list of, of, of ways of, of fitting nonlinear functions. There's another whole family of, of methods called local regression. In fact, Rob, wasn't there a PhD thesis from Stanford on, on local regression? [S2] I don't remember. [S1] Rob's PhD thesis was called Local Likelihood and was an extension of these local regression ideas to a whole range of problems and it's still used widely today.
29.44
3.184272
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p49_BV11t411A7Ym_p49_m4-dialogue_1034577.mp3
[ 48, 3673, 6810, 7279, 10676, 10465, 3675, 12129, 8693, 11252, 4583, 2005, 3852, 5035, 7594, 9651, 6515, 9452, 9843, 7339, 2115, 4635, 1133, 8558, 10664, 10600, 10482, 2797, 1180, 6179, 8810, 9834, 11411, 5738, 6690, 9363, 9345, 394, 8118, ...
[ 0.8346568942070007, -5.306182861328125, -0.13416528701782227, -0.0009276240016333759, 0.4623354971408844, -0.010295264422893524, 0.28927168250083923, -0.08604037761688232, 0.08075515180826187, -3.9960851669311523, 0.540765106678009, 0.0613103024661541, -0.18458452820777893, -0.073334388434...
736
29.44
bilibili_data_13232720_BV11t411A7Ym_p49_BV11t411A7Ym_p49_m4-dialogue_1034578
[S1] I wouldn't say wildly. [S2] You're too modest, Rob. [S1] Yes, yes.
4.32
2.904474
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p49_BV11t411A7Ym_p49_m4-dialogue_1034578.mp3
[ 8914, 3811, 7031, 9502, 2669, 5935, 5634, 6293, 2913, 466, 6959, 3276, 3178, 2666, 3383, 4406, 9637, 4685, 858, 5277, 12182, 10186, 7602, 6515, 3873, 9654, 12509, 7953, 5537, 8160, 7474, 6738, 7446, 10013, 6360, 8442, 11881, 3667, 12700, ...
[ 0.7658279538154602, -5.780812740325928, -0.13255295157432556, 0.23124703764915466, 0.3435732126235962, -0.15458254516124725, 0.4931055009365082, -0.040761686861515045, 0.14402969181537628, -3.967100143432617, 0.06876248866319656, 0.12931515276432037, -0.14535051584243774, -0.06246463209390...
108
4.32
bilibili_data_13232720_BV11t411A7Ym_p49_BV11t411A7Ym_p49_m4-dialogue_1034579
[S1] ... probably fair to say that, uh, Lowe's and cubic splines are, cubic smoothing splines are probably the two best ways of doing smoothing. And if you set the degrees of freedom of both to, to be equal, roughly, they look pretty similar too, in general. Would you agree? [S2] That's right. [S1] So, either one of these is a good choice.
14.8
3.265989
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p49_BV11t411A7Ym_p49_m4-dialogue_1034579.mp3
[ 4705, 3139, 7261, 9784, 9312, 5115, 8092, 9515, 7510, 6512, 4084, 8164, 3781, 3047, 794, 6096, 10165, 11749, 9935, 2906, 11639, 3733, 11344, 10769, 1481, 9007, 6447, 6519, 6366, 5852, 1105, 4187, 5729, 1181, 4467, 11825, 8203, 12510, 12590,...
[ 0.6754032969474792, -5.615063190460205, -0.1575276404619217, 0.041641585528850555, 0.2533435523509979, 0.0732564777135849, 0.3471986651420593, -0.0842054933309555, 0.08349665999412537, -4.01329231262207, 0.31517651677131653, 0.15379808843135834, -0.08197581022977829, 0.00017978022515308112...
370
14.8
bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478486
[S1] One, uh, node per observation. [S2] So a very bushy tree has got high variance in. [S1] Right, it's got high variance, uh, low- [S2] It's overfitting the data. [S1] Low bias, but is overfitting and probably not gonna predict well.
10.2
3.073116
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478486.mp3
[ 4196, 8377, 1107, 2806, 11198, 12446, 5646, 10956, 4309, 9827, 9894, 8503, 3958, 4612, 3605, 10893, 11126, 10356, 5198, 6367, 8256, 5738, 7089, 2909, 3470, 6014, 4662, 6003, 5000, 2852, 1228, 3379, 4966, 7366, 4955, 10128, 3671, 6155, 6753,...
[ 1.1279462575912476, -5.603692054748535, -0.149387389421463, 0.15725596249103546, 0.2236778289079666, -0.08098039031028748, 0.7061444520950317, -0.05170690640807152, 0.14192846417427063, -3.9104833602905273, 0.5307911038398743, 0.28218019008636475, -0.263802707195282, -0.0602521076798439, ...
255
10.2
bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478487
[S1] We, we set aside one part, we fit trees of various sizes on the K-1 parts, and then evaluate the, the, the prediction error on the part we've left out. [S2] Oh, so that's clever. The whole thing's controlled by alpha, then. Alpha decides how big the tree is, and you use cross-validation just to pick alpha. [S1] Exactly, right.
17.64
3.319244
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478487.mp3
[ 9248, 9777, 620, 11598, 7636, 8161, 3025, 6252, 6755, 7269, 1196, 8414, 7491, 10712, 3548, 12284, 9499, 8108, 9012, 8672, 475, 11646, 11581, 12220, 11742, 8899, 740, 2677, 8566, 12075, 10891, 6690, 9137, 3381, 8510, 8164, 8073, 8850, 10944,...
[ 0.8524765372276306, -5.593687057495117, -0.2049711048603058, 0.11821191757917404, 0.26657626032829285, 0.016292255371809006, 0.5613898038864136, -0.1473774015903473, 0.10864093154668808, -3.916579008102417, 0.649940013885498, 0.18797168135643005, -0.26563820242881775, -0.02542770281434059,...
441
17.64
bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478488
[S1] Right. And so we, we choose the alpha and we'll see cross validation curves in a few slides, but it's going to tell us the, a good idea of the, the best value of alpha to trade off the fit with the size of the tree. Um, having chosen alpha, we then go back to the, the, the, the full tree and find the sub-tree that has the, the smallest- [S2] You mean the t grown on all the training data? [S1] Exactly. So let's see what this looks like for the, for the baseball data. So
25.88
3.325204
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478488.mp3
[ 2933, 7605, 2129, 8962, 10027, 9385, 9314, 4645, 4213, 4268, 2731, 4269, 4260, 4330, 8297, 6193, 7268, 9655, 11695, 8854, 7686, 6146, 1347, 6170, 9817, 9249, 6746, 9273, 9779, 4251, 4193, 12571, 12531, 9371, 4709, 4512, 9152, 6519, 5359, ...
[ 0.920885443687439, -5.439282417297363, -0.18751153349876404, 0.051538895815610886, 0.10597837716341019, 0.0735870972275734, 0.4885239005088806, -0.12459161877632141, 0.15284748375415802, -3.840649366378784, 0.6827453374862671, 0.34218621253967285, -0.08414816111326218, 0.013563787564635277...
647
25.88
bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478489
[S1] ... the size of the, the, the term- we, we don't split a terminal node that has fewer than, say, five observations. And that gave us this tree with how many nodes? One, two, three, four, five, six, seven, eight, nine, 10, 11, 12. Right? But probably not all of these are, are, are, are predictive. [S2] Why, why are some of the arms of the tree long and some shorter? [S1] Uh, good question. [S2] Tell me why.
22.48
3.069391
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478489.mp3
[ 7201, 6192, 284, 1515, 8080, 3035, 6014, 6526, 12277, 12550, 2944, 8670, 10982, 867, 2577, 6052, 4885, 2202, 3600, 859, 9631, 4811, 528, 1434, 9639, 4382, 1096, 3161, 8736, 11306, 8753, 8402, 8745, 10323, 3593, 3076, 3082, 3097, 585, 732,...
[ 1.0840984582901, -5.479235649108887, -0.15028129518032074, 0.18303824961185455, 0.07293223589658737, 0.02909919247031212, 0.41448378562927246, -0.15275293588638306, 0.09020989388227463, -3.8257369995117188, 0.6906905174255371, 0.39410439133644104, -0.15056212246418, -0.02080998569726944, ...
562
22.48
bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478490
[S1] Not me. It's a good idea though. [S2] So it's the, whoever the author of the tree growing program in R is. [S1] Yeah. [S2] Okay. So here's the result of cross validation, which is going to give us the pruned tree. So what do we see here? Well, we see along the horizontal axis, tree size, which is the alpha parameter. So as we, as we vary alpha, it's in, it's in one-to-one correspondence with tree size, right? Uh,
19.72
2.868133
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478490.mp3
[ 9014, 9441, 7466, 9443, 11765, 5708, 9433, 4187, 8469, 7995, 9939, 6411, 6811, 10988, 11731, 12555, 9606, 5007, 6859, 9435, 12709, 12789, 11692, 12780, 12789, 7469, 6834, 5193, 2960, 1173, 9755, 8596, 8845, 3021, 11062, 6950, 12783, 4677, 3...
[ 0.9174361824989319, -5.549104690551758, -0.17424872517585754, 0.07032407820224762, 0.10980871319770813, -0.09008060395717621, 0.5175853967666626, -0.12935984134674072, 0.17279332876205444, -3.868457794189453, 0.6013780236244202, 0.24713045358657837, -0.14036910235881805, -0.015221172943711...
493
19.72
bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478491
[S1] So cross validation, on the other hand, you fit for a while, it's, it's, it's, it's, it's happy with the splits, and then it looks like it's overfitting, just increasing variance and, and not helping prediction error. Uh, the test error, which we're, we can evaluate here because we have, we've set aside a separate test set, is, it's roughly tracking this CV curve. It's, it's minimized maybe around the same place. [S2] So it looks like about three terminal nodes is, is, is good, which is the, was the first tree we showed. [S1] Exactly. So it's on the basis of this graph that we,
29.28
3.352506
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p53_BV11t411A7Ym_p53_m4-dialogue_0478491.mp3
[ 4652, 6817, 4305, 6016, 5422, 3302, 12467, 4530, 5534, 11123, 11581, 3939, 1489, 2867, 12150, 2678, 1174, 8001, 11251, 8592, 11090, 7570, 4560, 8782, 4308, 9118, 6455, 2906, 5486, 844, 10099, 10135, 12710, 7095, 11871, 3075, 4491, 11947, 98...
[ 1.0273770093917847, -5.390557765960693, -0.07588550448417664, 0.06007693335413933, 0.23089294135570526, 0.042430609464645386, 0.3303375542163849, -0.1121814176440239, 0.0659092515707016, -3.878923177719116, 0.6621416211128235, 0.3567654490470886, -0.15004445612430573, 0.02699129469692707, ...
732
29.28
bilibili_data_13232720_BV11t411A7Ym_p54_BV11t411A7Ym_p54_m4-dialogue_0511630
[S1] And so we gr- we ran the tree growing process with cross-validation, and we see what we get in the next figure. At the top, you see the full tree grown to- to all the data, and you can see it's quite a bushy tree, um, with an early split- uh, split on thal. [S2] It's actually, it's a- it's a thalium stress test. [S1] A thalium stress test, okay.
24.72
3.415736
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p54_BV11t411A7Ym_p54_m4-dialogue_0511630.mp3
[ 7209, 3641, 12619, 10081, 11876, 10223, 7606, 12726, 10174, 8725, 8847, 9748, 8385, 4059, 7323, 9387, 12002, 9322, 10939, 6689, 9779, 7265, 3113, 3802, 5968, 414, 8943, 34, 5862, 12613, 4114, 6699, 12061, 339, 5461, 11035, 4317, 9902, 9430,...
[ 0.8479945659637451, -5.591557502746582, -0.10762767493724823, 0.04984968155622482, 0.16241003572940826, -0.07132107019424438, 0.6244043707847595, -0.16326844692230225, -0.005102274939417839, -3.8283448219299316, 0.6998128294944763, 0.21003690361976624, -0.28107795119285583, 0.0191373825073...
618
24.72
bilibili_data_13232720_BV11t411A7Ym_p54_BV11t411A7Ym_p54_m4-dialogue_0511631
[S1] And, and then the two, the left and right nodes were split on, on CA. [S2] Uh, calcium, I think. [S1] Which is calcium. And, and then the subsequent splits, it's hard to see here, but you get, um, this, these pictures are in the book. So quite a bushy tree.
17.28
3.236883
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p54_BV11t411A7Ym_p54_m4-dialogue_0511631.mp3
[ 9898, 5665, 6861, 6877, 9902, 10030, 12263, 10215, 9207, 12207, 8261, 9222, 8833, 3035, 4821, 6754, 12506, 11873, 8745, 5345, 8225, 8736, 10275, 12632, 5161, 9761, 8234, 7201, 5161, 3632, 9317, 12271, 6959, 6158, 2979, 12716, 11740, 12701, ...
[ 0.9446770548820496, -5.689849376678467, -0.1394605040550232, 0.07470761984586716, 0.17014820873737335, -0.1273939162492752, 0.6611514687538147, -0.12984640896320343, 0.04443005844950676, -3.784428358078003, 0.6617788076400757, 0.18729759752750397, -0.23936647176742554, -0.01321649644523859...
432
17.28
bilibili_data_13232720_BV11t411A7Ym_p55_BV11t411A7Ym_p55_m4-dialogue_1034580
[S1] If we go over the left here, we'll see actually results of a single tree. So in this case, bagging proved a single tree maybe just by 1% error. [S2] Actually, the dotted line is a single tree, right? [S1] Okay.
9.96
3.245733
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p55_BV11t411A7Ym_p55_m4-dialogue_1034580.mp3
[ 7267, 6753, 10029, 3869, 498, 9862, 12130, 5484, 10361, 7768, 7925, 3362, 1244, 6006, 3740, 8109, 11121, 8867, 1321, 9610, 10089, 11750, 5518, 3173, 2741, 3301, 8161, 6104, 7524, 11972, 6575, 8446, 11574, 5010, 10192, 11991, 5859, 7438, 396...
[ 0.8158510327339172, -5.83002233505249, -0.1310693323612213, 0.14590029418468475, 0.28847184777259827, -0.06603812426328659, 0.5155383348464966, -0.06821521371603012, 0.18006852269172668, -3.923424482345581, 0.401822030544281, 0.318948358297348, -0.004148370120674372, -0.10983739048242569, ...
249
9.96
bilibili_data_13232720_BV11t411A7Ym_p55_BV11t411A7Ym_p55_m4-dialogue_1034581
[S1] ... training sample, if you run this, if you grow two tre- two trees, you'll get two different trees, 'cause- [S2] Exactly. [S1] ... by chance, it'll pick different variables each time. [S2] Okay, so let's see how Random Forest does in the-
11.52
2.88343
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p55_BV11t411A7Ym_p55_m4-dialogue_1034581.mp3
[ 12050, 7709, 9370, 11857, 9697, 2982, 11638, 12186, 12032, 10763, 12107, 12042, 8212, 8978, 4056, 1491, 8503, 757, 3823, 6693, 11624, 3182, 7837, 5828, 12197, 12565, 5929, 2467, 11525, 2911, 734, 738, 3550, 9150, 9891, 10837, 8899, 8503, 39...
[ 0.7948043942451477, -5.737488269805908, -0.11467139422893524, 0.17412924766540527, 0.21307280659675598, -0.13291114568710327, 0.6698658466339111, -0.09892012178897858, -0.0695539265871048, -3.943159341812134, 0.43015724420547485, 0.18539506196975708, -0.29373422265052795, -0.00556250428780...
288
11.52
bilibili_data_13232720_BV11t411A7Ym_p55_BV11t411A7Ym_p55_m4-dialogue_1034582
[S1] That's a good question. Is it, is it cheating? Is it biasing things? Well, it's not because when, when, the, the outcome, the class is not being used to, to choose the genes. It would have been a problem if we chose the genes that, that, that, that vary the most from one class to another. In other words, if we, if we actually, if, if we use the class label to choose the genes. But because we're, we're just doing this, doing this just looking at the overall variance, without regard to the class label, this is, this is fine. It's not gonna bias our results. [S2] Oh, so it's unsupervised screening. [S1] Exactly. So supervised screening is, is,
29.36
3.29197
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p55_BV11t411A7Ym_p55_m4-dialogue_1034582.mp3
[ 4124, 3121, 4245, 11629, 9005, 1568, 3344, 9078, 11117, 8980, 5988, 8483, 8509, 9313, 401, 8668, 7965, 9941, 1883, 9003, 2932, 12214, 7086, 5020, 9600, 6349, 7005, 4300, 1161, 7269, 12134, 6980, 3008, 2902, 1098, 3104, 9258, 6809, 6812, 9...
[ 0.8550264239311218, -5.467484951019287, 0.011683040298521519, -0.095976322889328, 0.3468685746192932, -0.06250601261854172, 0.418119341135025, -0.0532662533223629, 0.12642844021320343, -3.9055633544921875, 0.6367548108100891, 0.41375643014907837, -0.15103685855865479, 0.06224564090371132, ...
734
29.36
bilibili_data_13232720_BV11t411A7Ym_p55_BV11t411A7Ym_p55_m4-dialogue_1034583
[S1] Its error is, um, upwards of 50 or 60%. Now remember, there's 15 classes, so an error of 60 or 70% isn't crazy, right? Um, because there's so many classes, but it's still, it's still not very good. [S2] It's interesting, Rob, how quickly the error comes down with random forests. [S1] Right. [S2] And then sort of levels off. It doesn't take long before it, by about 100 trees, it's kind of leveled off and not really changing that much.
24.68
3.256489
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p55_BV11t411A7Ym_p55_m4-dialogue_1034583.mp3
[ 7338, 3624, 9437, 9567, 5137, 338, 1875, 11812, 9958, 10142, 11636, 8027, 8005, 2821, 276, 12127, 12071, 324, 3025, 5512, 3430, 6447, 3902, 3822, 5158, 11310, 6757, 12403, 12338, 7905, 12305, 11801, 7730, 4938, 6684, 7205, 4127, 6706, 4701,...
[ 0.8144109845161438, -5.511756896972656, -0.08123540878295898, 0.04051661118865013, 0.16892318427562714, -0.13326184451580048, 0.5033081769943237, -0.13573580980300903, -0.021090038120746613, -3.8815417289733887, 0.679031252861023, 0.325744092464447, -0.24809473752975464, 0.0034681991674005...
617
24.68
bilibili_data_13232720_BV11t411A7Ym_p56_BV11t411A7Ym_p56_m4-dialogue_0269969
[S1] So there's these two components, growing a tree to the residuals and then adding in some shrunken version of it into your current model. [S2] And that lambda's pretty small, right? It's like, we're going to see about 0.01, for example, as a value of lambda. [S1] So really shrinking-
12.8
3.230265
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p56_BV11t411A7Ym_p56_m4-dialogue_0269969.mp3
[ 7342, 6321, 6866, 848, 934, 9463, 3718, 282, 12133, 11685, 3412, 2888, 274, 9566, 12613, 848, 8754, 11728, 8618, 11019, 6739, 12203, 2987, 7923, 8734, 7202, 9200, 9662, 6447, 10599, 7310, 10950, 1271, 4163, 1362, 4033, 7449, 9834, 12452, ...
[ 0.817520022392273, -5.741697311401367, -0.04531911760568619, 0.11175303161144257, 0.09165989607572556, -0.07754986733198166, 0.633685290813446, -0.05244177207350731, 0.07573971897363663, -3.8843729496002197, 0.5377955436706543, 0.16310544312000275, -0.27975431084632874, 0.00205466337502002...
320
12.8
bilibili_data_13232720_BV11t411A7Ym_p56_BV11t411A7Ym_p56_m4-dialogue_0269970
[S1] four and eight. That might be a typical example, depending on, on the size of your data set and the number of predictors. [S2] Oh, so if, if D is one, each little tree can only involve a single variable. [S1] Right. So it's, it's actually an, an additive function of, of single variables, so there's no interactions allowed. [S2] And if D equals two, it can, uh, involve at most two variables. [S1] Right. So that's pairwise interactions. [S2] Interesting. [S1] So, uh, that's one tuning parameter. Um, the number of trees is also a tuning parameter.
29.36
3.061068
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p56_BV11t411A7Ym_p56_m4-dialogue_0269970.mp3
[ 7088, 8624, 2857, 7927, 8239, 8318, 2948, 990, 9294, 4829, 12661, 7646, 7621, 4996, 4641, 6594, 5043, 4851, 5810, 1154, 9726, 7269, 7789, 9655, 9622, 3625, 8688, 7494, 9430, 2221, 10825, 8682, 12591, 5865, 9322, 2606, 718, 3293, 1291, 321...
[ 1.1188157796859741, -5.309113502502441, -0.11649829894304276, 0.03422752395272255, 0.15896376967430115, -0.07924595475196838, 0.4064803421497345, -0.08888977020978928, 0.1327771544456482, -3.9070239067077637, 0.7717362642288208, 0.421710729598999, -0.21224138140678406, 0.019171705469489098...
734
29.36
bilibili_data_13232720_BV11t411A7Ym_p56_BV11t411A7Ym_p56_m4-dialogue_0269971
[S1] Unlike in random forest where the number of trees, you just went far enough so that you, uh, till you stopped, uh, getting the benefit of averaging. [S2] Right. I think it's still the case that, that the number of trees is not a hugely important parameter. It's possible to overfit, but it takes, I think, a very large number to typically start to cause overfitting. And here we see we're up to 5,000 and, uh, not much is really happening yet in terms of overfitting. [S1] I think it's
21.92
3.066187
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p56_BV11t411A7Ym_p56_m4-dialogue_0269971.mp3
[ 9304, 12002, 10010, 11645, 9533, 10882, 8404, 8036, 8571, 10012, 12562, 10210, 12677, 5125, 8917, 404, 12789, 12076, 5644, 5581, 11430, 5165, 9315, 5553, 509, 11003, 8490, 5524, 6455, 1809, 3904, 4436, 3300, 12183, 3652, 2924, 12382, 8709, ...
[ 0.8972695469856262, -5.506076812744141, -0.16529794037342072, 0.020497873425483704, 0.18579600751399994, -0.09574076533317566, 0.4260914921760559, -0.11952987313270569, 0.17191240191459656, -3.9317736625671387, 0.6528124809265137, 0.1972716748714447, -0.2409088909626007, -0.022415578365325...
548
21.92
bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269972
[S1] Um, x2 squared, x1 cubed, x1, x2, and so on, right? Polynomial expansions. So you can go from a p dimensional space, in this case, two, to a higher dimensional space. And the more transform variables you add, um, the more likely you are to be, to be able to get separation in this higher dimensional space. [S2] I just noticed something. The M there, of course, is not the same M as we use for margin. [S1] Oh. [S2] So that's- [S1] Good point, Rob. [S2] Maybe we could, should use a different letter.
28.84
3.087031
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269972.mp3
[ 6873, 4180, 9772, 8810, 9371, 9828, 9518, 4599, 438, 3447, 3510, 6173, 8294, 9869, 12050, 9889, 8848, 8284, 7277, 12773, 10134, 12002, 5090, 9168, 4057, 8905, 9193, 11093, 10501, 8455, 4001, 9168, 7633, 11875, 3820, 8038, 11732, 11669, 1167...
[ 1.0339291095733643, -5.217985153198242, -0.09255531430244446, 0.08393412083387375, 0.47195538878440857, 0.030114589259028435, 0.21172374486923218, -0.11038171499967575, 0.09311709553003311, -4.040534973144531, 0.681945264339447, 0.22290050983428955, -0.2682904601097107, -0.1135675460100174...
721
28.84
bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269973
[S1] Okay. Thank you. That was a bad choice of, uh, of letter. [S2] That's usually the kind of mistake I make. [S1] [LAUGHS] [S2] I'm glad, I'm glad I caught you on one, finally. [S1] It's the kind of mistake I hate making.
11.72
3.378087
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269973.mp3
[ 6867, 4307, 4645, 4774, 7023, 5814, 3806, 11874, 12057, 7094, 9575, 2468, 2253, 4827, 6899, 4204, 9843, 6289, 12336, 12376, 10795, 8241, 9868, 12515, 9379, 3856, 11703, 10158, 12325, 12443, 10137, 4422, 774, 3716, 3883, 9443, 9889, 4650, 21...
[ 0.9157103896141052, -5.446868896484375, -0.21106842160224915, 0.35596439242362976, 0.3724381923675537, -0.06848408281803131, 0.4026464819908142, -0.17611758410930634, 0.11694686859846115, -3.966416120529175, 0.20997051894664764, 0.16684238612651825, -0.09391149878501892, -0.127443790435791...
293
11.72
bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269974
[S1] That's a good point, Robin. You know, and when you think of it like that, you say, suppose you had a thousand points, and it ends up that there's 10 support points. You think, "Oh, great, you could have thrown away the other 990 points." Well, not really, because you had to have them all there to decide which ones would be the support points. Once you found them, you can throw them away, but it doesn't really help much in the, in the computations. [S2] Right. [S1] So it's a, it's a very interesting topic, support vector machines, and it took us a while to understand
29
3.076396
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269974.mp3
[ 12450, 6241, 3797, 12598, 3273, 2952, 9044, 9250, 6556, 10517, 8200, 6680, 12392, 6632, 5237, 7970, 9105, 8707, 6674, 3987, 2918, 5430, 6740, 10613, 12254, 7693, 9356, 11994, 9876, 12013, 9971, 12588, 9452, 12011, 8874, 4746, 11085, 8852, 8...
[ 0.8553888201713562, -5.353925704956055, -0.07234606891870499, 0.057326968759298325, 0.4868273138999939, 0.011639848351478577, 0.2364087998867035, -0.12965014576911926, -0.01141070481389761, -4.057007789611816, 0.686913251876831, 0.052813101559877396, -0.4076513946056366, -0.056013811379671...
725
29
bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269975
[S1] You'll notice when you, if you, if you went through this little example over here, you'll notice that it was a, it did give you the inner product between a degree two polynomial, but there were coefficients in front of these, and those are, are what give you the squashing factors. [S2] So Trevor, in that example, I'm, in polynomial kernels, I could take D to be a million, right? [S1] Yes. [S2] And I have a huge number of polynomial functions, which if I did with the, the feature expansion method, I would, things would get out of control. So, so what happens here? The magic- [S1] You'd run into trials.
28.32
3.218714
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269975.mp3
[ 7411, 8874, 5194, 8965, 2606, 3221, 11396, 10740, 10600, 11469, 11072, 8171, 10524, 5464, 4993, 2803, 11740, 9870, 8845, 12749, 11158, 8015, 5903, 3917, 4186, 6691, 12517, 11669, 2921, 4065, 6598, 5462, 3717, 5891, 3163, 6755, 12069, 12693, ...
[ 0.8647171258926392, -5.274172782897949, -0.06923435628414154, 0.1522042453289032, 0.3827284276485443, 0.036938730627298355, 0.2148323655128479, -0.14741067588329315, 0.0427609458565712, -4.0918192863464355, 0.6509352326393127, 0.1743156611919403, -0.25779157876968384, -0.0784628838300705, ...
708
28.32
bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269976
[S1] You'd run into trouble raising power of... [S2] Right. [S1] ... to a million. [S2] Yeah, with a polynomial kernel, I can get away with that. [S1] You can get away with it, yeah. [S2] And that's because of the squishing. [S1] Because of all the squishing down, yeah.
9.2
3.227033
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269976.mp3
[ 3235, 6155, 429, 5909, 8863, 4453, 10158, 3868, 10685, 11572, 10307, 12187, 12365, 9296, 917, 4374, 4781, 12113, 427, 10675, 11894, 7905, 7861, 5845, 3868, 404, 9135, 12693, 12109, 1344, 11605, 11924, 6155, 6265, 7000, 12158, 12734, 9006, 6...
[ 0.8422067165374756, -5.6049065589904785, -0.1260901689529419, 0.2610261142253876, 0.33991265296936035, 0.0394267700612545, 0.49866893887519836, -0.15977630019187927, -0.02759680524468422, -4.007374286651611, 0.15190492570400238, 0.19107086956501007, -0.20885729789733887, -0.072681114077568...
230
9.2
bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269977
[S1] So if gamma's really large, it's like having a small standard deviation, and, and you get much more wiggly decision boundaries. Whereas if gamma's small, the, the decision boundaries gets smoother. [S2] That's a good point. [S1] So we're gonna take this machinery in, in the next segment and, and look at a, a, an example.
25.44
3.074083
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p61_BV11t411A7Ym_p61_m4-dialogue_0269977.mp3
[ 6680, 1626, 3569, 6024, 3045, 11054, 10998, 10454, 5846, 1749, 4770, 12394, 8418, 9825, 8761, 7196, 12566, 11743, 670, 5552, 6129, 3504, 4145, 12051, 11751, 8691, 11256, 11298, 10476, 11567, 1475, 4545, 3362, 8149, 11721, 9492, 7955, 12236, ...
[ 1.1754025220870972, -5.2604146003723145, -0.08096949011087418, 0.16440004110336304, 0.2976839244365692, 0.00918581709265709, 0.12248922139406204, -0.12834425270557404, -0.03395259752869606, -3.9526047706604004, 0.5925313830375671, 0.08570299297571182, -0.32904914021492004, -0.1778408586978...
636
25.44
bilibili_data_13232720_BV11t411A7Ym_p62_BV11t411A7Ym_p62_m4-dialogue_0965731
[S1] In the left panel. [S2] In the right panel. [S1] In the right panel, we compare the linear support vector classifier, which is the red curve, to the SVM using a radial kernel, um, with different values of gamma, right? And you'll notice that it's not monotone. So when gamma is 10 to the minus 1, we seem to do really well.
25.96
3.335682
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p62_BV11t411A7Ym_p62_m4-dialogue_0965731.mp3
[ 6176, 8817, 4629, 12509, 11541, 8195, 2997, 6959, 5771, 11740, 10656, 11616, 6449, 3952, 6185, 10728, 9205, 11252, 9205, 11925, 7991, 7867, 7897, 3723, 1628, 6754, 12402, 8354, 9761, 9769, 12338, 9744, 7704, 11339, 8729, 9867, 9265, 4633, 1...
[ 1.045523762702942, -5.372611999511719, -0.1262638121843338, 0.07867614924907684, 0.4967329800128937, 0.04739270359277725, 0.2711208164691925, -0.13985179364681244, 0.05735194683074951, -3.9695961475372314, 0.7723881006240845, 0.163163959980011, -0.2498631626367569, -0.19510656595230103, ...
649
25.96
bilibili_data_13232720_BV11t411A7Ym_p62_BV11t411A7Ym_p62_m4-dialogue_0965732
[S1] Which means gamma is another tuning parameter for the support vector, uh, classifier. [S2] And just keep in mind, what they're telling us, it's not really a fair comparison on the right panel, right? 'Cause the, the, uh, gamma smaller has more complexity, so it's gonna fit more. So, and this is a bigger? [S1] Yeah. [S2] Yeah. So the, a training error comparison isn't a fair comparison. [S1] Right. So we didn't do it yet, but if we made, if we made, uh, gamma even much bigger, we'd probably get an area under the curve of one.
28
3.148697
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p62_BV11t411A7Ym_p62_m4-dialogue_0965732.mp3
[ 9451, 9378, 3234, 7381, 11017, 7632, 3099, 10602, 11072, 11736, 8724, 11524, 2984, 3392, 1097, 6282, 5802, 7187, 1569, 11410, 9189, 8692, 10745, 12073, 7779, 10676, 10670, 7006, 10070, 12189, 11614, 3465, 2898, 9655, 7693, 6054, 10676, 11812,...
[ 1.0150799751281738, -5.2941083908081055, -0.08419695496559143, 0.10166922956705093, 0.38617658615112305, 0.012597142718732357, 0.26274481415748596, -0.10929854214191437, 0.10327450186014175, -4.083386421203613, 0.5787096619606018, 0.14962397515773773, -0.2693294584751129, -0.07960398495197...
700
28
bilibili_data_13232720_BV11t411A7Ym_p67_BV11t411A7Ym_p67_m4-dialogue_0652383
[S1] ... or group them into, into, uh, uh, customers that are similar with regard to these features. And why do we want to do that? Because m- maybe if they're similar with regard to these features, then the, the, the, the kind of advertising we use for that subgroup will be, uh, important. So we use a certain kind of advertising for one subgroup, like maybe young males who have a lot of money, one subgroup. Another subgroup might be- [S2] Like me. [S1] Like, uh, yeah. [S2] Young male. [S1] Mm-hmm.
25.12
3.254528
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p67_BV11t411A7Ym_p67_m4-dialogue_0652383.mp3
[ 3112, 2592, 3758, 8093, 2829, 5675, 3940, 2900, 7898, 10513, 8282, 7400, 8200, 3027, 4087, 8285, 10348, 8220, 4132, 12442, 9746, 8402, 4891, 4629, 5722, 9738, 7699, 4717, 3243, 4268, 4200, 2656, 9428, 12110, 8205, 8578, 11102, 7894, 5390, ...
[ 1.4439276456832886, -5.194937229156494, -0.35169535875320435, -0.015131937339901924, 0.21242591738700867, -0.0388539582490921, 0.21333205699920654, -0.031528741121292114, 0.28328990936279297, -4.058526039123535, 0.9249829053878784, 0.3623881936073303, -0.33612531423568726, -0.0640606731176...
628
25.12
bilibili_data_13232720_BV11t411A7Ym_p67_BV11t411A7Ym_p67_m4-dialogue_0652384
[S1] It's, what's also interesting, Rob, is that variables that are somewhat responsible for clusters, like for example, this, this second variable- [S2] Mm-hmm. [S1] ... that we have over here, um, also tend to have high variance because, you know, if, if they separate the data in, in, in, in clusters, they, there tends to be variance. So, so there's some connection between principle components and clustering, but in, in a, in a more abstract sense.
24.12
3.343619
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p67_BV11t411A7Ym_p67_m4-dialogue_0652384.mp3
[ 7411, 48, 3650, 6366, 3136, 848, 4168, 2594, 3758, 11062, 6153, 3850, 712, 6757, 7859, 8339, 6480, 3036, 9559, 11741, 11348, 11467, 6113, 8029, 3923, 3408, 6433, 9109, 9542, 7693, 349, 358, 11062, 4124, 5548, 12695, 8982, 832, 3428, 12279...
[ 1.271203637123108, -5.234489440917969, -0.28097930550575256, 0.05641017109155655, 0.4288673996925354, 0.07404741644859314, 0.431562215089798, -0.05053527653217316, 0.01458045095205307, -3.882509469985962, 0.761026918888092, 0.20998284220695496, -0.37483662366867065, -0.06934747844934464, ...
603
24.12
bilibili_data_13232720_BV11t411A7Ym_p67_BV11t411A7Ym_p67_m4-dialogue_0652385
[S1] So, that's just the detail of the previous figure, which we talked about. Now, oh, sorry, I missed a point here, which is that it's not, this algorithm, although it gives you a, a local minimum, it's not guaranteed to give a global minimum. Why not? Well- [S2] What does that mean, Rob? A local minimum? [S1] Oh.
16
3.190581
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p67_BV11t411A7Ym_p67_m4-dialogue_0652385.mp3
[ 6185, 2577, 4521, 6560, 4433, 3904, 836, 11118, 11703, 5943, 5871, 5863, 7847, 7775, 5271, 3812, 3945, 12403, 9866, 9802, 9810, 9810, 9745, 9739, 9768, 9762, 2626, 9795, 12297, 569, 3712, 4389, 7038, 9500, 6592, 4002, 3740, 5904, 465, 309...
[ 1.4891561269760132, -5.229709625244141, -0.40847277641296387, 0.07838553190231323, 0.2801882028579712, -0.12892740964889526, 0.37949344515800476, -0.06482439488172531, 0.1667129099369049, -4.061025619506836, 0.711540699005127, 0.33921024203300476, -0.2549896240234375, -0.20889216661453247,...
400
16
bilibili_data_13232720_BV11t411A7Ym_p67_BV11t411A7Ym_p67_m4-dialogue_0652386
[S1] Remember, starting the algorithm was this random assignment of points to, to the number of clusters you're using. [S2] Right. So when we start the algorithm from different places, we get actually quite different solutions. Don't worry about the fact that the colors are chosen differently. Like this, these are gold and these are green, that the coloring is arbitrary, but the, the partitioning is quite different. Um, and we, typically we pick the lowest value. There's, let's see, this guy, well, these- [S1] Looks like we got three different solutions there.
29.28
3.336347
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p67_BV11t411A7Ym_p67_m4-dialogue_0652386.mp3
[ 12451, 4258, 1179, 3919, 9431, 7916, 10728, 11610, 5661, 8693, 9631, 3936, 6617, 4817, 9033, 11060, 10354, 8050, 6740, 6400, 9556, 6428, 6285, 8685, 10014, 12653, 10557, 7989, 5916, 6292, 9451, 2964, 8092, 8846, 282, 10678, 8854, 8295, 8807...
[ 1.493223786354065, -5.033672332763672, -0.40707483887672424, 0.06669193506240845, 0.35646751523017883, 0.019173063337802887, 0.3175804913043976, -0.014474141411483288, 0.11371232569217682, -4.091217517852783, 0.6930384635925293, 0.2727722227573395, -0.3744352161884308, -0.20504245162010193...
732
29.28
bilibili_data_13232720_BV11t411A7Ym_p67_BV11t411A7Ym_p67_m4-dialogue_0652387
[S1] ... three or four different, three different, yeah, three different solutions, right, I guess. [S2] The one, the ones you colored in red at the top all have exactly the same distance. [S1] Yeah. [S2] So they're actually all the same. The colorings are different, but as Rob said, the colorings are arbitrary. [S1] Yeah, so the, the, the, these solutions, these four panels give us the, the solution with lowest- [S2] But the, but the top left and the bottom right are actually different solutions again. [S1] Right.
22.84
3.260617
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p67_BV11t411A7Ym_p67_m4-dialogue_0652387.mp3
[ 6817, 228, 8900, 9485, 2909, 1251, 9081, 946, 7734, 7916, 6542, 3933, 6500, 8916, 4331, 3819, 11502, 3594, 7210, 12330, 9290, 7253, 9770, 12339, 4121, 9416, 5184, 1385, 3285, 11660, 9346, 11101, 3816, 6317, 1570, 4690, 12117, 8566, 4853, ...
[ 1.4519799947738647, -5.090041160583496, -0.4850539565086365, 0.06000826507806778, 0.4112876057624817, 0.025565512478351593, 0.48099085688591003, -0.08124369382858276, 0.09377256035804749, -4.009371280670166, 0.6318502426147461, 0.3731546998023987, -0.37387368083000183, -0.22812588512897491...
571
22.84
bilibili_data_13232720_BV11t411A7Ym_p73_BV11t411A7Ym_p73_m4-dialogue_0652388
[S1] Well, we have a guest here today, Dr. John Chambers. Um, John was the inventor of the S-language, which is the language that we use in, in R. John was also my department head at AT&T Bell Labs when I worked there in the, in the 80s and 90s. So, John, welcome. [S2] Okay. [S1] And- [S2] Great to be here. [S1] Thank you. And please tell us about S and how it became R. [S2] Okay.
21.44
3.31769
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p73_BV11t411A7Ym_p73_m4-dialogue_0652388.mp3
[ 6363, 246, 10149, 10546, 7795, 3283, 4257, 12379, 3180, 11099, 12228, 5091, 11259, 8494, 3060, 11181, 7316, 12115, 9190, 11233, 11681, 7593, 8656, 4370, 8392, 10060, 11676, 8487, 9481, 9184, 9126, 3653, 2947, 7031, 7159, 11233, 11088, 12121, ...
[ 0.8482228517532349, -5.570706844329834, -0.14470042288303375, -0.037337444722652435, 0.4730345904827118, -0.00782054103910923, 0.7080973982810974, -0.09211397171020508, 0.1837446689605713, -3.9912853240966797, 0.5616380572319031, -0.06683044880628586, -0.46238479018211365, -0.1261857897043...
536
21.44
bilibili_data_13232720_BV11t411A7Ym_p73_BV11t411A7Ym_p73_m4-dialogue_0652389
[S1] Uh, and meanwhile, the next event, uh, on the floor above us at Murray Hill, there were some guys in computer science research who were busy creating a new kind of operating system, which eventually was called UNIX. [S2] Oh. [S1] And in particular, in around 1978 or so, a new version of UNIX was developed that was based on the C programming language and, uh, was portable amongst different computers.
28.12
3.203431
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p73_BV11t411A7Ym_p73_m4-dialogue_0652389.mp3
[ 5672, 4130, 4406, 3966, 3958, 3959, 3446, 3454, 3454, 6006, 7030, 1811, 4634, 4644, 7203, 9910, 7533, 7574, 10070, 10078, 10087, 7334, 5645, 8710, 3593, 6099, 3892, 7346, 5155, 6815, 9895, 3940, 6258, 6201, 1045, 11036, 11648, 10696, 11401,...
[ 1.0959502458572388, -5.16337251663208, -0.1837516576051712, 0.01216049399226904, 0.5771908164024353, -0.10544531792402267, 0.3641086220741272, -0.105758897960186, 0.17785382270812988, -4.076606273651123, 0.5000042915344238, 0.07483706623315811, -0.4635505974292755, -0.2137906402349472, -...
703
28.12
bilibili_data_13232720_BV11t411A7Ym_p73_BV11t411A7Ym_p73_m4-dialogue_0652390
[S1] Not, not so similar. There, there were, uh, two books that Rick and I wrote that described that. If you looked at it and just looked at what somebody written down, say, to do a linear regression, it would look similar. [S2] Yeah. [S1] But as soon as you started to do anything non-trivial, it was very different. [S2] Yeah. [S1] And in fact, that's sort of the next part of the story. [S2] Oh. [S1] Uh, so then some other events started to take, take place. One was that in 1981, I left statistics for a few years and went to head what was called the Advanced Software Department.
29.88
3.064448
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p73_BV11t411A7Ym_p73_m4-dialogue_0652390.mp3
[ 7410, 3601, 4621, 5650, 8329, 4251, 7196, 6149, 11091, 10673, 10608, 12067, 8208, 4104, 4747, 8204, 8098, 10552, 12019, 8393, 8161, 5594, 10546, 7966, 6560, 2970, 10729, 11557, 5671, 8182, 5806, 2829, 6575, 3487, 5454, 8022, 3210, 2925, 111...
[ 0.82981276512146, -5.310061454772949, -0.08679202944040298, -0.06509188562631607, 0.6889902353286743, -0.03917717933654785, 0.4844379425048828, -0.11901712417602539, 0.14907516539096832, -4.114579200744629, 0.4882833659648895, 0.12760207056999207, -0.44902855157852173, -0.05980996787548065...
747
29.88
bilibili_data_13232720_BV11t411A7Ym_p73_BV11t411A7Ym_p73_m4-dialogue_0652392
[S1] Uh, and on that date, version 1.0 of R was, was produced. I wasn't part of R-Core at that time, but I was very friendly with them, and one of my most treasured souvenirs is serial number one of the CD-ROMs, uh, that were produced at that time for R, autographed by all the members of- [S2] Oh, fantastic. [S1] ... the R-Core group. [S2] Yeah. [S1] Very treasured momentum. [S2] Yeah. [S1] Um, so that basically, you know, in a sense, that, that was it, and the rest is kind of been history. Uh, R has really taken off and has introduced a, a number
29.88
3.222223
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p73_BV11t411A7Ym_p73_m4-dialogue_0652392.mp3
[ 7338, 5154, 6958, 6455, 9006, 5998, 8566, 7094, 4901, 11957, 7598, 7519, 8461, 8503, 5942, 7181, 8180, 10736, 12265, 11777, 3105, 7945, 12727, 12686, 12166, 9485, 16, 530, 6706, 4114, 4122, 4628, 3604, 3434, 8470, 8012, 8077, 8526, 7553, ...
[ 0.965562105178833, -5.200615406036377, -0.21554650366306305, -0.01669817790389061, 0.6898847818374634, 0.0032741473987698555, 0.4074176549911499, -0.043922070413827896, 0.16194333136081696, -4.101860523223877, 0.44739672541618347, 0.16738036274909973, -0.48718687891960144, -0.1737368404865...
747
29.88
bilibili_data_13232720_BV11t411A7Ym_p73_BV11t411A7Ym_p73_m4-dialogue_0652393
[S1] for many things now, not all, but many things now, people produce the code, they produce an R package, and instantly, everyone else can use what they've got freely and contribute their ideas to let things evolve. And that, I think, is, to my mind, the most beneficial result that's happened. [S2] Well, thank you very much, John. [S1] My pleasure. [S2] John retired from Bell Labs a few years back, and we managed to lure him to Stanford, where he's a consulting professor in the Statistics Department. Thanks very much, John. [S1] Thanks, Trevor. [S2] Bye-bye.
28.04
3.21001
24,000
audio/en/bilibili_data_13232720_BV11t411A7Ym_p73_BV11t411A7Ym_p73_m4-dialogue_0652393.mp3
[ 3099, 3496, 6128, 5493, 10587, 8142, 2893, 3219, 10854, 11429, 8805, 10971, 10736, 10672, 12202, 8915, 12169, 11658, 12685, 3536, 3562, 9583, 5014, 7381, 11934, 1284, 1344, 5644, 10158, 5103, 2412, 1779, 1331, 181, 3110, 1180, 9237, 8205, 5...
[ 1.06319260597229, -5.2677083015441895, -0.1880122870206833, 0.006033145822584629, 0.5200542211532593, -0.10651492327451706, 0.5848588943481445, -0.12604735791683197, 0.14013671875, -4.07820463180542, 0.625187337398529, 0.07565702497959137, -0.5323562026023865, -0.1259290724992752, -0.010...
701
28.04
bilibili_data_13276612_BV1Ls411z7VH_BV1Ls411z7VH_m4-dialogue_0551826
[S1] No, I don't think that's the case. Um, the thing is for us, you know, a lot of what we do is actually sometimes a lot of it is traveling. So you would be very bored sometimes, I think, on your own. [S2] Yeah. [S1] And, um, it's really good to have each other around and, and kind of buzz off each other. And even with performing, you know, it's, it's very nerve-wracking. We've performed in front of, say, 20,000 people. And to do that on your own, for me, would just be a bit too much. It's nice to rest on the other boys. [S2] Mm-hmm.
29
2.818294
24,000
audio/en/bilibili_data_13276612_BV1Ls411z7VH_BV1Ls411z7VH_m4-dialogue_0551826.mp3
[ 10026, 9313, 6723, 10516, 11186, 8950, 9142, 11771, 9922, 10651, 11121, 1620, 3840, 6122, 12629, 4636, 1218, 9150, 3901, 3923, 256, 2905, 9965, 11603, 4076, 11757, 9647, 12175, 12678, 7508, 2977, 2834, 2898, 336, 843, 6243, 8802, 12470, 964...
[ 0.3319063186645508, -5.631253719329834, 0.18909429013729095, 0.1819557547569275, 0.37553900480270386, 0.017531687393784523, 0.4338564872741699, -0.02195744588971138, 0.11426069587469101, -3.5779600143432617, 0.09836488217115402, -0.5101807713508606, -0.1362844705581665, 0.2259410321712494,...
725
29
bilibili_data_13276612_BV1Ls411z7VH_BV1Ls411z7VH_m4-dialogue_0551835
[S1] I've already been to Eiffel Tower twice. [S2] Yeah. [S1] So, and I do love it, but I would, I think I'd just go to a really nice French restaurant. [S2] Yeah. [S1] I'd actually, I've never tried, like, snails or what else is the French signature? [S2] Frog's legs? [S1] Frog's legs, yeah. Never tried either, and I would definitely try both of them. [S2] Are they good? [S1] Yeah, it's delicious. [S2] I've heard frog's legs taste a little bit like chicken.
28.8
2.913521
24,000
audio/en/bilibili_data_13276612_BV1Ls411z7VH_BV1Ls411z7VH_m4-dialogue_0551835.mp3
[ 9451, 9379, 4252, 3383, 10173, 11502, 3050, 10482, 7979, 9502, 8513, 4766, 5555, 11739, 11332, 9932, 6025, 3429, 7599, 822, 12222, 12262, 6378, 5552, 2869, 6390, 10752, 4074, 11710, 11638, 2862, 4015, 6225, 11202, 3505, 5237, 9022, 3839, 49...
[ 0.5471051335334778, -5.570855140686035, 0.13977673649787903, 0.25709372758865356, 0.32090944051742554, 0.091858871281147, 0.49615955352783203, -0.02503371238708496, 0.2251504808664322, -3.6690914630889893, 0.13411307334899902, -0.5115353465080261, -0.09080333262681961, 0.040653131902217865...
720
28.8
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043771
[S1] Elon Musk, great to see you. How are you? [S2] Good, how are you? [S1] I mean, we're here at the Texas Giga Factory the day before this thing opens. It's been pretty crazy out there. Thank you so much for making time. [S2] Very welcome. [S1] I would love you to help us kind of cast our minds, I don't know, 10, 20, maybe 30 years into the future. And
22.32
3.372397
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043771.mp3
[ 7209, 8249, 9929, 10150, 8027, 8059, 8877, 10917, 10685, 11134, 10038, 7665, 1448, 3537, 7330, 12059, 4074, 7154, 4979, 4466, 7347, 5729, 11947, 357, 11244, 4941, 10944, 4061, 6056, 8080, 986, 11613, 9478, 4932, 4485, 1429, 733, 646, 3244, ...
[ 0.6073667407035828, -5.708722114562988, -0.03913656994700432, 0.037205055356025696, 0.5569118857383728, 0.160328209400177, 0.3779972195625305, -0.07971217483282089, -0.0016239617252722383, -3.767195224761963, 0.12262852489948273, -0.3480466902256012, 0.044793106615543365, 0.152924954891204...
558
22.32
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043772
[S1] And help us try to picture what it would take to build a future that's worth getting excited about. You've often said it. [S2] Sure. [S1] The last time you spoke at ten, you said that, that was really just a big driver. It's, you know, you can talk about lots of other reasons to do the work you're doing, but fundamentally, you want to think about the future and not think that it sucks. [S2] [LAUGHS]
21.4
3.217972
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043772.mp3
[ 9769, 8881, 11430, 12222, 12147, 9298, 10837, 10827, 6145, 8217, 11818, 10785, 9904, 10251, 4841, 7730, 9310, 5036, 10227, 9019, 7779, 6697, 5480, 8049, 11562, 1513, 2000, 11986, 1466, 5549, 12132, 8960, 6566, 5649, 5744, 8682, 10162, 12694, ...
[ 0.6446990966796875, -5.723208904266357, 0.021707182750105858, 0.061629731208086014, 0.500353217124939, 0.11937430500984192, 0.4445528984069824, -0.08241081237792969, 0.048651136457920074, -3.86625599861145, 0.29886963963508606, -0.29557710886001587, 0.0007318061543628573, 0.076977625489234...
535
21.4
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043773
[S1] And then, uh, ultimately, uh, you, you, it's not really possible to make electric rockets, but you can make the propellant used in, in, in rockets, uh, using sustainable energy. [S2] Right. [S1] So, uh, ultimately we can have a fully sustainable energy economy, uh, and, um, and it's, it's those three things, solar, wind, uh, stationary battery- [S2] Right. [S1] ... Black electric vehicles. So, so then, uh, what, what are the limiting factors on progress? The limiting factor really will be, uh, battery cell production.
27.56
3.057193
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043773.mp3
[ 6688, 1613, 12661, 4535, 8933, 3734, 3941, 7534, 10159, 9951, 8270, 3990, 7274, 9839, 4535, 4470, 3446, 886, 1461, 1831, 3934, 4884, 9900, 9020, 5301, 5912, 7323, 10541, 9428, 723, 9502, 7068, 10211, 10211, 10218, 7666, 9394, 12329, 2134, ...
[ 0.7493929862976074, -5.617140769958496, 0.13285265862941742, -0.061444368213415146, 0.6967002153396606, 0.0026106340810656548, 0.31876376271247864, -0.05920577794313431, -0.2943943738937378, -3.84677791595459, 0.27589312195777893, -0.026655806228518486, -0.11612707376480103, 0.033063650131...
689
27.56
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043774
[S1] If, if, if that's the, the, the- [S2] That's the angle. [S1] ... rough, very rough numbers. And I certainly would invite others to check our calculations, 'cause they may arrive at a different, a different conclusions. But, um, in order to, uh, transition, uh, not just, um, current electricity production, but also, uh, heating, uh, and transport, uh, which roughly triples the amount of electricity that you need, uh, it amounts to approximately 300 terawatt hours of installed capacity.
29.04
3.163236
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043774.mp3
[ 8857, 11995, 9518, 8675, 12052, 3728, 9761, 11493, 11748, 6814, 2665, 8793, 12526, 10071, 2576, 5458, 12284, 12085, 5976, 3008, 792, 11670, 7492, 4249, 4122, 2697, 12693, 9159, 1804, 5720, 9258, 7201, 5256, 5014, 8588, 10739, 7065, 5480, 56...
[ 0.9280873537063599, -5.593968868255615, 0.1923885941505432, -0.05508309230208397, 0.6313707828521729, -0.0016050830017775297, 0.30270686745643616, -0.10076574981212616, -0.18370938301086426, -3.9079935550689697, 0.3008306920528412, -0.04873274639248848, -0.14459136128425598, 0.012672659009...
726
29.04
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043775
[S1] We'll probably do more than that, but yes, that's, hopefully we get there within a couple of years. [S2] Right. But I mean, that, so that is one- [S1] Point, point, point one terawatt hours. [S2] But that's still one, one hundredth of what's needed. How much of the rest of that 100 is, is Tesla planning to take on between, let's say between now and 2030, 2040, um, when, when, uh, you know, we really need to see this scale up happen.
29.8
2.949294
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043775.mp3
[ 5218, 5855, 9502, 3425, 1837, 8051, 7929, 5331, 9107, 6923, 6353, 8542, 8462, 6229, 5294, 5302, 5950, 6229, 2968, 2862, 8725, 5718, 4022, 12198, 11191, 9438, 9360, 8737, 7785, 942, 3999, 9346, 11288, 6801, 8001, 8064, 3970, 4683, 11810, 8...
[ 0.8550004959106445, -5.500391483306885, 0.10400526970624924, 0.08907725661993027, 0.539840042591095, 0.19477210938930511, 0.3818812072277069, -0.05464622378349304, -0.21581359207630157, -3.825549364089966, 0.23070943355560303, -0.08042600750923157, 0.07891116291284561, 0.15264396369457245,...
745
29.8
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043776
[S1] I mean, these are just guesses. I mean, um, so please, you know, people just shouldn't hold me to these things. It's not like this is like, uh, some, what, what does happen is I'll, I'll, I'll, I'll make some, like, you know, best guess and then people will, in five years, there'll be some jerk that writes an article, "Elon said this would happen," [S2] [LAUGHS] [S1] "and it didn't happen. He's a liar and a fool." [S2] [LAUGHS] [S1] Uh, it's very annoying when that happens. Uh, so these are just guesses. This is just a conversation. [S2] Right. [S1] Um,
27.08
3.195566
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043776.mp3
[ 4713, 4632, 9022, 11690, 3100, 8476, 9484, 8547, 12041, 6410, 3467, 4244, 9089, 10612, 3940, 5960, 8738, 9555, 12775, 11260, 11630, 4977, 6024, 1380, 1390, 798, 1218, 1299, 9526, 7254, 8557, 12678, 9415, 7749, 8782, 4611, 6176, 5281, 9291, ...
[ 0.6908318996429443, -5.663654804229736, 0.1076970249414444, -0.09993056207895279, 0.8077273368835449, 0.015157272107899189, 0.2581687867641449, -0.03788254037499428, -0.36195048689842224, -3.8354086875915527, 0.2197943478822708, -0.04587397351861, -0.005303595680743456, 0.03768381476402283...
677
27.08
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043777
[S1] that same grid probably is offering the world really low cost energy, isn't it? Compared, compared with now. [S2] Yeah. [S1] And I'm, I'm curious about, like, do, should people, are people entitled to get a little bit excited about the possibilities of that, that world?
17.56
3.180168
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043777.mp3
[ 9761, 6168, 3034, 11697, 11642, 4339, 1675, 867, 8097, 8080, 1003, 11699, 12266, 12692, 12557, 7782, 8863, 7253, 11491, 357, 2965, 11684, 11099, 9631, 2534, 2051, 2819, 5665, 8232, 7857, 11296, 8761, 11187, 3562, 5556, 10418, 6184, 2931, 83...
[ 0.731099009513855, -5.723182678222656, -0.08692564815282822, 0.1361275315284729, 0.636144757270813, 0.22857937216758728, 0.4714329242706299, -0.12136024236679077, 0.07053539901971817, -3.827573776245117, 0.3261586129665375, -0.23121477663516998, 0.007980014197528362, 0.022668959572911263, ...
439
17.56
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043778
[S1] ... use a lot of energy. But if you've got, um, a lot of s- uh, sustainable energy from wind and solar, uh, you can actually sequester carbon, so you can re- reverse the CO2- [S2] Right. [S1] ... uh, p- parts per million of the atmosphere and- and- and oceans. Um, and- and also, uh, uh, you can really have as much fresh water- fresh water as you want. Um, earth is mostly water. We should call earth water. It's 70% water by surface area. Now, most of that's sea water, but it's still, it's like, uh, we just have it on the- the bit that's land. [S2] Right. Um, and with energy, you can turn sea water into-
29.36
3.280154
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043778.mp3
[ 8736, 6683, 7244, 4243, 6733, 9867, 8518, 1293, 4138, 7301, 9541, 2895, 2513, 3794, 2782, 4029, 4308, 1254, 3436, 5044, 6284, 1758, 12744, 2445, 7193, 3242, 8542, 8917, 3952, 12108, 3157, 11938, 8043, 10608, 9021, 3138, 9444, 10094, 9916, ...
[ 0.6893990635871887, -5.598897933959961, 0.19440434873104095, -0.07934071123600006, 0.695146381855011, -0.008037255145609379, 0.34014010429382324, -0.050047583878040314, -0.18362320959568024, -3.8514580726623535, 0.27698132395744324, -0.05243401229381561, -0.09349682182073593, 0.11659811437...
734
29.36
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043780
[S1] Yes, exactly. [S2] Yeah, yeah. [S1] 'Cause like, like when you, when you burn fossil fuels, there's all, there's all these like, uh, side reactions and, and, and, and toxic gases of various kinds. Um, and like, like, like, sort of, uh, little particulates that, that are bad for your lungs. Like, there's, there's all sorts of bad things that are happening that will go away. [S2] Okay. [S1] And, and the sky will be cleaner and feel quieter. Uh, uh, fear's gonna be good.
25.12
3.141921
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043780.mp3
[ 7030, 6956, 11220, 11245, 4469, 6804, 842, 1730, 4782, 9764, 3921, 5540, 11773, 8852, 3867, 6433, 3301, 6926, 4358, 4438, 6411, 9491, 4259, 7283, 4755, 7445, 4574, 1395, 1899, 4298, 5013, 1853, 6802, 12548, 4868, 12509, 12742, 4867, 11428, ...
[ 0.7069823145866394, -5.683932781219482, 0.14095531404018402, -0.12115328013896942, 0.7438819408416748, 0.033223096281290054, 0.29890763759613037, -0.0187970120459795, -0.12170644849538803, -3.850538492202759, 0.2826741635799408, -0.10226734727621078, -0.09896780550479889, -0.01707932539284...
628
25.12
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043781
[S1] I want us to switch now t- to think a bit about artificial intelligence. And, but the segue there, let, you, you mentioned how, how annoying it is when people call you up for bad predictions in the past. So, I'm, I'm possibly gonna be, um, uh, annoying now. But, um, I, I, I'm curious about- [S2] Yeah. [S1] ... your timelines and how you predict and how come some things are so amazingly on the money and some aren't. So, when it comes to predicting-
25.96
3.378747
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043781.mp3
[ 8243, 11774, 5798, 8059, 11419, 9133, 6129, 1217, 10561, 3374, 10712, 5024, 5428, 4821, 4994, 11720, 4355, 11102, 7671, 886, 4270, 11073, 1490, 9454, 4233, 6688, 4145, 8128, 3573, 4317, 5720, 6642, 12053, 12435, 7585, 10054, 7912, 11759, 37...
[ 0.5318764448165894, -5.602531909942627, -0.03458147123456001, -0.011125009506940842, 0.6380034685134888, 0.24771365523338318, 0.4580959975719452, -0.09229160845279694, -0.026238858699798584, -3.796999216079712, 0.18486234545707703, -0.25306960940361023, 0.14873814582824707, 0.1442285925149...
649
25.96
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043782
[S1] You did almost exactly half of it. You were scoffed in 2014 because no one- [S2] Yeah. [S1] ... since Henry Ford with the Model T had, had come close to that kind of growth rate- [S2] Yes. [S1] ... for, for cars. You were scoffed and you actually hit 500,000 cars and- [S2] Yeah. [S1] ... and 510,000 or whatever produced. But five years ago,
18.04
3.12848
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043782.mp3
[ 9760, 4179, 12164, 7958, 6723, 8649, 11236, 9501, 5469, 8317, 2076, 2806, 5025, 5520, 9813, 12067, 385, 11196, 4780, 3347, 7580, 6460, 5814, 3496, 6166, 11171, 7460, 3104, 7273, 8737, 11179, 8096, 10592, 11547, 9429, 4893, 8857, 7331, 6680,...
[ 0.6875860095024109, -5.742409706115723, -0.05757270008325577, 0.08141905069351196, 0.5767093300819397, 0.09978627413511276, 0.5211342573165894, -0.09642302989959717, 0.03985483571887016, -3.812628984451294, 0.19383269548416138, -0.17524774372577667, -0.01997155323624611, 0.0930001884698867...
451
18.04
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043783
[S1] Last time you came to TED, we, um, I asked you about full self-driving and, um, you said, "Yep, this very year, where I am confident that we will have a car going from LA to New York, uh, without any intervention." [S2] Yeah, I, I don't want to blow your mind, but I'm not always right. [S1] [LAUGHS] [S2] Um... [S1] So talk, talk, what's the difference between those two? Why, why, why has full self-driving in particular been so hard to predict?
27.04
2.948246
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043783.mp3
[ 11922, 9826, 6740, 3349, 5885, 6461, 2896, 8977, 12150, 5733, 12629, 7252, 10145, 7606, 7518, 37, 11601, 1005, 4618, 4505, 10095, 9647, 9199, 7652, 10147, 10218, 4403, 3827, 8035, 12100, 4875, 10140, 4811, 12004, 10037, 9460, 9526, 9390, 43...
[ 0.5147814154624939, -5.6137590408325195, 0.07978341728448868, -0.07777956873178482, 0.7247962951660156, 0.15831027925014496, 0.4634194076061249, -0.059574611485004425, -0.028378691524267197, -3.8708577156066895, 0.11608453840017319, -0.19336123764514923, 0.05266469717025757, 0.176787883043...
676
27.04
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043784
[S1] Uh, you, you just hit a ceiling. Um, and, and, uh, uh, if you, because what happened, if you, if you were to plot the progress, the, the progress looks like a log curve. So it's like, yeah, a series of log curves. So, uh, most people don't know what a log curve is, I suppose, but it, it, it- [S2] Sure, sure, the same thing. [S1] It, it-
17.52
3.106623
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043784.mp3
[ 9769, 6292, 7294, 11701, 6006, 6454, 12279, 4371, 2113, 12492, 8967, 6356, 2193, 9942, 5455, 9353, 12674, 8055, 6552, 4040, 10090, 11749, 7364, 3055, 1900, 11752, 3472, 12630, 12758, 3238, 5613, 5085, 9814, 9892, 7275, 9835, 11809, 12322, 9...
[ 0.9003801941871643, -5.716394901275635, 0.03569096326828003, 0.02307983487844467, 0.713121771812439, -0.0676826685667038, 0.32966455817222595, -0.06428131461143494, -0.11716984212398529, -3.9330666065216064, 0.2945360839366913, 0.017508111894130707, -0.04190429300069809, -0.042179066687822...
438
17.52
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043785
[S1] It goes, it goes up sort of a, you know, sort of a fairly straight way and then it starts tailing off and, and, and, and you start- [S2] And there's a kind of ocean. [S1] ... getting diminishing returns. Uh, yeah, and, and you're like, uh-oh, this, it was trending up and now it's sort of curving over and not, and, and, and you, you start getting to these, what I call, uh, local, local maxima, uh, where, uh, you don't realize basically how dumb you were. Uh, that's, uh, and then, and then, and then it happens again. So,
26.4
3.007478
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043785.mp3
[ 8988, 11757, 4696, 3925, 8036, 7448, 11436, 11236, 4827, 9258, 12262, 11227, 4753, 9251, 12645, 12183, 9377, 7077, 10676, 10597, 9558, 5968, 3330, 9250, 12142, 12566, 11426, 6566, 10612, 10590, 3472, 899, 12534, 10682, 11059, 3178, 4531, 9176...
[ 0.6929891109466553, -5.657282829284668, 0.19952861964702606, -0.25864529609680176, 0.7104342579841614, -0.044995009899139404, 0.40915408730506897, 0.006314214784651995, -0.1617332547903061, -3.8705015182495117, 0.21045595407485962, -0.17375054955482483, -0.04313895106315613, 0.059127975255...
660
26.4
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043786
[S1] Well, I mean, admittedly, these, these, uh, may be an infamous, uh, last words, but I, I actually am confident that we will solve it this year. Uh, that we will exceed, uh, you're, you're saying, like, what, the, the probability of an accident, uh, at what point do you exceed that of the average person? [S2] Right. [S1] Um, I think we will exceed that this year. [S2] What are you seeing behind the scenes that gives you that confidence?
22.64
3.004251
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043786.mp3
[ 2803, 10602, 10545, 10413, 3403, 7276, 11582, 12275, 6236, 11683, 11714, 8772, 3349, 6802, 7276, 9893, 11175, 1820, 4234, 4380, 6493, 1235, 4437, 9699, 10212, 10212, 5042, 9504, 10268, 9959, 5869, 4892, 8426, 8250, 266, 4918, 4259, 4765, 31...
[ 0.8250789046287537, -5.6385111808776855, 0.2214396893978119, -0.08169028162956238, 0.6129928827285767, 0.017815645784139633, 0.3158624768257141, -0.07134844362735748, -0.1591619849205017, -3.880582571029663, 0.3691093325614929, -0.10543783009052277, 0.004867591429501772, -0.036250688135623...
566
22.64
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043787
[S1] ... ambiguity. But if you look at a, at a video segment of a few seconds of video, that ambiguity resolves. [S2] Mm-hmm. [S1] Um, so the, so the first thing we have to do is sort of tie all eight cameras together so they're syn- they're synchronized, so the, all, all the frames are looked at simultaneously and labeled simultaneously by, by a one person, 'cause we still need human labeling. [S2] Mm-hmm. [S1] Um,
20.88
3.26086
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043787.mp3
[ 9825, 8211, 7351, 3999, 4372, 4381, 4316, 7381, 8737, 9825, 8737, 10789, 12286, 10166, 8734, 6313, 4503, 9810, 7507, 9103, 2766, 3430, 2335, 6466, 2334, 4866, 7363, 4875, 9906, 10282, 12382, 6902, 5805, 9398, 9339, 5221, 4844, 8738, 8051, ...
[ 0.8305961489677429, -5.739602565765381, 0.16426019370555878, -0.03416817635297775, 0.6459486484527588, -0.04169900715351105, 0.3050798773765564, -0.029892249032855034, -0.2728220522403717, -3.8166110515594482, 0.3098444938659668, 0.03305738419294357, -0.008011173456907272, -0.0183098781853...
522
20.88
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043788
[S1] create, uh, an order labeling, uh, create order labeling software to amplify the efficiency of human labelers because it's quite hard to label video. It takes, in the beginning it was taking several hours to label a 10 second video clip. [S2] Mm-hmm. [S1] This is not scalable. [S2] Right.
16.8
2.999193
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043788.mp3
[ 9241, 11825, 12452, 9570, 1948, 7208, 8732, 4784, 12515, 4083, 284, 11685, 12748, 7556, 2648, 3905, 6185, 10789, 12662, 4015, 6447, 3951, 4014, 1462, 4079, 6494, 603, 5673, 11825, 8347, 9832, 11316, 12313, 12449, 8361, 11337, 9482, 9498, 87...
[ 0.7940191030502319, -5.789025783538818, 0.1584423929452896, -0.09947877377271652, 0.6015121936798096, -0.057815369218587875, 0.3352116346359253, -0.037456490099430084, -0.24811412394046783, -3.84822416305542, 0.2916850447654724, -0.04194479435682297, -0.03126722574234009, -0.02826607227325...
420
16.8
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043789
[S1] What you're saying is that, uh, that you, you think that, I mean, the, the result of this is that you're effectively giving the car a 3D model of the actual objects that are all around it. It knows what they are, and it knows how fast they are moving. And th- the remaining task is to- [S2] Yes. [S1] ... is to predict-
21.72
3.310687
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043789.mp3
[ 9945, 8365, 7923, 11626, 9984, 10541, 2862, 8664, 3051, 10736, 11672, 11924, 11757, 6511, 779, 5456, 2688, 810, 9015, 1820, 3088, 9777, 2824, 3446, 1399, 302, 4740, 5840, 3445, 4836, 4184, 7243, 12547, 11222, 11091, 10516, 6342, 1813, 4651,...
[ 0.7565491795539856, -5.654158592224121, 0.003932146821171045, -0.0038328724913299084, 0.5658079385757446, 0.17197653651237488, 0.4025431275367737, -0.06764201819896698, -0.04441419243812561, -3.7942159175872803, 0.2573125660419464, -0.18584032356739044, 0.09736641496419907, 0.0650795027613...
543
21.72
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043790
[S1] What the quirky behaviors are, that, that, you know, that when a pedestrian is walking down the road with a smaller pedestrian, that maybe that smaller pedestrian might do something unpredictable or, like things like that. [S2] Yeah. [S1] You have to build into it before you can really call it safe.
15.24
3.188036
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043790.mp3
[ 3676, 8723, 11827, 10794, 1062, 2740, 5236, 10549, 3829, 1064, 1538, 5557, 9314, 8809, 11612, 3570, 4019, 3322, 11117, 11181, 11183, 4836, 12194, 7656, 12612, 11525, 8297, 4013, 4583, 2461, 10161, 7606, 10151, 9551, 11598, 6413, 2856, 4877, ...
[ 0.6911190152168274, -5.845350742340088, -0.0500124916434288, -0.001360435038805008, 0.4613824486732483, 0.19993756711483002, 0.38149961829185486, -0.09540877491235733, -0.10839571803808212, -3.80499005317688, 0.23843087255954742, -0.25842827558517456, 0.10419561713933945, 0.107600413262844...
381
15.24
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043791
[S1] So you have to say, how much are you going to try to remember? Um, and, but like if, it's very common for things to be occluded. So like if you talk about, say, a, a pedestrian walking past a, a truck where you saw the p- pedestrian, um, st- start on one side of the truck, then they, then they're occluded by the truck. [S2] Right. [S1] Uh, you ne- but, but you need to think. You, you would know intuitively, okay, this, that pedestrian's gonna pop out the other side most likely. [S2] Right. [S1] And, and, and- [S2] The computer doesn't know- [S1] So you need to slow down. [S2] I mean, it's,
27.56
3.163267
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043791.mp3
[ 4142, 9891, 6938, 6632, 6080, 8559, 2847, 7428, 8654, 9519, 2923, 6552, 3996, 12247, 2412, 10235, 11646, 8502, 8228, 7990, 12006, 12753, 9153, 2279, 8893, 862, 12546, 4358, 8996, 5935, 3358, 9885, 9810, 12746, 1756, 11666, 497, 3446, 10173,...
[ 0.7971219420433044, -5.595880508422852, 0.23313090205192566, -0.07025988399982452, 0.7495640516281128, -0.020378485321998596, 0.24083641171455383, -0.04344853013753891, -0.2215234339237213, -3.833592414855957, 0.2831023931503296, -0.04950451850891113, -0.04157114401459694, 0.08590506762266...
689
27.56
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043792
[S1] ... computer doesn't know until it's full. [S2] So you need to slow down. [S1] I mean, a skeptic is going to say that every year for the last five years, you've kind of said, "Well, no, this is the year." We're confident that it, we're, we're there in a year or two or, you know, like it's, it's always been about that, that far away. But you're,
16.52
3.204012
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043792.mp3
[ 10030, 9036, 9690, 4443, 6497, 11650, 10578, 11469, 9168, 8631, 9, 11137, 11702, 10558, 897, 5549, 11373, 2123, 8534, 9207, 2927, 6106, 11668, 8837, 11156, 6477, 6580, 5807, 8612, 8694, 6608, 409, 8053, 8630, 2838, 6228, 8165, 10601, 11624,...
[ 0.6459288597106934, -5.806176662445068, -0.04257209226489067, 0.010615374892950058, 0.6425706744194031, 0.0783599242568016, 0.39803263545036316, -0.10513358563184738, -0.08909565210342407, -3.7652032375335693, 0.11100459098815918, -0.23737794160842896, 0.14539365470409393, 0.13705170154571...
413
16.52
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043793
[S1] Um, yes. I mean, the, the car currently drives me around Austin most of the time with no interventions. So it's not like, um, and, and, and, and we, we have, uh, over 100,000 people in our, uh, uh, full-stop driving beard program. Uh, so you can look at the videos that they post online. Um- [S2] I do. [S1] Okay, great. Um, and, uh- [S2] [LAUGHS]
21.04
3.290601
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043793.mp3
[ 8801, 12443, 12386, 11765, 10610, 10676, 5661, 4243, 9874, 12620, 12197, 2423, 6616, 1499, 5661, 12188, 5635, 5556, 4371, 1565, 5457, 10596, 3796, 4124, 2979, 9965, 11995, 5041, 6459, 11057, 8048, 10659, 6737, 11875, 7594, 4538, 11637, 8170, ...
[ 0.7976915240287781, -5.682247161865234, 0.08483762294054031, -0.008642852306365967, 0.6480976343154907, 0.051474444568157196, 0.32827794551849365, -0.06704969704151154, -0.23416028916835785, -3.821972131729126, 0.24007651209831238, -0.024485426023602486, -0.04391709715127945, -0.0483414232...
526
21.04
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043794
[S1] Okay, great. Um, and, uh- [S2] Some of them are great and some of them are a little terrifying. I mean, occasionally- [S1] Yeah. [S2] ... the car seems to sort of like veer off and scare the hell out of people. Um, but, um- [S1] It's still better. [S2] [LAUGHS] It's still better. But, but you, but the, the, but you're behind the scenes looking at the data. You're seeing enough improvement to, to, to believe that a, a, this year timeline is-
19.96
3.036838
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043794.mp3
[ 5921, 12260, 5541, 1942, 10004, 9197, 5468, 8157, 4564, 11245, 7443, 6304, 11695, 6118, 2927, 5614, 8934, 8869, 9123, 6636, 8694, 10070, 9732, 9052, 8044, 9061, 5542, 2990, 3567, 4380, 6818, 4453, 10712, 5602, 11437, 10475, 11178, 5842, 694...
[ 0.6241776943206787, -5.714883804321289, -0.08470361679792404, -0.04432717710733414, 0.671171247959137, 0.1458396166563034, 0.5965973138809204, -0.08605100214481354, -0.012304062023758888, -3.839994430541992, 0.03881096467375755, -0.14036723971366882, 0.06536790728569031, 0.0916976332664489...
499
19.96
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043795
[S1] drive people to be ambitious. Without that, nothing gets done. [S2] Well, I generally believe, um, in terms of internal, uh, timelines that we want to set, set the most aggressive timeline that we can, um, uh, because there's sort of like a law of gases expansion where, for schedules where whatever time you set, it's, it's not gonna be less than that. It's very rare- [S1] Right. [S2] ... that it'll be less than that. Um, but, and as far as my predictions are concerned, um, what, what tends to happen in the media is that they will report all the wrong ones and ignore all the right ones.
29.48
3.277028
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043795.mp3
[ 6689, 8746, 604, 5845, 7249, 8745, 9827, 1557, 11530, 410, 5990, 6461, 12788, 3868, 1761, 9137, 12756, 9029, 8362, 2812, 3149, 8658, 4635, 6629, 11733, 9597, 8223, 3689, 4581, 11610, 7515, 12697, 2011, 1845, 1699, 1691, 4008, 2849, 2841, ...
[ 0.755035936832428, -5.593441009521484, 0.21712225675582886, -0.11591128259897232, 0.7273277640342712, 0.019588185474276543, 0.29214367270469666, -0.0024497194681316614, -0.178649440407753, -3.8288402557373047, 0.25960758328437805, -0.07613122463226318, -0.08142326027154922, 0.0865577086806...
737
29.48
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043796
[S1] ... percent. [S2] Yes. [S1] So, uh, but they don't mention that one. Uh, so it, it, I mean, I'm not sure what my exact track record is on predictions. They're more optimistic than pessimistic, but they're not all optimistic. Um, some of them, uh, are exceeded, uh, probably more are later, um, but they, uh,
18.64
3.294664
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043796.mp3
[ 5938, 853, 1444, 12256, 3521, 10174, 7667, 2155, 2083, 11584, 7649, 9833, 10275, 9748, 7219, 4692, 2086, 4700, 3171, 7202, 4243, 11677, 9572, 4395, 6884, 6453, 886, 1965, 1893, 5028, 4316, 2324, 2259, 2323, 4811, 4747, 9893, 4685, 858, 61...
[ 0.7018117308616638, -5.806836128234863, 0.07041463255882263, -0.15964765846729279, 0.7305654883384705, 0.10342540591955185, 0.3570270538330078, -0.013491088524460793, -0.06084089353680611, -3.831345558166504, 0.24304038286209106, -0.1134483814239502, -0.07205261290073395, 0.030142901465296...
466
18.64
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043797
[S1] they, they do come true. It's very rare that they do not come true. It's sort of like, uh, you know, uh, you know, if, if, if there's some radical technology prediction, uh, the, the point is not that it was a few years late, but that it happened at all. [S2] Right. [S1] Yeah, that's the, that's the more important part. [S2] [LAUGHS]
19.24
3.078528
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043797.mp3
[ 8804, 7381, 3354, 7259, 9325, 7446, 3865, 3746, 3593, 6582, 11748, 7622, 6235, 12571, 2458, 4176, 6006, 7630, 7171, 11400, 9171, 10905, 6284, 12451, 7592, 7034, 11582, 6182, 3629, 11985, 10138, 4585, 6642, 1394, 5459, 10457, 7964, 286, 724,...
[ 0.7329627871513367, -5.751105308532715, -0.0028440768364816904, -0.05205058678984642, 0.6983093619346619, 0.0525900162756443, 0.35996320843696594, -0.0746663510799408, -0.20310543477535248, -3.8993608951568604, 0.2929672300815582, 0.009552170522511005, -0.14619578421115875, -0.032369654625...
481
19.24
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043798
[S1] ... generalize that to a robot on legs as well. The, the two hard parts, I think, like it's not, obviously companies like Boston Dynamics have shown that it's possible to make, uh, uh, quite compelling, sometimes alarming robots. [S2] Right. [S1] Um, you know, so, so this is, from a sensors and actuator standpoint, it's certainly, uh,
18.6
3.172823
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043798.mp3
[ 9243, 2077, 11523, 4564, 9647, 12086, 2908, 5853, 5222, 5550, 7550, 10198, 7502, 3344, 521, 3298, 7039, 4982, 7614, 10231, 9745, 10832, 8656, 7202, 9257, 8762, 9736, 8144, 6040, 1960, 3798, 3854, 5894, 5326, 2766, 3732, 12414, 3374, 3293, ...
[ 0.7159139513969421, -5.764488697052002, 0.0637722760438919, -0.08206551522016525, 0.6565453410148621, 0.06825077533721924, 0.41213327646255493, -0.058789998292922974, -0.22126609086990356, -3.8189892768859863, 0.20141343772411346, -0.05010933429002762, -0.05116833746433258, -0.007726229261...
465
18.6
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043799
[S1] So let's dig into exactly that. I mean, in one way, it's actually an easier problem than full self-driving because you, instead of an object going along at 60 miles an hour, which if it gets it wrong, someone will die, this is an object that's engineered to only go at, what, 3 or 4 or 5 miles an hour. [S2] Yeah, cool. Walking speed, basically. [S1] And so a mistake is that there aren't lives at stake. There might be embarrassment at stake. [S2] So much so that AI doesn't take it over and, uh, uh, [S1] Right. [S2] ... or something.
26.92
3.263236
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043799.mp3
[ 12450, 9274, 1234, 2912, 3027, 6446, 10652, 9654, 472, 4240, 8129, 11171, 12499, 11757, 12124, 10318, 9298, 8664, 4020, 11183, 8623, 1821, 9834, 9325, 11734, 7366, 11355, 4883, 11938, 9274, 9318, 9111, 6926, 11931, 3925, 6241, 6691, 12133, ...
[ 0.7232154607772827, -5.5821757316589355, 0.03833729401230812, -0.13534384965896606, 0.6419674754142761, 0.06259765475988388, 0.5113130211830139, -0.0803329348564148, 0.02053806371986866, -3.8456263542175293, 0.23835745453834534, -0.2719555199146271, 0.024395396932959557, 0.1536697596311569...
673
26.92
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043800
[S1] Right. [S2] [LAUGHS] [S1] But, um, but so, so talk about, I mean, I, I think the first applications you, you've mentioned are probably going to be manufacturing, but eventually the vision is to, to have these available for people at home. [S2] Right. [S1] If you had a robot that really understood the 3D architecture of your house and knew where every object in that house
24.16
3.119205
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043800.mp3
[ 9844, 9323, 3366, 2836, 211, 3292, 3292, 3292, 3292, 3292, 3348, 3340, 5918, 9534, 5100, 2517, 4096, 8850, 9452, 12603, 9517, 12212, 11744, 8665, 7132, 10148, 10092, 6962, 5721, 7398, 8565, 10549, 10622, 8494, 6492, 6556, 7651, 7146, 10666,...
[ 0.7378568053245544, -5.496486663818359, -0.039349425584077835, 0.12273909896612167, 0.5958004593849182, 0.15076470375061035, 0.5431943535804749, -0.07494037598371506, 0.07662826776504517, -3.8537983894348145, 0.1859104484319687, -0.1841769814491272, 0.05147889256477356, 0.07712157815694809...
604
24.16
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043802
[S1] When it could recognize, obviously recognize everyone in the home. [S2] Yeah. [S1] Could play catch with your kids.
5.32
3.189346
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043802.mp3
[ 3746, 5934, 11019, 12204, 9235, 12202, 2917, 3227, 338, 8150, 12534, 12065, 4596, 10765, 5477, 11070, 11692, 12772, 8918, 449, 2905, 2960, 3409, 9770, 7274, 12378, 11646, 3237, 3438, 848, 5456, 9621, 542, 5866, 5582, 10046, 11426, 4972, 820...
[ 0.5665543079376221, -5.911228179931641, -0.08619052171707153, 0.20366498827934265, 0.42778700590133667, -0.0007619122043251991, 0.7202128171920776, -0.15425506234169006, 0.06076851859688759, -3.90395450592041, -0.014621750451624393, -0.12294494360685349, -0.11017833650112152, 0.09490491449...
133
5.32
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043803
[S1] That, that sounds wise. [S2] And I do think there should be a regulatory agency for AI. I've said this for many years. I don't, I don't love being regulated, but I, you know, I think this is an important thing for public safety. [S1] Let, let, let's come back to that. But I'm, I'm just, I, I don't think many people have really sort of taken seriously the notion of, you know, a, a robot at home. I mean, at the start of the computing revolution, you know, Bill Gates said, "There's gonna be a computer in every home," and people at the time said, "Yeah, whatever." [S2] Right. [S1] Who, who would even want that? [S2] Yeah, now we have a computer in our pocket.
27.88
2.960619
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043803.mp3
[ 9369, 931, 10620, 6445, 520, 6178, 3728, 3572, 10619, 4189, 346, 2960, 6006, 9591, 3894, 3318, 6758, 8783, 4235, 2913, 840, 2149, 8738, 54, 8574, 6463, 4478, 5108, 2514, 4373, 1227, 3040, 337, 4265, 9882, 6186, 9491, 4252, 6818, 10022, ...
[ 0.6738752126693726, -5.541544437408447, 0.05512731894850731, -0.05073152482509613, 0.6992486119270325, 0.06984244287014008, 0.43624386191368103, -0.11263557523488998, -0.032844383269548416, -3.891608476638794, 0.18319648504257202, -0.1808444857597351, 0.01159645151346922, 0.099870793521404...
697
27.88
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043804
[S1] Do you think there will be basically like in say, say 2050 or whatever, that like a, a, a robot in most homes is, is what there will be and people will, will- [S2] Yeah, I think there probably will. [S1] ... will love them and count on them. You'll have your own butler basically. [S2] Yeah, you'll have your sort of buddy robot. Probably. Yeah. [S1] I mean, how much of a buddy? Do you, like, do you, do you, how, how many applications do you thought? Is there, you know, can you have a romantic partner, a sex partner? [S2] I mean- [S1] A lot of lonely people out there. [S2] ... it's probably inevitable.
25.6
2.990322
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043804.mp3
[ 9950, 7560, 1850, 4139, 8448, 11212, 6412, 1554, 10880, 9169, 11073, 8460, 6112, 6498, 11300, 2075, 2907, 12275, 9007, 50, 2812, 10973, 10347, 10460, 4187, 6572, 10054, 2316, 8297, 1391, 4588, 5006, 4017, 400, 1389, 2080, 11283, 7584, 5915,...
[ 0.8032376766204834, -5.506178855895996, 0.05639033019542694, -0.017369138076901436, 0.5397988557815552, 0.050932493060827255, 0.3525160253047943, -0.09354254603385925, -0.08731774985790253, -3.8623533248901367, 0.26557570695877075, -0.24951685965061188, 0.007260986603796482, 0.085451386868...
640
25.6
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043805
[S1] I mean, it's probably inevitable. I mean, I did promise the internet that I would make cat girls. We, we could make a robot cat girl. [S2] [LAUGHS] [S1] I mean, I mean, I mean, I mean, it's- [S2] Be careful what you promise the internet, you know? [LAUGHS] [S1] Yeah. Um, so, yeah, I, I guess, uh, it'll be what, what, whatever people want, really, you know? So. [S2] What, what sort of timeline should we be thinking about of the first, the first models that are actually made and sold?
27.88
2.941025
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043805.mp3
[ 2166, 9645, 2125, 5997, 6551, 3989, 6606, 4558, 6277, 3285, 6220, 7276, 9834, 12321, 8808, 6682, 9433, 9315, 3213, 5550, 8958, 3409, 5943, 3411, 4315, 5981, 6590, 9948, 3940, 8044, 7862, 5460, 10198, 6093, 4759, 6638, 9711, 8061, 5805, 43...
[ 0.6753852367401123, -5.507092475891113, 0.13324624300003052, -0.03888623043894768, 0.6888436675071716, 0.060264114290475845, 0.49461254477500916, -0.05972645431756973, -0.06061987578868866, -3.852722644805908, 0.09787905216217041, -0.12196434289216995, -0.021052522584795952, 0.109088301658...
697
27.88
bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043806
[S1] Initially just selling to businesses, or when do you picture you'll sell, you'll start selling them where you can buy your parents one for Christmas or something? [S2] I'd say less than 10 years. [S1] How, how, how, how, help me on the economics of this. So, what, what do you picture the cost of one of these being?
17.32
3.172469
24,000
audio/en/bilibili_data_1330415626_BV1Lu411k7cv_BV1Lu411k7cv_m4-dialogue_0043806.mp3
[ 6693, 9331, 9819, 10156, 12357, 11012, 11244, 9690, 12241, 3490, 11157, 10182, 8833, 12225, 10678, 4011, 3408, 6123, 12094, 7900, 10005, 12438, 11728, 4709, 6067, 9063, 3914, 403, 8215, 4381, 7968, 3400, 6951, 12126, 11619, 11547, 3348, 5464,...
[ 0.8100352883338928, -5.682348251342773, -0.08256732672452927, 0.12393768876791, 0.5888543128967285, 0.1973886936903, 0.4273076355457306, -0.08964415639638901, -0.04543713107705116, -3.764207124710083, 0.1942111998796463, -0.08611845225095749, 0.06210126355290413, 0.05074816569685936, -0....
433
17.32