-
Notifications
You must be signed in to change notification settings - Fork 0
/
subtitles.srt
520 lines (390 loc) · 12.8 KB
/
subtitles.srt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
1
00:00:00,200 --> 00:00:03,230
Hi everyone, I'm Patrick and in this video I show you how I would learn
2
00:00:03,262 --> 00:00:07,334
machine learning if I could start over. For context, I'm a machine learning developer
3
00:00:07,374 --> 00:00:10,910
advocate at AssemblyAI and before that I worked several years as
4
00:00:10,942 --> 00:00:14,854
software developer and ML engineer and I also teach Python and machine
5
00:00:14,894 --> 00:00:18,326
learning on my own YouTube channel. So I would say I'm pretty experienced
6
00:00:18,350 --> 00:00:21,750
in the field, but I know that the available courses out there
7
00:00:21,782 --> 00:00:25,078
can be overwhelming for beginners, so I hope to give you some guidance
8
00:00:25,126 --> 00:00:29,006
with this video. The demand for machine learning engineers is still increasing
9
00:00:29,110 --> 00:00:33,380
every year, so it's a great skill to have. I divided this learning path into
10
00:00:33,452 --> 00:00:36,924
seven steps that should take you about three months to finish. Of course,
11
00:00:36,964 --> 00:00:40,364
this can vary depending on how much time and effort you want to put into
12
00:00:40,404 --> 00:00:44,364
this, and I know that everyone learns differently or might have different goals.
13
00:00:44,484 --> 00:00:47,884
So this is just my personal take on how to learn machine
14
00:00:47,924 --> 00:00:51,436
learning. You can use this guide if you just want to explore machine learning as
15
00:00:51,460 --> 00:00:54,444
a hobby, but also if you plan to find a job in the field,
16
00:00:54,564 --> 00:00:57,900
I will mention a few more tips about the job search in the end.
17
00:00:58,012 --> 00:01:01,556
So lets jump into the study plan. The first thing I recommend is to lay
18
00:01:01,580 --> 00:01:04,804
the foundation with some math basics. Now you might say
19
00:01:04,844 --> 00:01:08,524
math is not really necessary anymore and this is partly true.
20
00:01:08,644 --> 00:01:12,004
The available machine learning frameworks abstract the math away
21
00:01:12,084 --> 00:01:15,588
and I know many machine learning engineers so dont need it in their day job
22
00:01:15,636 --> 00:01:19,116
at all. However, in my opinion, knowing the underlying math
23
00:01:19,180 --> 00:01:23,180
provides you with a better foundation and better understanding of how the algorithms
24
00:01:23,252 --> 00:01:26,996
work and it makes your life easier when you run into problems.
25
00:01:27,140 --> 00:01:30,212
Also, I think there is beauty in the underlying math
26
00:01:30,268 --> 00:01:33,940
that makes the machines learn. So for me, knowing the math
27
00:01:34,012 --> 00:01:37,844
sparked my excitement even more. Now you dont need to get too deep into
28
00:01:37,884 --> 00:01:41,636
this. A great website with free resources is Khan Academy.
29
00:01:41,740 --> 00:01:45,132
So my recommendation is just to take some basic courses and
30
00:01:45,148 --> 00:01:49,036
then move on. And then later when you do the actual machine learning course
31
00:01:49,100 --> 00:01:52,604
and dont understand everything, then come back here and learn the missing
32
00:01:52,644 --> 00:01:56,374
topics. Oh and by the way, you find all the resources and recommended courses
33
00:01:56,414 --> 00:01:59,766
in the description below. The next step is to learn Python. It is
34
00:01:59,790 --> 00:02:03,142
the number one programming language for machine learning and there is
35
00:02:03,198 --> 00:02:06,950
no way around it. All major machine learning frameworks are built
36
00:02:06,982 --> 00:02:10,718
with it and all major courses use Python for their exercises.
37
00:02:10,846 --> 00:02:14,110
So having decent Python skills is essential to build machine
38
00:02:14,142 --> 00:02:17,574
learning projects. Now you don't need to become an advanced software
39
00:02:17,614 --> 00:02:20,652
developer, but a little bit more than the beginner level would be great.
40
00:02:20,758 --> 00:02:24,112
One great thing about Python is that it is very beginner friendly,
41
00:02:24,168 --> 00:02:27,504
and in my opinion it's the best first programming language you can
42
00:02:27,544 --> 00:02:31,216
learn. I recommend two free courses on YouTube, one four
43
00:02:31,280 --> 00:02:34,392
hour beginner course and one six hour intermediate course,
44
00:02:34,448 --> 00:02:37,680
and then you should have a solid base. This step is what I call the
45
00:02:37,712 --> 00:02:41,792
machine learning tech stack and consists of the most important python libraries
46
00:02:41,848 --> 00:02:45,424
for machine learning, data science, and data visualization. This step
47
00:02:45,464 --> 00:02:48,896
is a bit optional because you can also pick up these skills later when you
48
00:02:48,920 --> 00:02:52,270
do the actual machine learning course, but I think it's great to build
49
00:02:52,302 --> 00:02:55,926
the foundation first and then it will be easier later. The three libraries
50
00:02:55,990 --> 00:02:59,262
I recommend at this point are numpy, which is the base for
51
00:02:59,318 --> 00:03:02,630
everything pandas, which is important for data handling,
52
00:03:02,702 --> 00:03:06,166
and matplotlib, which is needed for visualization. These libraries
53
00:03:06,230 --> 00:03:09,806
are used in almost every machine learning project. That's why I would include
54
00:03:09,870 --> 00:03:13,510
them in your learning path at this point. Again, you don't need
55
00:03:13,542 --> 00:03:16,598
to learn too much here. I recommend just following one
56
00:03:16,646 --> 00:03:20,000
free crash course for each library and then later pick
57
00:03:20,032 --> 00:03:23,552
up more advanced concepts if you need them. At this point you dont have
58
00:03:23,568 --> 00:03:27,488
to learn the machine learning courses like scikit learn, Tensorflow, or Pytorch.
59
00:03:27,576 --> 00:03:30,744
You could of course if you want, but these are included in the machine
60
00:03:30,784 --> 00:03:33,960
learning course I show you in a moment, so you can pick these up
61
00:03:33,992 --> 00:03:37,504
later. Now that weve covered the coding skills, it is finally time
62
00:03:37,544 --> 00:03:41,560
for the actual machine learning course. There are many great ones available,
63
00:03:41,672 --> 00:03:45,192
but the most popular, and in my opinion also one of the best
64
00:03:45,248 --> 00:03:48,680
ones, is the machine learning specialization by Andrew Ng on
65
00:03:48,712 --> 00:03:51,990
Coursera. This specialization includes its three courses.
66
00:03:52,062 --> 00:03:55,638
It got revamped just a few months ago and now includes Python
67
00:03:55,686 --> 00:03:58,766
with Numpy, scikit learn and Tensorflow for the code.
68
00:03:58,870 --> 00:04:02,422
So you not only learn all the essential machine learning concepts,
69
00:04:02,478 --> 00:04:05,758
but also get your first hands on experience with the ML
70
00:04:05,806 --> 00:04:09,182
libraries. It is extensive and it takes several weeks to finish,
71
00:04:09,238 --> 00:04:13,014
but it's worth it after these courses. I have one more recommendation for
72
00:04:13,054 --> 00:04:16,662
you. I suggest to implement a few algorithms from scratch
73
00:04:16,718 --> 00:04:20,056
in Python using only pure Python and numpy,
74
00:04:20,150 --> 00:04:24,092
for example by following my ML from scratch playlist here on YouTube.
75
00:04:24,148 --> 00:04:27,396
This is completely optional, but it helped me to properly understand
76
00:04:27,500 --> 00:04:31,340
some of the concepts from Andrew's course, and a lot of students have told
77
00:04:31,372 --> 00:04:34,660
me the same feedback, so check it out if you want to. Also, we plan
78
00:04:34,692 --> 00:04:38,180
to release an updated version of the ML from scratch course here
79
00:04:38,212 --> 00:04:41,396
on the AssemblyAI channel, so make sure to subscribe to
80
00:04:41,420 --> 00:04:45,108
our channel and don't miss it. Now I recommend getting even more hands on
81
00:04:45,156 --> 00:04:48,938
and learning more about data preparation. For this kaggle has
82
00:04:48,986 --> 00:04:52,066
awesome free courses on their website. I recommend at
83
00:04:52,090 --> 00:04:55,362
least the intro to ML and intermediate ML courses.
84
00:04:55,458 --> 00:04:58,826
They are lightweight compared to the previous one and some material is
85
00:04:58,850 --> 00:05:02,970
just a refresher for you, but youll learn more about data preprocessing
86
00:05:03,002 --> 00:05:06,202
and data preparation with pandas. Each lesson has a theory
87
00:05:06,258 --> 00:05:09,586
part and then some coding exercises. It also gives you a
88
00:05:09,610 --> 00:05:12,970
gentle introduction to the Kaggle platform and you learn how to make
89
00:05:13,002 --> 00:05:16,456
code submissions on Kaggle which is perfect for the next point.
90
00:05:16,560 --> 00:05:19,808
Now its time to practice as much as possible and
91
00:05:19,856 --> 00:05:23,096
apply your knowledge to real world machine learning problems.
92
00:05:23,200 --> 00:05:26,064
For this the best platform is kaggle.com.
93
00:05:26,184 --> 00:05:29,872
it provides thousands of different datasets and challenges where you
94
00:05:29,888 --> 00:05:33,240
can participate. Participating in challenges can motivate
95
00:05:33,272 --> 00:05:37,000
you a lot. Now I wouldnt try making it to the top or even winning
96
00:05:37,032 --> 00:05:40,976
prize money with this because to be honest, this requires true expertise
97
00:05:41,040 --> 00:05:44,828
and also a lot of GPU power. But I would still try
98
00:05:44,876 --> 00:05:48,132
to tweak your solutions multiple times by learning
99
00:05:48,188 --> 00:05:52,396
more about data preprocessing and also about hyperparameter tuning.
100
00:05:52,460 --> 00:05:55,820
You can then use Kaggle competitions to build your portfolio and put
101
00:05:55,852 --> 00:05:59,548
them on your cv. So in my opinion, Kaggle is an awesome
102
00:05:59,596 --> 00:06:03,412
platform and you should practice here as much as possible. Of course, you can also
103
00:06:03,468 --> 00:06:07,244
tackle other machine learning problems outside of Kaggle. It just makes your
104
00:06:07,284 --> 00:06:11,176
life a little bit easier because it provides you with the datasets, a platform
105
00:06:11,240 --> 00:06:14,736
to evaluate the projects, and there is a whole community around it.
106
00:06:14,800 --> 00:06:17,584
At this point you can already be super proud of yourself.
107
00:06:17,664 --> 00:06:21,144
And now in this last section, I want to give you a few more tips.
108
00:06:21,184 --> 00:06:24,864
If your goal is to get a job, the tasks of ML engineers vary
109
00:06:24,904 --> 00:06:27,952
a lot and it's not possible to know everything. For example,
110
00:06:28,008 --> 00:06:31,416
some positions are specialized in computer vision or NLP,
111
00:06:31,560 --> 00:06:35,448
or they require you to have experience with a specific ML framework
112
00:06:35,496 --> 00:06:38,978
or even mlops requirements like how to deploy and scale scale
113
00:06:39,026 --> 00:06:42,650
ML apps. MLops is a whole field on its own, so I may cover this
114
00:06:42,682 --> 00:06:46,306
in a separate video. My point is, you have to decide in which field you
115
00:06:46,330 --> 00:06:50,426
want to work and then look at the requirements in some corresponding job descriptions
116
00:06:50,490 --> 00:06:53,930
and then specialize in this direction. Another great tip I can
117
00:06:53,962 --> 00:06:57,754
give you, and this is something I wish I had done earlier in my career,
118
00:06:57,834 --> 00:07:01,546
is to start a blog. You can write tutorials, share what you've learned,
119
00:07:01,610 --> 00:07:05,066
which projects you have built, which problems you have faced along the way,
120
00:07:05,090 --> 00:07:08,442
and how you've solved them. By writing about a topic, you can deepen your
121
00:07:08,498 --> 00:07:11,984
knowledge and then you can use this as a resource on your cv.
122
00:07:12,104 --> 00:07:15,848
Trust me, this will increase your chances to get an interview a lot.
123
00:07:15,936 --> 00:07:19,216
Alright, that's my recommendation for a machine learning study guide.
124
00:07:19,280 --> 00:07:22,568
Again, this might not be suited for everyone. This is just how
125
00:07:22,616 --> 00:07:26,016
I would learn machine learning if I had to start over. Just one more
126
00:07:26,040 --> 00:07:29,608
quick addition. If you prefer learning with books, then you can check out these
127
00:07:29,656 --> 00:07:32,896
two books. Let me know in the comments if this was helpful,
128
00:07:32,960 --> 00:07:36,168
or if you have any other suggestions you would add to the plan.
129
00:07:36,296 --> 00:07:39,952
Don't forget to check out the resource list I put in the description below and
130
00:07:39,968 --> 00:07:41,808
then I hope to see you in the next video.