-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathMachineLearning.html
More file actions
1513 lines (1241 loc) · 143 KB
/
Copy pathMachineLearning.html
File metadata and controls
1513 lines (1241 loc) · 143 KB
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
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>ML Mastery — 90-Day Roadmap</title>
<link rel="preconnect" href="https://fonts.googleapis.com">
<link href="https://fonts.googleapis.com/css2?family=IBM+Plex+Mono:wght@400;500;600&family=IBM+Plex+Sans:wght@300;400;500;600&display=swap" rel="stylesheet">
<style>
* { box-sizing: border-box; margin: 0; padding: 0; }
:root {
--bg: #f7f6f2;
--surface: #ffffff;
--border: #d8d5cc;
--border-light: #e8e5de;
--text: #1a1a1a;
--text-dim: #6b6860;
--text-faint: #a09d96;
--accent: #1a1a1a;
--ink: #1a1a1a;
--tag-math: #2d4a3e;
--tag-ml: #3d2b1f;
--tag-dl: #2b2d4a;
--tag-eval: #4a3d2b;
--tag-unsup: #3a2b4a;
--tag-rl: #2b4a42;
--tag-nlp: #4a2b3a;
--tag-rev: #3a3a2b;
--rule: 1px solid var(--border);
}
body {
background: var(--bg);
color: var(--text);
font-family: 'IBM Plex Sans', system-ui, sans-serif;
font-size: 13px;
line-height: 1.6;
}
/* ── HEADER ── */
.header {
background: var(--ink);
color: #f7f6f2;
padding: 48px 56px 40px;
border-bottom: 3px solid var(--ink);
}
.header-eyebrow {
font-family: 'IBM Plex Mono', monospace;
font-size: 10px;
letter-spacing: 0.18em;
text-transform: uppercase;
color: #888;
margin-bottom: 14px;
}
.header-title {
font-size: 36px;
font-weight: 300;
letter-spacing: -0.02em;
line-height: 1.1;
color: #f7f6f2;
margin-bottom: 8px;
}
.header-title strong {
font-weight: 600;
}
.header-sub {
font-size: 13px;
color: #888;
font-weight: 300;
margin-bottom: 36px;
max-width: 600px;
}
.stats-row {
display: flex;
gap: 40px;
align-items: flex-end;
flex-wrap: wrap;
}
.stat {
display: flex;
flex-direction: column;
gap: 2px;
}
.stat-num {
font-family: 'IBM Plex Mono', monospace;
font-size: 24px;
font-weight: 600;
color: #f7f6f2;
line-height: 1;
}
.stat-label {
font-size: 10px;
text-transform: uppercase;
letter-spacing: 0.12em;
color: #666;
}
.progress-block {
flex: 1;
min-width: 160px;
}
.progress-bar {
height: 2px;
background: #333;
margin-bottom: 4px;
}
.progress-fill {
height: 100%;
background: #f7f6f2;
transition: width 0.3s ease;
}
.progress-pct {
font-family: 'IBM Plex Mono', monospace;
font-size: 10px;
color: #666;
text-align: right;
}
/* ── NAV STRIP ── */
.nav-strip {
background: var(--surface);
border-bottom: var(--rule);
padding: 0 56px;
display: flex;
gap: 0;
align-items: stretch;
overflow-x: auto;
}
.nav-btn {
background: none;
border: none;
border-bottom: 2px solid transparent;
padding: 14px 18px;
font-family: 'IBM Plex Mono', monospace;
font-size: 10px;
text-transform: uppercase;
letter-spacing: 0.1em;
color: var(--text-dim);
cursor: pointer;
white-space: nowrap;
transition: color 0.1s, border-color 0.1s;
}
.nav-btn:hover { color: var(--text); }
.nav-btn.active {
color: var(--text);
border-bottom-color: var(--text);
}
/* ── LEGEND ── */
.legend-strip {
padding: 12px 56px;
background: var(--bg);
border-bottom: var(--rule);
display: flex;
gap: 20px;
flex-wrap: wrap;
align-items: center;
}
.leg {
display: flex;
align-items: center;
gap: 6px;
font-size: 10px;
color: var(--text-dim);
font-family: 'IBM Plex Mono', monospace;
text-transform: uppercase;
letter-spacing: 0.06em;
}
.leg-sq {
width: 8px;
height: 8px;
flex-shrink: 0;
}
/* ── MATH GUIDE SECTION ── */
.math-guide {
margin: 40px 56px 0;
background: var(--surface);
border: var(--rule);
border-left: 3px solid var(--ink);
}
.math-guide-header {
padding: 20px 28px 16px;
border-bottom: var(--rule);
display: flex;
align-items: baseline;
gap: 16px;
}
.math-guide-title {
font-size: 14px;
font-weight: 600;
letter-spacing: 0.01em;
}
.math-guide-sub {
font-size: 11px;
color: var(--text-dim);
font-family: 'IBM Plex Mono', monospace;
}
.math-guide-body {
display: grid;
grid-template-columns: repeat(3, 1fr);
gap: 0;
}
.math-col {
padding: 24px 28px;
border-right: var(--rule);
}
.math-col:last-child { border-right: none; }
.math-col-label {
font-family: 'IBM Plex Mono', monospace;
font-size: 9px;
text-transform: uppercase;
letter-spacing: 0.14em;
color: var(--text-faint);
margin-bottom: 14px;
}
.math-topic {
margin-bottom: 16px;
}
.math-topic-name {
font-size: 12px;
font-weight: 600;
margin-bottom: 4px;
}
.math-topic-what {
font-size: 11px;
color: var(--text-dim);
line-height: 1.5;
}
.math-topic-why {
margin-top: 5px;
font-size: 10px;
font-family: 'IBM Plex Mono', monospace;
color: var(--text-faint);
padding-left: 10px;
border-left: 2px solid var(--border);
line-height: 1.4;
}
.math-note {
margin: 0 56px;
padding: 14px 0;
border-top: none;
}
/* ── COVERAGE AUDIT ── */
.audit-section {
margin: 32px 56px 0;
background: var(--surface);
border: var(--rule);
border-left: 3px solid var(--text-faint);
}
.audit-header {
padding: 16px 28px;
border-bottom: var(--rule);
display: flex;
align-items: center;
gap: 12px;
}
.audit-title {
font-size: 13px;
font-weight: 600;
}
.audit-note {
font-size: 11px;
color: var(--text-dim);
font-family: 'IBM Plex Mono', monospace;
}
.audit-body {
display: grid;
grid-template-columns: repeat(2, 1fr);
gap: 0;
}
.audit-col {
padding: 20px 28px;
border-right: var(--rule);
}
.audit-col:last-child { border-right: none; }
.audit-col-label {
font-family: 'IBM Plex Mono', monospace;
font-size: 9px;
letter-spacing: 0.14em;
text-transform: uppercase;
color: var(--text-faint);
margin-bottom: 12px;
}
.audit-item {
display: flex;
gap: 10px;
margin-bottom: 10px;
align-items: flex-start;
}
.audit-status {
font-family: 'IBM Plex Mono', monospace;
font-size: 10px;
font-weight: 600;
padding: 2px 6px;
flex-shrink: 0;
margin-top: 1px;
}
.status-yes { background: #e8f5e9; color: #2d4a3e; }
.status-gap { background: #ffeee8; color: #5a2000; }
.status-partial { background: #fdf6e0; color: #5a4000; }
.audit-item-text {
font-size: 11px;
color: var(--text-dim);
line-height: 1.4;
}
.audit-item-text strong {
color: var(--text);
font-weight: 500;
}
/* ── HOW TO LEARN ── */
.howto-section {
margin: 32px 56px 0;
background: var(--surface);
border: var(--rule);
border-left: 3px solid #2b4a42;
}
.howto-header {
padding: 16px 28px;
border-bottom: var(--rule);
}
.howto-title {
font-size: 13px;
font-weight: 600;
}
.howto-body {
display: grid;
grid-template-columns: repeat(3, 1fr);
}
.howto-col {
padding: 24px 28px;
border-right: var(--rule);
}
.howto-col:last-child { border-right: none; }
.howto-col-label {
font-family: 'IBM Plex Mono', monospace;
font-size: 9px;
letter-spacing: 0.14em;
text-transform: uppercase;
color: var(--text-faint);
margin-bottom: 14px;
}
.howto-rule {
margin-bottom: 12px;
}
.howto-rule-name {
font-size: 12px;
font-weight: 600;
margin-bottom: 3px;
}
.howto-rule-desc {
font-size: 11px;
color: var(--text-dim);
line-height: 1.5;
}
/* ── PHASE BLOCK ── */
.phase-block {
margin: 40px 56px 0;
padding: 16px 24px;
background: var(--surface);
border: var(--rule);
border-top: 2px solid var(--ink);
display: flex;
align-items: baseline;
gap: 20px;
}
.phase-id {
font-family: 'IBM Plex Mono', monospace;
font-size: 10px;
text-transform: uppercase;
letter-spacing: 0.12em;
color: var(--text-faint);
flex-shrink: 0;
}
.phase-name {
font-size: 15px;
font-weight: 600;
}
.phase-desc {
font-size: 11px;
color: var(--text-dim);
font-family: 'IBM Plex Mono', monospace;
margin-left: auto;
}
/* ── TABLE ── */
.table-wrap {
padding: 0 56px 4px;
overflow-x: auto;
}
table {
width: 100%;
border-collapse: collapse;
min-width: 860px;
margin-top: 0;
background: var(--surface);
border: var(--rule);
border-top: none;
}
thead th {
padding: 10px 16px;
text-align: left;
font-family: 'IBM Plex Mono', monospace;
font-size: 9px;
font-weight: 500;
text-transform: uppercase;
letter-spacing: 0.12em;
color: var(--text-faint);
background: var(--bg);
border-bottom: var(--rule);
border-right: var(--rule);
}
thead th:last-child { border-right: none; }
tbody tr {
border-bottom: var(--rule);
transition: background 0.08s;
}
tbody tr:hover { background: var(--bg); }
tbody tr.done { background: #f0f7f0; }
tbody tr.done td { opacity: 0.55; }
td {
padding: 16px 16px;
vertical-align: top;
font-size: 12px;
border-right: var(--rule);
}
td:last-child { border-right: none; }
.col-day { width: 60px; }
.col-topic { width: 22%; }
.col-vid { width: 34%; }
.col-paper { width: 22%; }
/* ── DAY CELL ── */
.day-wrap {
display: flex;
flex-direction: column;
align-items: center;
gap: 8px;
}
.day-num {
font-family: 'IBM Plex Mono', monospace;
font-size: 18px;
font-weight: 600;
color: var(--text-faint);
line-height: 1;
}
.day-num.done-num { color: #ccc; }
.day-check {
appearance: none;
width: 14px;
height: 14px;
border: 1px solid var(--border);
background: transparent;
cursor: pointer;
position: relative;
flex-shrink: 0;
}
.day-check:checked {
background: var(--ink);
border-color: var(--ink);
}
.day-check:checked::after {
content: '✓';
position: absolute;
top: -1px;
left: 1px;
font-size: 10px;
color: white;
font-weight: 600;
}
/* ── TOPIC CELL ── */
.track-pill {
display: inline-block;
font-family: 'IBM Plex Mono', monospace;
font-size: 8px;
font-weight: 500;
padding: 2px 7px;
text-transform: uppercase;
letter-spacing: 0.08em;
margin-bottom: 6px;
color: #f7f6f2;
}
.pill-math { background: var(--tag-math); }
.pill-ml { background: var(--tag-ml); }
.pill-dl { background: var(--tag-dl); }
.pill-eval { background: var(--tag-eval); }
.pill-unsup { background: var(--tag-unsup); }
.pill-rl { background: var(--tag-rl); }
.pill-nlp { background: var(--tag-nlp); }
.pill-rev { background: var(--tag-rev); }
.topic-name {
font-size: 13px;
font-weight: 600;
color: var(--text);
margin-bottom: 3px;
line-height: 1.3;
}
.topic-sub {
font-size: 10.5px;
color: var(--text-dim);
line-height: 1.45;
}
.impl-row {
margin-top: 8px;
display: flex;
flex-wrap: wrap;
gap: 3px;
}
.impl-tag {
font-family: 'IBM Plex Mono', monospace;
font-size: 9px;
padding: 1px 5px;
background: var(--bg);
border: var(--rule);
color: var(--text-dim);
}
/* ── VIDEO CELL ── */
.vid-list { list-style: none; display: flex; flex-direction: column; gap: 10px; }
.src-badge {
font-family: 'IBM Plex Mono', monospace;
font-size: 8px;
font-weight: 500;
padding: 2px 5px;
border: var(--rule);
color: var(--text-dim);
white-space: nowrap;
background: var(--bg);
text-transform: uppercase;
letter-spacing: 0.05em;
}
.vid-name {
font-size: 11px;
font-weight: 500;
color: var(--text);
margin-bottom: 3px;
}
.vid-target {
font-size: 10px;
color: var(--text-dim);
padding-left: 8px;
border-left: 2px solid var(--border);
line-height: 1.4;
}
.vid-target strong {
color: var(--text);
font-weight: 500;
}
.vid-header {
display: flex;
align-items: center;
gap: 6px;
margin-bottom: 2px;
}
/* ── PAPER CELL ── */
.paper-title {
font-size: 11px;
font-weight: 600;
color: var(--text);
margin-bottom: 4px;
line-height: 1.3;
}
.paper-focus {
font-size: 10px;
color: var(--text-dim);
padding-left: 8px;
border-left: 2px solid var(--border);
line-height: 1.4;
}
.no-paper {
font-size: 10px;
color: var(--text-faint);
font-style: italic;
}
/* ── MILESTONE ── */
tr.milestone td {
background: var(--ink);
color: #f7f6f2;
font-family: 'IBM Plex Mono', monospace;
font-size: 10px;
letter-spacing: 0.04em;
padding: 12px 20px;
border: none;
}
tr.milestone.ms-rl td { background: #1a2e2b; }
tr.milestone.ms-nlp td { background: #2e1a24; }
tr.milestone.ms-final td { background: var(--ink); color: #f7f6f2; font-size: 11px; line-height: 1.6; }
/* ── FINAL SUMMARY ── */
.final-bar {
margin: 40px 56px;
padding: 28px 32px;
background: var(--ink);
color: #f7f6f2;
}
.final-bar-title {
font-family: 'IBM Plex Mono', monospace;
font-size: 11px;
letter-spacing: 0.1em;
text-transform: uppercase;
color: #888;
margin-bottom: 12px;
}
.final-bar-text {
font-size: 12px;
line-height: 1.7;
color: #ccc;
}
.final-bar-text strong { color: #f7f6f2; }
@media (max-width: 900px) {
.header, .nav-strip, .legend-strip, .phase-block, .table-wrap,
.math-guide, .audit-section, .howto-section, .final-bar {
padding-left: 16px; padding-right: 16px; margin-left: 0; margin-right: 0;
}
.math-guide-body, .howto-body { grid-template-columns: 1fr; }
.audit-body { grid-template-columns: 1fr; }
.math-col, .howto-col, .audit-col { border-right: none; border-bottom: var(--rule); }
.col-vid, .col-paper { display: none; }
}
</style>
</head>
<body>
<!-- HEADER -->
<div class="header">
<div class="header-eyebrow">90-day programme · ML + Math + DL + RL + NLP</div>
<div class="header-title">Machine Learning <strong>Mastery</strong></div>
<div class="header-sub">Math foundations through production systems. Every algorithm. Every paper. Implemented from scratch.</div>
<div class="stats-row">
<div class="stat">
<div class="stat-num" id="done-count">0</div>
<div class="stat-label">Days done / 90</div>
</div>
<div class="stat">
<div class="stat-num">90+</div>
<div class="stat-label">Papers</div>
</div>
<div class="stat">
<div class="stat-num">60+</div>
<div class="stat-label">Algorithms</div>
</div>
<div class="stat">
<div class="stat-num">7</div>
<div class="stat-label">Phases</div>
</div>
<div class="progress-block">
<div class="progress-bar"><div class="progress-fill" id="progress-fill" style="width:0%"></div></div>
<div class="progress-pct" id="progress-pct">0 %</div>
</div>
</div>
</div>
<!-- NAV -->
<div class="nav-strip">
<button class="nav-btn active" onclick="filterRows('all',this)">All</button>
<button class="nav-btn" onclick="filterRows('math',this)">Math</button>
<button class="nav-btn" onclick="filterRows('ml',this)">Classical ML</button>
<button class="nav-btn" onclick="filterRows('dl',this)">Deep Learning</button>
<button class="nav-btn" onclick="filterRows('eval',this)">Eval</button>
<button class="nav-btn" onclick="filterRows('unsup',this)">Unsupervised</button>
<button class="nav-btn" onclick="filterRows('rl',this)">RL</button>
<button class="nav-btn" onclick="filterRows('nlp',this)">NLP</button>
<button class="nav-btn" onclick="filterRows('undone',this)">Remaining</button>
</div>
<!-- LEGEND -->
<div class="legend-strip">
<div class="leg"><div class="leg-sq" style="background:#2d4a3e"></div>Math</div>
<div class="leg"><div class="leg-sq" style="background:#3d2b1f"></div>Classical ML</div>
<div class="leg"><div class="leg-sq" style="background:#2b2d4a"></div>Deep Learning</div>
<div class="leg"><div class="leg-sq" style="background:#4a3d2b"></div>Eval</div>
<div class="leg"><div class="leg-sq" style="background:#3a2b4a"></div>Unsupervised</div>
<div class="leg"><div class="leg-sq" style="background:#2b4a42"></div>RL</div>
<div class="leg"><div class="leg-sq" style="background:#4a2b3a"></div>NLP</div>
<div class="leg"><div class="leg-sq" style="background:#3a3a2b"></div>Review</div>
<span style="font-family:'IBM Plex Mono',monospace;font-size:10px;color:#a09d96;margin-left:auto;">checkboxes saved in browser</span>
</div>
<!-- ═══════════════════════════════════════
MATH GUIDE SECTION (NEW)
════════════════════════════════════════════ -->
<div class="math-guide" id="math-guide-section">
<div class="math-guide-header">
<div class="math-guide-title">What to actually learn from the math sections</div>
<div class="math-guide-sub">· not what the syllabus says — what you need to carry into every algorithm</div>
</div>
<div class="math-guide-body">
<div class="math-col">
<div class="math-col-label">Linear Algebra · Days 1–8</div>
<div class="math-topic">
<div class="math-topic-name">Vectors & Matrix operations</div>
<div class="math-topic-what">Learn matrix multiplication as function composition, not row-times-column arithmetic. Know why Ax = b has 0, 1 or ∞ solutions.</div>
<div class="math-topic-why">→ Every layer of a neural net is a matrix multiply. Understanding this visually lets you debug shapes instantly.</div>
</div>
<div class="math-topic">
<div class="math-topic-name">Projections & Least squares</div>
<div class="math-topic-what">Derive P = A(AᵀA)⁻¹Aᵀ yourself. Understand that least squares is projecting y onto the column space of X.</div>
<div class="math-topic-why">→ Linear regression, PCA, attention — all fundamentally projections. If you get this, you get all three.</div>
</div>
<div class="math-topic">
<div class="math-topic-name">Eigenvalues & SVD</div>
<div class="math-topic-what">Learn SVD as three geometric operations: rotate → scale → rotate. Know Eckart-Young (best rank-k approximation). Know that PCA eigenvectors = right singular vectors of centred X.</div>
<div class="math-topic-why">→ PCA, dimensionality reduction, noise filtering, recommendation systems. SVD is everywhere.</div>
</div>
<div class="math-topic">
<div class="math-topic-name">What to actually memorise</div>
<div class="math-topic-what">Don't memorise determinant formulas. DO memorise: (1) four fundamental subspaces, (2) rank-nullity theorem, (3) projection matrix formula, (4) SVD A = UΣVᵀ, (5) PCA = eigen(XᵀX).</div>
<div class="math-topic-why">→ These five things appear constantly. Everything else you can re-derive.</div>
</div>
</div>
<div class="math-col">
<div class="math-col-label">Probability & Statistics · Days 9–11, 15–16</div>
<div class="math-topic">
<div class="math-topic-name">Bayes theorem — deeply</div>
<div class="math-topic-what">Don't just know the formula. Know why posterior = prior × likelihood / evidence. Know the difference between MAP (point estimate) and full posterior (distribution).</div>
<div class="math-topic-why">→ MAP estimate = L2-regularised MLE. Bayesian models = full posterior. This distinction matters for every regularisation decision.</div>
</div>
<div class="math-topic">
<div class="math-topic-name">MLE — derive it for 3 distributions</div>
<div class="math-topic-what">Gaussian MLE → gives you μ̂ and σ̂². Bernoulli MLE → gives you cross-entropy loss. Poisson MLE → gives you count regression. Do all three on paper.</div>
<div class="math-topic-why">→ Every loss function in ML is a negative log-likelihood. Understanding this removes all mysticism from loss choice.</div>
</div>
<div class="math-topic">
<div class="math-topic-name">KL divergence & entropy</div>
<div class="math-topic-what">Know why KL(P‖Q) ≠ KL(Q‖P). Know that cross-entropy = KL + H(P). Understand information gain = entropy reduction.</div>
<div class="math-topic-why">→ VAE loss, decision tree splits, knowledge distillation, diffusion models — all use these. One mental model covers all.</div>
</div>
<div class="math-topic">
<div class="math-topic-name">What to skip</div>
<div class="math-topic-what">Don't spend days on CDF tables or hypothesis test lookup tables. You'll use scipy for that. Focus on intuition: what is a p-value really saying, what does power mean geometrically.</div>
<div class="math-topic-why">→ You need the concepts for A/B test design decisions, not to hand-compute test statistics.</div>
</div>
</div>
<div class="math-col">
<div class="math-col-label">Optimisation · Days 12–14</div>
<div class="math-topic">
<div class="math-topic-name">Gradient descent — the geometry</div>
<div class="math-topic-what">Understand why we subtract the gradient (it points uphill, we want downhill). Know that SGD with noise can escape local minima that full-batch gradient descent cannot. Understand learning rate as step size, not a magic number.</div>
<div class="math-topic-why">→ Every training loop you ever write depends on this intuition. Debugging loss curves requires it.</div>
</div>
<div class="math-topic">
<div class="math-topic-name">Adam — derive it once, understand it forever</div>
<div class="math-topic-what">Adam = momentum (m̂) + per-parameter learning rates (v̂). Bias correction exists because m and v start at zero, making early estimates too small. Know defaults: β₁=0.9, β₂=0.999, ε=1e-8 and why they are those values.</div>
<div class="math-topic-why">→ You'll use Adam in 90% of DL projects. Knowing what it's doing lets you tune it when training is unstable.</div>
</div>
<div class="math-topic">
<div class="math-topic-name">Convexity — just enough</div>
<div class="math-topic-what">Know what a convex function is (bowl shape, any local min = global min). Know that logistic regression loss is convex but neural nets are not. Know why that matters — and why we train neural nets anyway.</div>
<div class="math-topic-why">→ Lets you reason about when gradient descent is guaranteed to work versus when it might fail.</div>
</div>
<div class="math-topic">
<div class="math-topic-name">Matrix calculus — practical focus</div>
<div class="math-topic-what">Learn numerator layout convention. Know ∂(Wx)/∂W = xᵀ, ∂(xᵀAx)/∂x = 2Ax. Practice deriving the gradient for a 2-layer net by hand — once. After that, PyTorch autograd does it.</div>
<div class="math-topic-why">→ Doing backprop once by hand builds permanent intuition. You don't need to do it every time.</div>
</div>
</div>
</div>
</div>
<!-- ═══════════════════════════════════════
COVERAGE AUDIT (NEW)
════════════════════════════════════════════ -->
<div class="audit-section">
<div class="audit-header">
<div class="audit-title">Coverage audit — what your plan covers and what it's missing</div>
<div class="audit-note">· reviewed against standard ML engineering and research curricula</div>
</div>
<div class="audit-body">
<div class="audit-col">
<div class="audit-col-label">What your plan covers well ✓</div>
<div class="audit-item"><span class="audit-status status-yes">✓</span><div class="audit-item-text"><strong>Classical ML</strong> — Linear/Ridge/Lasso/Logistic, SVM, Decision Trees, RF, GBM, XGBoost, LDA, GPs, Anomaly Detection, Time Series. Excellent breadth.</div></div>
<div class="audit-item"><span class="audit-status status-yes">✓</span><div class="audit-item-text"><strong>Deep Learning</strong> — Backprop, CNN, ViT, Word2Vec, GNN, LSTM, Attention, Transformer, VAE, Contrastive, Diffusion, RLHF, LoRA. Very thorough.</div></div>
<div class="audit-item"><span class="audit-status status-yes">✓</span><div class="audit-item-text"><strong>Evaluation</strong> — ROC-AUC, CV, Calibration, SHAP, A/B testing, SMOTE, Monitoring, Causal Inference. Better than most programmes.</div></div>
<div class="audit-item"><span class="audit-status status-yes">✓</span><div class="audit-item-text"><strong>RL</strong> — MDP through PPO through SAC. Covers value-based, policy-based, actor-critic, model-based. Solid.</div></div>
<div class="audit-item"><span class="audit-status status-yes">✓</span><div class="audit-item-text"><strong>NLP/LLMs</strong> — BPE, BERT, T5, RAG, Fine-tuning, DPO, Agents, Scaling Laws. Genuinely modern stack.</div></div>
<div class="audit-item"><span class="audit-status status-yes">✓</span><div class="audit-item-text"><strong>Math foundations</strong> — SVD, PCA, MLE, Adam, KL, Bias-Variance, Bayesian/MCMC. The right topics chosen.</div></div>
</div>
<div class="audit-col">
<div class="audit-col-label">Gaps to consider adding</div>
<div class="audit-item"><span class="audit-status status-gap">GAP</span><div class="audit-item-text"><strong>Spectral Clustering</strong> — uses graph Laplacian eigenvalues to cluster. Foundational for graph ML and image segmentation. Add a half-day after DBSCAN (Day 40).</div></div>
<div class="audit-item"><span class="audit-status status-gap">GAP</span><div class="audit-item-text"><strong>Hidden Markov Models (HMMs)</strong> — Viterbi, forward-backward algorithm. Still used in speech, bioinformatics, and NLP baselines. Add alongside time series or after LSTM.</div></div>
<div class="audit-item"><span class="audit-status status-gap">GAP</span><div class="audit-item-text"><strong>Markov Chains & Stationary Distributions</strong> — foundational for MCMC (which you do study), PageRank, RL theory. A day of explicit study would strengthen Days 16 and 69.</div></div>
<div class="audit-item"><span class="audit-status status-partial">PARTIAL</span><div class="audit-item-text"><strong>Conditional Random Fields (CRFs)</strong> — structured prediction model used in NER, sequence labelling. You cover BERT-NER but not CRF as a standalone model. Worth a short section.</div></div>
<div class="audit-item"><span class="audit-status status-partial">PARTIAL</span><div class="audit-item-text"><strong>Mean Shift Clustering</strong> — kernel-density-based, finds modes automatically. Good contrast to DBSCAN. Easy to add alongside Day 40 material.</div></div>
<div class="audit-item"><span class="audit-status status-partial">PARTIAL</span><div class="audit-item-text"><strong>Linear Programming basics</strong> — simplex method, duality. Appears in SVM duality, LP-relaxation for combinatorial problems. Your SVM section touches it but a cleaner LP framing helps.</div></div>
<div class="audit-item"><span class="audit-status status-gap">GAP</span><div class="audit-item-text"><strong>Conformal Prediction</strong> — distribution-free uncertainty quantification. Increasingly used in production ML. Worth a section near evaluation (Days 31–38).</div></div>
<div class="audit-item"><span class="audit-status status-partial">PARTIAL</span><div class="audit-item-text"><strong>Mixture of Experts (MoE)</strong> — you mention it briefly in NLP scaling laws (Day 89) but don't implement it. Given it's in GPT-4, Gemini, Mistral — worth a standalone deep dive in deep learning phase.</div></div>
</div>
</div>
</div>
<!-- ═══════════════════════════════════════
HOW TO LEARN ML (NEW)
════════════════════════════════════════════ -->
<div class="howto-section">
<div class="howto-header">
<div class="howto-title">How to actually learn ML — what to do at each stage</div>
</div>
<div class="howto-body">
<div class="howto-col">
<div class="howto-col-label">The learning loop (per topic)</div>
<div class="howto-rule">
<div class="howto-rule-name">1. Watch — don't take notes</div>
<div class="howto-rule-desc">First pass of any video: just watch. Resist the urge to pause and note everything. You're building a mental scaffold, not copying a textbook. Note only things that surprised you.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">2. Derive on paper — before code</div>
<div class="howto-rule-desc">After every video, close the screen and rederive the key equation from memory on paper. For linear regression, write the normal equations. For backprop, draw the computation graph. This is the hardest step and the most important one.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">3. Implement from scratch — no copy-paste</div>
<div class="howto-rule-desc">Write from_scratch.py with the algorithm doc closed. Use numpy, not sklearn. When you hit a shape error, that's the learning. Compare output to sklearn's version — same numbers means correct.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">4. Explain it out loud</div>
<div class="howto-rule-desc">Set a timer for 90 seconds and explain the algorithm to nobody. If you stumble on the motivation ("why does this work?"), that's the gap to fill. You don't understand something until you can explain it simply.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">5. Read the paper — after implementation</div>
<div class="howto-rule-desc">Read the original paper after you've implemented, not before. Your implementation gives you a question to bring to the paper. Without it, papers feel abstract. With it, they feel like answers.</div>
</div>
</div>
<div class="howto-col">
<div class="howto-col-label">Common traps to avoid</div>
<div class="howto-rule">
<div class="howto-rule-name">Passive watching is not studying</div>
<div class="howto-rule-desc">3B1B and StatQuest are excellent but watching them without implementing is like watching someone do push-ups. The video is the map — the code is the territory. If you haven't opened a text editor, you haven't studied yet.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">Don't get stuck on proofs forever</div>
<div class="howto-rule-desc">For each topic, there are proofs you must do once (SVD derivation, MLE for Gaussian, Bellman optimality), and proofs you can read and move on (convergence rates, Mercer's theorem). Know the difference. Spending 3 days on a convergence proof instead of implementing the algorithm is a trap.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">sklearn first is backwards</div>
<div class="howto-rule-desc">Most people learn sklearn.RandomForestClassifier() before they know what a tree is. This roadmap correctly inverts this — always implement from scratch first. Once you do, sklearn becomes a verified reference implementation, not a black box.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">Don't collect resources — use one per topic</div>
<div class="howto-rule-desc">Your plan already has good videos for everything. Don't spend the day of Day 5 finding "the best" eigenvalue video. Use what's listed, implement, move on. Resource-hunting is procrastination in disguise.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">Treat review days seriously</div>
<div class="howto-rule-desc">Days 63–68 look easy on the plan. They're not. The goal is to rederive every major result without notes. If you can't, that's a real gap — go back. These days are the difference between "studied ML" and "knows ML".</div>
</div>
</div>
<div class="howto-col">
<div class="howto-col-label">What builds actual skill</div>
<div class="howto-rule">
<div class="howto-rule-name">Build a project that breaks</div>
<div class="howto-rule-desc">Apply every algorithm to a real dataset (Kaggle, UCI, your own data). When your model underperforms, diagnose it: is it the model, the features, the hyperparameters, the evaluation? Diagnosing failures teaches more than studying successes.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">Connect algorithms across phases</div>
<div class="howto-rule-desc">When you study VAE (Day 55), connect it backwards to KL divergence (Day 10) and MLE (Day 11). When you study PPO (Day 75), connect the clipping to TRPO and to the trust region concept. Your plan has these connections implicit — make them explicit in your notes.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">Interview yourself daily</div>
<div class="howto-rule-desc">Before ending each day: "What would an interviewer ask about what I learned today?" Answer it. This is not about interview prep — it's a forcing function to identify what you understood versus what you only saw. Common questions: "Why does X work?", "What are the assumptions?", "When does it fail?"</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">The ratio: 30% theory, 70% code</div>
<div class="howto-rule-desc">A rough split that works: 30% of your time understanding why, 70% of your time implementing and debugging. If you spend more than 2 hours on a paper without writing code, you're off-balance. Papers inform code. Code tests understanding of papers.</div>
</div>
<div class="howto-rule">
<div class="howto-rule-name">Write, don't just run</div>
<div class="howto-rule-desc">After each topic, write a short (half a page) explanation as if for a blog post. This is more effective than flashcards for ML — algorithms have structure that prose can preserve but cards cannot. Your flash_cards.md on Day 67 should be the result of 67 days of short writings.</div>