Newer
Older
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Functions, Conditionals, and Iteration in Python\n",
"\n",
"Let us create a Python function, and call it from a loop."
]
},
{
"cell_type": "code",
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Hello World, x was < 10\n",
"3\n"
]
}
],
"source": [
"def HelloWorldXY(x, y):\n",
" if (x < 10):\n",
" print(\"Hello World, x was < 10\")\n",
" elif (x < 20):\n",
" print(\"Hello World, x was >= 10 but < 20\")\n",
" else:\n",
" print(\"Hello World, x was >= 20\")\n",
" return x + y\n",
"\n",
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Now let us call the function `HelloWorldXY()` from a loop:"
]
},
{
"cell_type": "code",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"--- Now running with i: 8\n",
"Hello World, x was < 10\n",
"Result from HelloWorld: 16\n",
"\n",
"--- Now running with i: 13\n",
"Hello World, x was >= 10 but < 20\n",
"Result from HelloWorld: 26\n",
"\n",
"--- Now running with i: 18\n",
"Hello World, x was >= 10 but < 20\n",
"Result from HelloWorld: 36\n",
"\n",
"--- Now running with i: 23\n",
"Hello World, x was >= 20\n",
"Result from HelloWorld: 46\n"
]
}
],
"source": [
"for i in range(8, 25, 5): # i=8, 13, 18, 23 (start, stop, step)\n",
" print(\"\\n--- Now running with i: {}\".format(i))\n",
" r = HelloWorldXY(i,i)\n",
" print(\"Result from HelloWorld: {}\".format(r))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"If you want a loop starting at 0 to 2 (exclusive) you could do any of the following:"
]
},
{
"cell_type": "code",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Iterate over the items. `range(2)` is like a list [0,1].\n",
"0\n",
"1\n",
"Iterate over an actual list.\n",
"0\n",
]
}
],
"source": [
"print(\"Iterate over the items. `range(2)` is like a list [0,1].\")\n",
"for i in range(2):\n",
" print(i)\n",
"\n",
"print(\"Iterate over an actual list.\")\n",
"for i in [0,1]:\n",
" print(i)"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"While works\n",
"0\n",
"1\n",
"Python supports standard key words like continue and break\n",
"Entered while\n",
"while broken\n"
]
}
],
"source": [
"print(\"While works\")\n",
"i = 0\n",
"while i < 2:\n",
" print(i)\n",
" i += 1\n",
" \n",
"print(\"Python supports standard key words like continue and break\")\n",
"while True:\n",
" print(\"Entered while\")\n",
" break\n",
"print(\"while broken\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## NumPy\n",
"\n",
"### Introducing NumPy\n",
"\n",
"Python is convenient, but it can also be slow. However, it does \n",
"allow you to access libraries that execute faster code written in \n",
"languages like C. NumPy is one such library: it provides fast alternatives \n",
"to math operations in Python and is designed to work efficiently with \n",
"groups of numbers - like matrices.\n",
"\n",
"NumPy is a large library and we are only going to scratch the surface \n",
"of it here. If you plan on doing much math with Python, you should \n",
"definitely spend some time exploring its documentation to learn more.\n",
"\n",
"### Importing Numpy\n",
"\n",
"When importing the NumPy library, the convention you will see \n",
"used most often - including here - is to name it `np`, like so:"
]
},
{
"cell_type": "code",
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
"execution_count": 8,
"metadata": {},
"outputs": [],
"source": [
"import numpy as np"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Now you can use the library by prefixing the names of functions and \n",
"types with `np`, which you will see in the following examples.\n",
"\n",
"### Data Types and Shapes\n",
"\n",
"The most common way to work with numbers in NumPy is through `ndarray` \n",
"objects. They are similar to Python lists, but can have any number of \n",
"dimensions. Also, `ndarray` supports fast math operations, which \n",
"is just what we want.\n",
"\n",
"Since it can store any number of dimensions, you can use `ndarrays` \n",
"to represent any of the data types : scalars, vectors, \n",
"matrices, or tensors. "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Scalars\n",
"\n",
"Scalars in NumPy are a bit more involved than in Python. Instead of \n",
"Python's basic types like `int`, `float`, etc., NumPy lets \n",
"you specify signed and unsigned types, as well as different sizes.\n",
"So instead of Python's `int`, you have access to types \n",
"like `uint8`, `int8`, `uint16`, `int16`, and so on.\n",
"\n",
"These types are important because every object you make \n",
"(vectors, matrices, tensors) eventually stores scalars. And when you \n",
"create a NumPy array, you can specify the type - _but every item in the \n",
"array must have the same type_. In this regard, NumPy arrays are more \n",
"like C arrays than Python lists.\n",
"\n",
"If you want to create a NumPy array that holds a scalar, you do so \n",
"by passing the value to NumPy's `array` function, as follows:"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array(5)"
]
},
"execution_count": 11,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# Scalar\n",
"s = np.array(5)\n",
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
"s"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"You can display the number of axes of a NumPy `array` via the `ndim` attribute;\n",
"a scalar array has $0$ axes (`ndim` == 0). The number of axes of an array is also\n",
"called its _rank_. "
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"0"
]
},
"execution_count": 12,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"s.ndim"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"You can see the shape of your arrays by checking their `shape` attribute. So if \n",
"you executed this code:"
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"()"
]
},
"execution_count": 13,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"s.shape"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"it would print out the result, an empty pair of parenthesis, `()`. This \n",
"indicates that it has zero dimensions.\n",
"\n",
"Even though scalars are inside arrays, you still use them like a normal scalar. \n",
"So you could type:"
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"numpy.int64"
]
},
"execution_count": 14,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"x = s + 3\n",
"type(x)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"and $x$ would now equal $8$. If you were to check the type of \n",
"$x$, you would find it is probably `numPy.int64`, because \n",
"it is working with NumPy types, not Python types.\n",
"\n",
"By the way, even scalar types support most of the array functions. \n",
"So you can call `x.shape` and it would return `()` because \n",
"it has zero dimensions, even though it is not an array. If you tried \n",
"that with a normal Python scalar, you would get an error."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Vectors\n",
"\n",
"To create a vector, you would pass a Python list to the `array` function, like this:"
]
},
{
"cell_type": "code",
"execution_count": 18,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([1, 2, 3])"
]
},
"execution_count": 18,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"v = np.array([1,2,3])\n",
"v"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"This vector has three entries and so is called a 3-dimensional vector. If you check a vector's `shape` attribute, it will return a single number representing the \n",
"vector's one-dimensional length. In the above example, "
]
},
{
"cell_type": "code",
"execution_count": 19,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(3,)"
]
},
"execution_count": 19,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"v.shape"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Don’t confuse a 3D vector with a 3D array. A 3D vector has only one axis and has three dimensions along its axis, whereas a 3D array has three axes (and may have any number of dimensions along each axis). "
]
},
{
"cell_type": "code",
"execution_count": 21,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"1"
]
},
"execution_count": 21,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"v.ndim"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Dimensionality can denote either the number of entries along a specific\n",
"axis (as in the case of our 3D vector) or the number of axes in an array (such as a\n",
"3D array), which can be confusing at times.\n",
"\n",
"You can access an element within the vector using indices, like this:"
]
},
{
"cell_type": "code",
"execution_count": 22,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"2"
]
},
"execution_count": 22,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"v[1]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"NumPy also supports advanced indexing techniques. For example, to access the items from the \n",
"second element onward, you would say:"
]
},
{
"cell_type": "code",
"execution_count": 23,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([2, 3])"
]
},
"execution_count": 23,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"v[1:]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"NumPy slicing is quite powerful, \n",
"allowing you to access any combination of items in an `ndarray`. But it can also be a bit complicated, \n",
"so you should read up on it in the documentation."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Matrices\n",
"\n",
"You create matrices using `NumPy`'s array function, just you did for vectors. However, instead \n",
"of just passing in a list, you need to supply a list of lists, where each list represents \n",
"a row. So to create a $3\\times 3$ matrix containing the numbers one through nine, you could \n",
"do this:"
]
},
{
"cell_type": "code",
"execution_count": 27,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[1, 2, 3],\n",
" [4, 5, 6],\n",
" [7, 8, 9]])"
]
},
"execution_count": 27,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"m = np.array([[1,2,3], [4,5,6], [7,8,9]])\n",
"m"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The entries from the first axis are called the _rows_, and the entries from \n",
"the second axis are called the _columns_. A matrix thus has two axes or _rank_ 2:"
]
},
{
"cell_type": "code",
"execution_count": 28,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"2"
]
},
"execution_count": 28,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"m.ndim"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Checking its shape attribute would return the tuple `(3, 3)` to indicate it has two dimensions, each length $3$:"
]
},
{
"cell_type": "code",
"execution_count": 25,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(3, 3)"
]
},
"execution_count": 25,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"m.shape"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"You can access elements of matrices just like vectors, but using additional index values. So to find \n",
"the number $6$ in the above matrix, you would access"
]
},
{
"cell_type": "code",
"execution_count": 29,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"6"
]
},
"execution_count": 29,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"m[1][2]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Tensors\n",
"\n",
"Tensors are just like vectors and matrices, but they can have more dimensions. For example, to \n",
"create a $3\\times 3\\times 2\\times 1$ tensor, you could do the following:"
]
},
{
"cell_type": "code",
"execution_count": 31,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[[[ 1],\n",
" [ 2]],\n",
"\n",
" [[ 3],\n",
" [ 4]],\n",
"\n",
" [[ 5],\n",
" [ 6]]],\n",
"\n",
"\n",
" [[[ 7],\n",
" [ 8]],\n",
"\n",
" [[ 9],\n",
" [10]],\n",
"\n",
" [[11],\n",
" [12]]],\n",
"\n",
"\n",
" [[[13],\n",
" [14]],\n",
"\n",
" [[15],\n",
" [16]],\n",
"\n",
" [[17],\n",
" [17]]]])"
]
},
"execution_count": 31,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"t = np.array([[[[1],[2]],[[3],[4]],[[5],[6]]],[[[7],[8]],\\\n",
" [[9],[10]],[[11],[12]]],[[[13],[14]],[[15],[16]],[[17],[17]]]])\n",
"t"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"And `t.shape` returns `(3, 3, 2, 1)` and `t.ndim` indicates that we are dealing with a rank 4 tensor."
]
},
{
"cell_type": "code",
"execution_count": 32,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(3, 3, 2, 1)"
]
},
"execution_count": 32,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"t.shape"
]
},
{
"cell_type": "code",
"execution_count": 33,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"4"
]
},
"execution_count": 33,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"t.ndim"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"You can access items just like with matrices, but with more indices. So "
]
},
{
"cell_type": "code",
"execution_count": 34,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"16"
]
},
"execution_count": 34,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"t[2][1][1][0]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
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
"### Changing Shapes\n",
"\n",
"Sometimes you will need to change the shape of your data without actually changing \n",
"its contents. For example, you may have a vector, which is one-dimensional, but need \n",
"a matrix, which is two-dimensional. There are two ways you can do that.\n",
"\n",
"Let's say you have the following vector:"
]
},
{
"cell_type": "code",
"execution_count": 36,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(4,)"
]
},
"execution_count": 36,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"v = np.array([1,2,3,4])\n",
"v.shape"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Calling `v.shape` would return `(4,)`. But what if you want a $1\\times 4$ matrix? \n",
"You can accomplish that with the `reshape` function, like so:\n"
]
},
{
"cell_type": "code",
"execution_count": 37,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[1, 2, 3, 4]])"
]
},
"execution_count": 37,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"x = v.reshape(1,4)\n",
"x"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Calling `x.shape` would return `(1,4)`. If you wanted a $4\\times 1$ matrix, you \n",
"could do this:"
]
},
{
"cell_type": "code",
"execution_count": 38,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[1],\n",
" [2],\n",
" [3],\n",
" [4]])"
]
},
"execution_count": 38,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"x = v.reshape(4,1)\n",
"x"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The `reshape` function works for more than just adding a dimension of size $1$. Check out its \n",
"documentation for more examples.\n",
"\n",
"One more thing about reshaping NumPy arrays: if you see code from experienced NumPy users, you \n",
"will often see them use a special slicing syntax instead of calling `reshape`. Using this \n",
"syntax, the previous two examples would look like this:"
]
},
{
"cell_type": "code",
"execution_count": 39,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[1, 2, 3, 4]])"
]
},
"execution_count": 39,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"x = v[None, :]\n",
"x"
]
},
{
"cell_type": "code",
"execution_count": 42,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(4, 1)"
]
},
"execution_count": 42,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"x.shape"
]
},
{
"cell_type": "code",
"execution_count": 40,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[1],\n",
" [2],\n",
" [3],\n",
" [4]])"
]
},
"execution_count": 40,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"x = v[:, None]\n",
"x"
]
},
{
"cell_type": "code",
"execution_count": 43,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(4, 1)"
]
},
"execution_count": 43,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"x.shape"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Those lines create a slice that looks at all of the items of `v` but asks NumPy to add a new dimension \n",
"of size $1$ for the associated axis. It may look strange to you now, but it's a common technique so \n",
"it's good to be aware of it. "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Element-wise Operations\n",
"\n",
"#### The Python Way\n",
"\n",
"Suppose you had a list of numbers, and you wanted to add $5$ to every item in the list. \n",
"Without NumPy, you might do something like this:"
]
},
{
"cell_type": "code",
"name": "stdout",
"output_type": "stream",
"text": [
"853 ns ± 38.3 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)\n"
]
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
"values = [1,2,3,4,5]\n",
"for i in range(len(values)):\n",
" values[i] += 5\n",
" \n",
"values"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"That makes sense, but it's a lot of code to write and it runs slowly because \n",
"it's pure Python.\n",
"\n",
"__Note:__ Just in case you aren't used to using operators like `+=`, that just \n",
"means _add these two items and then store the result in the left item._ It is a more \n",
"succinct way of writing `values[i] = values[i] + 5`. The code you see in these examples \n",
"makes use of such operators whenever possible."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### The NumPy Way\n",
"\n",
"In NumPy, we could do the following:\n"
]
},
{
"cell_type": "code",
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"3.24 µs ± 170 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)\n"
]
}
],
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Creating that array may seem odd, but normally you'll be storing your data in `ndarrays` \n",
"anyway. So if you already had an `ndarray` named `values`, you could have just done:"
]
},
{
"cell_type": "code",
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"4.07 µs ± 262 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)\n"
]
}
],
"source": [
"%%timeit\n",
"values = [1,2,3,4,5]\n",
"values = np.array(values)\n",
"values += 5\n",
"values"
]
"source": [
"We should point out, NumPy actually has functions for things like adding, multiplying, etc. \n",
"But it also supports using the standard math operators. So the following two lines are equivalent:"
]
"outputs": [
{
"data": {
"text/plain": [
"array([ 5, 10, 15, 20])"
]
},
"execution_count": 59,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"x = np.multiply(v, 5)\n",
"x"
]
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
"outputs": [
{
"data": {
"text/plain": [
"array([ 5, 10, 15, 20])"
]
},
"execution_count": 60,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"x = v * 5\n",
"x"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We will usually use the operators instead of the functions because they are more convenient to \n",
"type and easier to read, but it's really just personal preference.\n",
"\n",
"One more example of operating with scalars and `ndarrays`. Let's say you have a matrix `m` and you want \n",
"to reuse it, but first you need to set all its values to zero. Easy, just multiply by zero and assign \n",
"the result back to the matrix, like this:"
]
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[0, 0, 0],\n",
" [0, 0, 0],\n",
" [0, 0, 0]])"
]
},
"execution_count": 61,
"metadata": {},
"output_type": "execute_result"
"m *= 0\n",
"m"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Element-wise Matrix Operations\n",
"The same functions and operators that work with scalars and matrices also work with \n",
"other dimensions. You just need to make sure that the items you perform the operation \n",
"on have compatible shapes.\n",
"Let's say you want to get the squared values of a matrix. That's simply `x = m * m` (or if you \n",
"want to assign the value back to m, it's just `m *= m`\n",
"This works because it's an element-wise multiplication between two identically-shaped matrices. \n",
"(In this case, they are shaped the same because they are actually the same object.)\n",
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
"Here's another example:"
]
},
{
"cell_type": "code",
"execution_count": 68,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[1, 3],\n",
" [5, 7]])"
]
},
"execution_count": 68,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
"a"
]
},
{
"cell_type": "code",
"execution_count": 69,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[2, 4],\n",
" [6, 8]])"
]
},
"execution_count": 69,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
"b"
]
},
{
"cell_type": "code",
"execution_count": 70,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 3, 7],\n",
" [11, 15]])"
]
},
"execution_count": 70,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"a + b"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"And if you try working with incompatible shapes, you would get an error:"
]
},
{
"cell_type": "code",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[1, 3],\n",
" [5, 7]])"
]
},
"execution_count": 71,
"metadata": {},
"output_type": "execute_result"
"outputs": [
{
"data": {
"text/plain": [
"array([[2, 3, 6],\n",
" [4, 5, 9],\n",
" [1, 8, 7]])"
]
},
"execution_count": 72,
"metadata": {},
"output_type": "execute_result"
}
],
]
},
{
"cell_type": "code",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(2, 2)"
]
},
"execution_count": 73,
"metadata": {},
"output_type": "execute_result"
"outputs": [
{
"data": {
"text/plain": [
"(3, 3)"
]
},
"execution_count": 74,
"metadata": {},
"output_type": "execute_result"
}
],
]
},
{
"cell_type": "code",
"metadata": {},
"outputs": [
{
"ename": "ValueError",
"evalue": "operands could not be broadcast together with shapes (2,2) (3,3) ",
"output_type": "error",
"traceback": [
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
"\u001b[0;31mValueError\u001b[0m Traceback (most recent call last)",
"\u001b[0;32m<ipython-input-75-e81e582b6fa9>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0ma\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0mc\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
"\u001b[0;31mValueError\u001b[0m: operands could not be broadcast together with shapes (2,2) (3,3) "
]
}
],
"source": [
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
"a + c"
]
},
{
"cell_type": "code",
"execution_count": 76,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[1, 2, 3],\n",
" [4, 5, 6]])"
]
},
"execution_count": 76,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"m = np.array([[1,2,3],[4,5,6]])\n",
"m"
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
{
"cell_type": "code",
"execution_count": 77,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[0.25, 0.5 , 0.75],\n",
" [1. , 1.25, 1.5 ]])"
]
},
"execution_count": 77,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"n = m * 0.25\n",
"n"
]
},
{
"cell_type": "code",
"execution_count": 79,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[0.25, 1. , 2.25],\n",
" [4. , 6.25, 9. ]])"
]
},
"execution_count": 79,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"m * n"
]
},
{
"cell_type": "code",
"execution_count": 80,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[0.25, 1. , 2.25],\n",
" [4. , 6.25, 9. ]])"
]
},
"execution_count": 80,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"np.multiply(m, n)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"To find the matrix product, you use NumPy's `matmul` function.\n",
"\n",
"If you have compatible shapes, then it's as simple as this:\n"
]
},
{
"cell_type": "code",
"execution_count": 81,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[1, 2, 3, 4],\n",
" [5, 6, 7, 8]])"
]
},
"execution_count": 81,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"a = np.array([[1,2,3,4],[5,6,7,8]])\n",
"a"
]
},
{
"cell_type": "code",
"execution_count": 82,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(2, 4)"
]
},
"execution_count": 82,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"a.shape"
]
},
{
"cell_type": "code",
"execution_count": 83,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 1, 2, 3],\n",
" [ 4, 5, 6],\n",
" [ 7, 8, 9],\n",
" [10, 11, 12]])"
]
},
"execution_count": 83,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"b = np.array([[1,2,3],[4,5,6],[7,8,9],[10,11,12]])\n",
"b"
]
},
{
"cell_type": "code",
"execution_count": 84,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(4, 3)"
]
},
"execution_count": 84,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"b.shape"
]
},
{
"cell_type": "code",
"execution_count": 85,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 70, 80, 90],\n",
" [158, 184, 210]])"
]
},
"execution_count": 85,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"c = np.matmul(a, b)\n",
"c"
]
},
{
"cell_type": "code",
"execution_count": 86,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(2, 3)"
]
},
"execution_count": 86,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"c.shape"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"If your matrices have incompatible shapes, you'll get an error, like the following:"
]
},
{
"cell_type": "code",
"execution_count": 87,
"metadata": {},
"outputs": [
{
"ename": "ValueError",
"evalue": "matmul: Input operand 1 has a mismatch in its core dimension 0, with gufunc signature (n?,k),(k,m?)->(n?,m?) (size 2 is different from 3)",
"output_type": "error",
"traceback": [
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
"\u001b[0;31mValueError\u001b[0m Traceback (most recent call last)",
"\u001b[0;32m<ipython-input-87-af3b88aa2232>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmatmul\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ma\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
"\u001b[0;31mValueError\u001b[0m: matmul: Input operand 1 has a mismatch in its core dimension 0, with gufunc signature (n?,k),(k,m?)->(n?,m?) (size 2 is different from 3)"
]
}
],
"source": [
"np.matmul(b, a)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### NumPy's `dot` function\n",
"\n",
"You may sometimes see NumPy's `dot` function in places where you would expect a `matmul`. \n",
"It turns out that the results of dot and matmul are the same if the matrices are two dimensional.\n",
"\n",
"So these two results are equivalent:"
]
},
{
"cell_type": "code",
"execution_count": 88,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[1, 2],\n",
" [3, 4]])"
]
},
"execution_count": 88,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"a = np.array([[1,2],[3,4]])\n",
"a"
]
},
{
"cell_type": "code",
"execution_count": 89,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 7, 10],\n",
" [15, 22]])"
]
},
"execution_count": 89,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"np.dot(a,a)"
]
},
{
"cell_type": "code",
"execution_count": 90,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 7, 10],\n",
" [15, 22]])"
]
},
"execution_count": 90,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"a.dot(a)"
]
},
{
"cell_type": "code",
"execution_count": 91,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 7, 10],\n",
" [15, 22]])"
]
},
"execution_count": 91,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"np.matmul(a,a)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"While these functions return the same results for two dimensional data, you should be careful \n",
"about which you choose when working with other data shapes. You can read more about the \n",
"differences, and find links to other NumPy functions, in the `matmul` and `dot` documentation. \n",
"\n",
"\n",
"\n",
"### Transpose\n",
"\n",
"Getting the transpose of a matrix is really easy in NumPy. Simply access \n",
"its `T` attribute. There is also a `transpose()` function which \n",
"returns the same thing, but you will rarely see that used anywhere because \n",
"typing `T` is so much easier. \n",
"\n",
"For example:\n"
]
},
{
"cell_type": "code",
"execution_count": 92,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 1, 2, 3, 4],\n",
" [ 5, 6, 7, 8],\n",
" [ 9, 10, 11, 12]])"
]
},
"execution_count": 92,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"m = np.array([[1,2,3,4], [5,6,7,8], [9,10,11,12]])\n",
"m"
]
},
{
"cell_type": "code",
"execution_count": 93,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 1, 5, 9],\n",
" [ 2, 6, 10],\n",
" [ 3, 7, 11],\n",
" [ 4, 8, 12]])"
]
},
"execution_count": 93,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"m.T"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"NumPy does this without actually moving any data in memory -\n",
"it simply changes the way it indexes the original matrix - \n",
"so it's quite efficient.\n",
"\n",
"However, that also means you need to be careful with how you modify objects, \n",
"because they are sharing the same data. For example, with the same matrix `m` \n",
"from above, let us make a new variable `m_t` that stores `m`'s transpose. \n",
"Then look what happens if we modify a value in `m_t`:\n"
]
},
{
"cell_type": "code",
"execution_count": 94,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 1, 5, 9],\n",
" [ 2, 6, 10],\n",
" [ 3, 7, 11],\n",
" [ 4, 200, 12]])"
]
},
"execution_count": 94,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"m_t = m.T\n",
"m_t[3][1] = 200\n",
"m_t"
]
},
{
"cell_type": "code",
"execution_count": 95,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 1, 2, 3, 4],\n",
" [ 5, 6, 7, 200],\n",
" [ 9, 10, 11, 12]])"
]
},
"execution_count": 95,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"m"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Notice how it modified both the transpose and the original matrix, too. \n",
"That's because they are sharing the same copy of data. So remember to \n",
"consider the transpose just as a different view of your matrix, rather \n",
"than a different matrix entirely."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Geometric Interpretation of Matrix Operation"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The following example will show what matrices and vectors are good for. \n",
"\n",
"We are going to start with a picture of a bug."
]
},
{
"cell_type": "code",
"execution_count": 102,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<matplotlib.image.AxesImage at 0x7f56299e9250>"
]
},
"execution_count": 102,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"%matplotlib inline \n",
"\n",
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"import matplotlib.image as mpimg\n",
"img = mpimg.imread('stinkbug4.JPG')\n",
"plt.imshow(img)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"If you open the picture with an image viewer, you can enlarge or reduce the picture or rotate it. Now the important question: What's going on behind the scene if you do rotate the picture? Or in other words: What does the computer do to achieve the rotation? \n",
"\n",
"One simple solution is using a matrices and vectors."
]
},
{
"cell_type": "code",
"execution_count": 105,
"metadata": {},
"outputs": [
{
"ename": "ModuleNotFoundError",
"evalue": "No module named 'skimage'",
"output_type": "error",
"traceback": [
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
"\u001b[0;31mModuleNotFoundError\u001b[0m Traceback (most recent call last)",
"\u001b[0;32m<ipython-input-105-16d8c7ff678e>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0;32mimport\u001b[0m \u001b[0mskimage\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 2\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mskimage\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtransform\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mrescale\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 3\u001b[0m \u001b[0mimg\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mrescale\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimg\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m.2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 4\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 5\u001b[0m \u001b[0mimg\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdelete\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimg\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m20\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'skimage'"
]
}
],
"source": [
"import skimage\n",
"from skimage.transform import rescale\n",
"img = rescale(img, .2)\n",
"\n",
"img = np.delete(img,np.arange(20), axis=0)\n",
"img = np.delete(img,np.arange(80,100), axis=0)\n",
"img = np.delete(img,np.arange(20), axis=1)\n",
"img = np.delete(img,np.arange(80,100), axis=1)\n",
"\n",
"col =np.zeros(img.shape[0]*img.shape[1])\n",
"\n",
"k = 0\n",
"for i in np.arange(img.shape[0]):\n",
" for j in np.arange(img.shape[1]):\n",
" col[k] = np.round(.99*img[i,j,:],0)\n",
" k = k+1\n",
" \n",
"image = np.array([np.repeat(np.arange(img.shape[0]),img.shape[1]), np.tile(np.arange(img.shape[1]),img.shape[0])]).T\n",
"image[:,0] = image[:,0] - 40\n",
"image[:,1] = image[:,1] - 40\n",
"\n",
"image = image[col==0,:]\n",
"\n",
"\n",
"plt.plot(image[:,0],image[:,1],\"o\",color=\"black\")\n",
"plt.axis('square')\n",
"plt.xlim(-40,40)\n",
"plt.ylim(-40,40)\n",
"plt.savefig(\"bug01.eps\",bbox_inches=\"tight\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## A real use case"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[[-0.27 0.45 0.64 0.31]] (1, 4)\n",
"[[ 0.02 0.001 -0.03 0.036]\n",
" [ 0.04 -0.003 0.025 0.009]\n",
" [ 0.012 -0.045 0.28 -0.067]] (3, 4)\n",
"Matrix multiplication gives:\n",
" [[-0.01299 0.00664 0.13494]] \n",
"or, equivalently:\n",
" [[-0.01299]\n",
" [ 0.00664]\n",
" [ 0.13494]]\n"
]
}
],
"source": [
"inputs = np.array([[-0.27, 0.45, 0.64, 0.31]])\n",
"print(inputs, inputs.shape)\n",
"\n",
"weights = np.array([[0.02, 0.001, -0.03, 0.036], \n",
" [0.04, -0.003, 0.025, 0.009], \n",
" [0.012, -0.045, 0.28, -0.067]])\n",
"print(weights, weights.shape)\n",
"\n",
"print(\"Matrix multiplication gives:\\n\", np.matmul(inputs, weights.T), \"\\nor, equivalently:\\n\", np.matmul(weights, inputs.T))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Some more useful Numpy methods"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"Showing some basic math on arrays\n",
"Max: 4\n",
"Average: 2.0\n",
"Max index: 2\n",
"\n",
"Use numpy to create a [3,3] dimension array with random number\n",
"[[0.92371879 0.58999086 0.76979433]\n",
" [0.48733651 0.44698554 0.91494542]\n",
" [0.59130531 0.69632003 0.32785335]]\n"
]
}
],
"source": [
"print(\"\\nShowing some basic math on arrays\")\n",
"\n",
"b = np.array([0,1,4,3,2])\n",
"print(\"Max: {}\".format(np.max(b)))\n",
"print(\"Average: {}\".format(np.average(b)))\n",
"print(\"Max index: {}\".format(np.argmax(b)))\n",
"\n",
"print(\"\\nUse numpy to create a [3,3] dimension array with random number\")\n",
"c = np.random.rand(3, 3)\n",
"print(c)"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.7.6"
}
},
"nbformat": 4,
"nbformat_minor": 4
}