ml/QUESTIONS at master · davejacobs/ml · GitHub

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
Machine learning
----------------

- Look more into vector norms, how they relate to the graph
- Look into partial derivatives

Testing
-------

What would my ideal ML test framework look like?

- Has default dataset appropriate for testing specified type/class of algorithm
- Dataset is easily modifiable
- Fast for known datasets because statistics are memoized or mocked out
- Test passes if specified performance metric passes specified threshold
- Automatically splices data into learning data, test data, cross-validation data

Options for gradient-descent
----------------------------

- Current, recursive: (gradient-descent x ys thetas alpha iterations)
- Implicit original thetas, alpha, iterations: (gradient-descent x ys)
- Repeatedly (to pass to take): (gradient-descent x ys alpha)

How should I permute features when creating new ones automatically
; Eventually will do something more general
; x1^1 * x2^0 * x3^0
; x1^1 * x2^1

;     feature position
;
; e   1 0 0
; x   0 1 0
; p   0 0 1
; o
; n   2 0
; e   1 1
; n   0 2
; t
;     3 0
;     2 1
;     1 2
;     0 3

; 2 0 0
; 2 0 1
; 2 0 2
; 1 1 0
; 1 1 1
; 1 1 2
; 0 2 0
; 0 2 1
; 0 2 2


; The order is:
; x1^1 * x2^0
; x1^0 * x2^1
; x1^2 * x2^0
; x1^1 * x2^1
; x1^0 * x2^2
; x1^3 * x2^0
; x1^2 * x2^1
; x1^1 * x2^2
; x1^0 * x2^3
; x1^4 * x2^0
; ...