-
Notifications
You must be signed in to change notification settings - Fork 10
/
Copy pathREADME-old.Rmd
174 lines (141 loc) · 4.8 KB
/
README-old.Rmd
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
Computer Science 499-007 (9235) at Northern Arizona University, Spring 2019
Optimization Algorithms for Machine Learning
TuTh 9:35-10:50
## Schedule
Jan 14 to May 10.
```r
all.day.vec <- seq(as.Date("2019-01-14"), as.Date("2019-05-10"), by = "day")
class.day.vec <- all.day.vec[strftime(all.day.vec, "%a") %in% c("Tue", "Thu")]
lecture.vec <- paste(
"* Lecture", seq_along(class.day.vec), strftime(class.day.vec, "%a %b %d"))
cat(lecture.vec, sep="\n")
```
Classification data sets
* ElemStatLearn::spam 2-class [4601, 57] output is last column (spam).
* ElemStatLearn::SAheart 2-class [462, 9] output is last column (chd).
* ElemStatLearn::zip.train: 10-class [7291, 256] output is first column.
* [Fashion-MNIST](https://github.com/zalandoresearch/fashion-mnist):
10-class [60000 x 784] Read into R via https://gist.github.com/brendano/39760
Regression data sets
* ElemStatLearn::prostate [97 x 8] output is lpsa column, ignore train column.
* ElemStatLearn::ozone [111 x 3] output is first column (ozone).
* nycflights13::flights, output is arr_delay.
```r
data(zip.train, package="ElemStatLearn")
library(data.table)
zip.wide <- data.table(zip.train)
setnames(zip.wide, c(
"digit", paste(rep(1:16, 16), rep(1:16, each=16))))
zip.wide[, example.i := 1:.N]
zip.tall <- melt(zip.wide, id.vars=c("example.i", "digit"))
zip.tall[, row := as.integer(sub(".* ", "", variable))]
zip.tall[, col := as.integer(sub(" .*", "", variable))]
zip.some <- zip.tall[example.i %in% c(1:10)]
library(ggplot2)
ggplot()+
facet_wrap("example.i")+
theme_bw()+
theme(panel.margin=grid::unit(0, "lines"))+
geom_tile(aes(
col, -row, fill=value),
data=zip.some)+
scale_fill_gradient(low="white", high="black")
```
Coding projects:
* lasso solver
* SVM solver
* k-means / GMM / hierarchical clustering
Topics:
* Support vector machines
* Neural networks
* Decision tree models
* Nearest neighbors
* K-means
* Gaussian mixture models
* Hierarchical clustering
* Hidden Markov models
* Optimal segmentation / dynamic programming
Murphy Chapter 1
- Logistic regression
- Overfitting / model selection
Murphy Chapter 16
* 16.5 Feedforward neural networks
LP/QP solver examples
* Murphy 7.4 Robust linear regression
* Lasso
* SVM
Murphy Chapter 7
* 7.5 Ridge regression.
Murphy Chapter 8
* 8.3.2 Steepest descent
* 8.3.3 Newton's method
* 8.3.5 Quasi-newton methods
* 8.3.6 L2 regularization
* 8.3.7 Multi-class logistic regression
Murphy Chapter 13
* Linear models
* Best subset selection
* Lasso
* Coordinate descent
* LARS
* Proximal and gradient projection methods
Applications
* Music transcription - TODO
* Translation - TODO google books?
* Spam filtering - OK
* Medical diagnosis - TODO slide Natl Geo, my NB
* Optical/intelligent character/word recognition TODO slide
* Image classification OK
* Image segmentation - TODO LabelMe slide
* Genomic segmentation - TODO copy slide.
## Lectures
* [Lecture 1 Tue Jan 15](2019-01-15-applications): Applications of
machine learning, syllabus, calculus/C quiz.
* Lecture 2 Thu Jan 17:
* Lecture 3 Tue Jan 22:
* Lecture 4 Thu Jan 24:
* Lecture 5 Tue Jan 29:
* Lecture 6 Thu Jan 31:
* Lecture 7 Tue Feb 05:
* Lecture 8 Thu Feb 07:
* Lecture 9 Tue Feb 12:
* Lecture 10 Thu Feb 14:
Coding project 1. R package for smooth optimization, linear/logistic
regression, compare L2 regularization with early stopping.
* Lecture 11 Tue Feb 19:
* Lecture 12 Thu Feb 21:
* Lecture 13 Tue Feb 26:
* Lecture 14 Thu Feb 28:
* Lecture 15 Tue Mar 05:
* Lecture 16 Thu Mar 07:
* Lecture 17 Tue Mar 12:
* Lecture 18 Thu Mar 14:
Coding project 2. R package for non-smooth optimization, Lasso and
SVM.
MIDTERM EXAM on smooth optimization algorithms for supervised learning
* Lecture 19 Tue Mar 19:
* Lecture 20 Thu Mar 21:
* Lecture 21 Tue Mar 26:
* Lecture 22 Thu Mar 28:
* Lecture 23 Tue Apr 02:
* Lecture 24 Thu Apr 04:
* Lecture 25 Tue Apr 09:
* Lecture 26 Thu Apr 11:
Coding project 3. R package for unsupervised learning
* Lecture 27 Tue Apr 16: paper review?
* Lecture 28 Thu Apr 18:
* Lecture 29 Tue Apr 23:
* Lecture 30 Thu Apr 25:
* Lecture 31 Tue Apr 30:
* Lecture 32 Thu May 02:
* Lecture 33 Tue May 07:
* Lecture 34 Thu May 09:
Coding project 4. R package for algo from ML literature.
-
FINAL EXAM on non-smooth, discrete, unsupervised algorithms
## Books
* https://www.cs.ubc.ca/~murphyk/MLbook/
* NAU library https://arizona-nau-primo.hosted.exlibrisgroup.com/primo-explore/fulldisplay?docid=01NAU_ALMA51166833980003842&context=L&vid=01NAU&lang=en_US&search_scope=Everything&adaptor=Local%20Search%20Engine&tab=default_tab&query=any,contains,murphy%20machine%20learning&sortby=rank&mode=Basic
* e-book http://eds.a.ebscohost.com/ehost/ebookviewer/ebook?sid=f4a68ba6-099a-4015-9d63-a342771786f3%40sdc-v-sessmgr02&vid=0&format=EB
* http://numerical.recipes/
* https://web.stanford.edu/~hastie/ElemStatLearn/