This repository has been archived by the owner on Nov 18, 2022. It is now read-only.
/
optimizetype.jl
178 lines (153 loc) · 5.25 KB
/
optimizetype.jl
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
"""
# OptimizeAlgorithm types
### Types defined
```julia
* Lbfgs::OptimizeAlgorithm : Euclidean manifold with unit metric
* Bfgs::OptimizeAlgorithm : Euclidean manifold with unit metric
* Newton::OptimizeAlgorithm : Euclidean manifold with unit metric
```
"""
abstract type OptimizeAlgorithm end
"""
# Lbfgs type and constructor
Settings for method=Lbfgs() in Optimize().
### Method
```julia
Lbfgs(;init_alpha=0.001, tol_obj=1e-8, tol_grad=1e-8, tol_param=1e-8, history_size=5)
```
### Optional arguments
```julia
* `init_alpha::Float64` : Linear search step size for first iteration
* `tol_obj::Float64` : Convergence tolerance for objective function
* `tol_rel_obj::Float64` : Relative change tolerance in objective function
* `tol_grad::Float64` : Convergence tolerance on norm of gradient
* `tol_rel_grad::Float64` : Relative change tolerance on norm of gradient
* `tol_param::Float64` : Convergence tolerance on parameter values
* `history_size::Int` : No of update vectors to use in Hessian approx
```
### Related help
```julia
?OptimizeMethod : List of available optimize methods
?Optimize : Optimize arguments
```
"""
mutable struct Lbfgs <: OptimizeAlgorithm
init_alpha::Float64
tol_obj::Float64
tol_rel_obj::Float64
tol_grad::Float64
tol_rel_grad::Float64
tol_param::Float64
history_size::Int64
end
Lbfgs(;init_alpha=0.001, tol_obj=1e-8, tol_rel_obj=1e4,
tol_grad=1e-8, tol_rel_grad=1e7, tol_param=1e-8, history_size=5) =
Lbfgs(init_alpha, tol_obj, tol_rel_obj, tol_grad, tol_rel_grad,
tol_param, history_size)
"""
# Bfgs type and constructor
Settings for method=Bfgs() in Optimize().
### Method
```julia
Bfgs(;init_alpha=0.001, tol_obj=1e-8, tol_rel_obj=1e4,
tol_grad=1e-8, tol_rel_grad=1e7, tol_param=1e-8)
```
### Optional arguments
```julia
* `init_alpha::Float64` : Linear search step size for first iteration
* `tol_obj::Float64` : Convergence tolerance for objective function
* `tol_rel_obj::Float64` : Relative change tolerance in objective function
* `tol_grad::Float64` : Convergence tolerance on norm of gradient
* `tol_rel_grad::Float64` : Relative change tolerance on norm of gradient
* `tol_param::Float64` : Convergence tolerance on parameter values
```
### Related help
```julia
?OptimizeMethod : List of available optimize methods
?Optimize : Optimize arguments
```
"""
mutable struct Bfgs <: OptimizeAlgorithm
init_alpha::Float64
tol_obj::Float64
tol_rel_obj::Float64
tol_grad::Float64
tol_rel_grad::Float64
tol_param::Float64
end
Bfgs(;init_alpha=0.001, tol_obj=1e-8, tol_rel_obj=1e4,
tol_grad=1e-8, tol_rel_grad=1e7, tol_param=1e-8) =
Bfgs(init_alpha, tol_obj, tol_rel_obj, tol_grad, tol_rel_grad, tol_param)
"""
# Newton type and constructor
Settings for method=Newton() in Optimize().
### Method
```julia
Newton()
```
### Related help
```julia
?OptimizeMethod : List of available optimize methods
?Optimize : Optimize arguments
```
"""
mutable struct Newton <: OptimizeAlgorithm
end
"""
# Optimize type and constructor
Settings for Optimize top level method.
### Method
```julia
Optimize(;
method=Lbfgs(),
iter=2000,
save_iterations=false
)
```
### Optional arguments
```julia
* `method::OptimizeMethod` : Optimization algorithm
* `iter::Int` : Total number of iterations
* `save_iterations::Bool` : Stream optimization progress to output
```
### Related help
```julia
?Stanmodel : Create a StanModel
?OptimizeAlgorithm : Available algorithms
```
"""
mutable struct Optimize <: Method
method::OptimizeAlgorithm
iter::Int64
save_iterations::Bool
end
Optimize(;method::OptimizeAlgorithm=Lbfgs(), iter::Number=2000,
save_iterations::Bool=false) =
Optimize(method, iter, save_iterations)
Optimize(method::OptimizeAlgorithm) = Optimize(method, 2000, false)
function optimize_show(io::IO, o::Optimize, compact::Bool)
if compact
println(io, "Optimize($(o.method), $(o.iter), $(o.save_iterations))")
else
println(io, " method = Optimize()")
if isa(o.method, Lbfgs)
println(io, " algorithm = Lbfgs()")
println(io, " init_alpha = ", o.method.init_alpha)
println(io, " tol_obj = ", o.method.tol_obj)
println(io, " tol_grad = ", o.method.tol_grad)
println(io, " tol_param = ", o.method.tol_param)
println(io, " history_size = ", o.method.history_size)
elseif isa(o.method, Bfgs)
println(io, " algorithm = Bfgs()")
println(io, " init_alpha = ", o.method.init_alpha)
println(io, " tol_obj = ", o.method.tol_obj)
println(io, " tol_grad = ", o.method.tol_grad)
println(io, " tol_param = ", o.method.tol_param)
else
println(io, " algorithm = Newton()")
end
println(io, " iterations = ", o.iter)
println(io, " save_iterations = ", o.save_iterations)
end
end
show(io::IO, o::Optimize) = optimize_show(io, o, false)