Cambios
En el instante 11 de octubre de 2025, 1:23:59 UTC,
-
Añadido recurso An evolutionary algorithm of linear complexity: application to training of deep neural networks a An evolutionary algorithm of linear complexity: application to training of deep neural networks
f | 1 | { | f | 1 | { |
2 | "author": "SI Valdez, A Rojas-Dom\u00ednguez", | 2 | "author": "SI Valdez, A Rojas-Dom\u00ednguez", | ||
3 | "author_email": null, | 3 | "author_email": null, | ||
4 | "creator_user_id": "a3da3ec9-3fd4-47a4-8d04-0a90b09614e0", | 4 | "creator_user_id": "a3da3ec9-3fd4-47a4-8d04-0a90b09614e0", | ||
5 | "extras": [ | 5 | "extras": [ | ||
6 | { | 6 | { | ||
7 | "key": "Publicaci\u00f3n", | 7 | "key": "Publicaci\u00f3n", | ||
8 | "value": "Revista" | 8 | "value": "Revista" | ||
9 | }, | 9 | }, | ||
10 | { | 10 | { | ||
11 | "key": "Tipo", | 11 | "key": "Tipo", | ||
12 | "value": "Publicaci\u00f3n" | 12 | "value": "Publicaci\u00f3n" | ||
13 | } | 13 | } | ||
14 | ], | 14 | ], | ||
15 | "groups": [ | 15 | "groups": [ | ||
16 | { | 16 | { | ||
17 | "description": "Este grupo integra las publicaciones | 17 | "description": "Este grupo integra las publicaciones | ||
18 | acad\u00e9micas derivadas de los proyectos de investigaci\u00f3n del | 18 | acad\u00e9micas derivadas de los proyectos de investigaci\u00f3n del | ||
19 | Observatorio Metropolitano CentroGeo. Incluye art\u00edculos | 19 | Observatorio Metropolitano CentroGeo. Incluye art\u00edculos | ||
20 | presentados en congresos nacionales e internacionales, manuscritos en | 20 | presentados en congresos nacionales e internacionales, manuscritos en | ||
21 | formato preprint, cap\u00edtulos de libro y trabajos publicados en | 21 | formato preprint, cap\u00edtulos de libro y trabajos publicados en | ||
22 | revistas cient\u00edficas especializadas. Estos materiales reflejan la | 22 | revistas cient\u00edficas especializadas. Estos materiales reflejan la | ||
23 | labor de investigaci\u00f3n, desarrollo metodol\u00f3gico y | 23 | labor de investigaci\u00f3n, desarrollo metodol\u00f3gico y | ||
24 | an\u00e1lisis territorial del observatorio, contribuyendo al avance | 24 | an\u00e1lisis territorial del observatorio, contribuyendo al avance | ||
25 | del conocimiento en temas urbanos, metropolitanos y geoespaciales.", | 25 | del conocimiento en temas urbanos, metropolitanos y geoespaciales.", | ||
26 | "display_name": "Publicaciones", | 26 | "display_name": "Publicaciones", | ||
27 | "id": "a15a6b77-ddf5-4594-acab-7e772938a5b0", | 27 | "id": "a15a6b77-ddf5-4594-acab-7e772938a5b0", | ||
28 | "image_display_url": "", | 28 | "image_display_url": "", | ||
29 | "name": "publicaciones", | 29 | "name": "publicaciones", | ||
30 | "title": "Publicaciones" | 30 | "title": "Publicaciones" | ||
31 | } | 31 | } | ||
32 | ], | 32 | ], | ||
33 | "id": "a5a7eb6d-e5e8-4e67-8260-b98accfbb5ec", | 33 | "id": "a5a7eb6d-e5e8-4e67-8260-b98accfbb5ec", | ||
34 | "isopen": false, | 34 | "isopen": false, | ||
35 | "license_id": null, | 35 | "license_id": null, | ||
36 | "license_title": null, | 36 | "license_title": null, | ||
37 | "maintainer": null, | 37 | "maintainer": null, | ||
38 | "maintainer_email": null, | 38 | "maintainer_email": null, | ||
39 | "metadata_created": "2025-10-11T01:23:59.289651", | 39 | "metadata_created": "2025-10-11T01:23:59.289651", | ||
n | 40 | "metadata_modified": "2025-10-11T01:23:59.289660", | n | 40 | "metadata_modified": "2025-10-11T01:23:59.878926", |
41 | "name": | 41 | "name": | ||
42 | ear-complexity-application-to-training-of-deep-neural-n-a261c0afdcb2", | 42 | ear-complexity-application-to-training-of-deep-neural-n-a261c0afdcb2", | ||
43 | "notes": "The performance of deep neural networks, such as Deep | 43 | "notes": "The performance of deep neural networks, such as Deep | ||
44 | Belief Networks formed by Restricted Boltzmann Machines (RBMs), | 44 | Belief Networks formed by Restricted Boltzmann Machines (RBMs), | ||
45 | strongly depends on their training, which is the process of adjusting | 45 | strongly depends on their training, which is the process of adjusting | ||
46 | their parameters. This process can be posed as an optimization problem | 46 | their parameters. This process can be posed as an optimization problem | ||
47 | over n dimensions. However, typical networks contain tens of thousands | 47 | over n dimensions. However, typical networks contain tens of thousands | ||
48 | of parameters, making this a High-Dimensional Problem (HDP). Although | 48 | of parameters, making this a High-Dimensional Problem (HDP). Although | ||
49 | different optimization methods have been employed for this goal, the | 49 | different optimization methods have been employed for this goal, the | ||
50 | use of most of the Evolutionary Algorithms (EAs) becomes prohibitive | 50 | use of most of the Evolutionary Algorithms (EAs) becomes prohibitive | ||
51 | due to their inability to deal with HDPs. For instance, the Covariance | 51 | due to their inability to deal with HDPs. For instance, the Covariance | ||
52 | Matrix Adaptation Evolutionary Strategy (CMA-ES) which is regarded as | 52 | Matrix Adaptation Evolutionary Strategy (CMA-ES) which is regarded as | ||
53 | one of the most effective EAs, exhibits the enormous disadvantage of | 53 | one of the most effective EAs, exhibits the enormous disadvantage of | ||
54 | requiring $O(n^2)$ memory and operations, making it unpractical for | 54 | requiring $O(n^2)$ memory and operations, making it unpractical for | ||
55 | problems with more than a few hundred variables. In this paper, we | 55 | problems with more than a few hundred variables. In this paper, we | ||
56 | introduce a novel EA that requires $O(n)$ operations and memory, but | 56 | introduce a novel EA that requires $O(n)$ operations and memory, but | ||
57 | delivers competitive solutions for the training stage of RBMs with | 57 | delivers competitive solutions for the training stage of RBMs with | ||
58 | over one million variables, when compared against CMA-ES and the | 58 | over one million variables, when compared against CMA-ES and the | ||
59 | Contrastive Divergence algorithm, which is the standard method for | 59 | Contrastive Divergence algorithm, which is the standard method for | ||
60 | training RBMs.", | 60 | training RBMs.", | ||
n | 61 | "num_resources": 0, | n | 61 | "num_resources": 1, |
62 | "num_tags": 13, | 62 | "num_tags": 13, | ||
63 | "organization": { | 63 | "organization": { | ||
64 | "approval_status": "approved", | 64 | "approval_status": "approved", | ||
65 | "created": "2022-05-19T00:10:30.480393", | 65 | "created": "2022-05-19T00:10:30.480393", | ||
66 | "description": "Observatorio Metropolitano CentroGeo", | 66 | "description": "Observatorio Metropolitano CentroGeo", | ||
67 | "id": "b3b3a79d-748a-4464-9471-732b6c74ec53", | 67 | "id": "b3b3a79d-748a-4464-9471-732b6c74ec53", | ||
68 | "image_url": | 68 | "image_url": | ||
69 | "2022-05-19-001030.456616FullColor1280x1024LogoOnly.png", | 69 | "2022-05-19-001030.456616FullColor1280x1024LogoOnly.png", | ||
70 | "is_organization": true, | 70 | "is_organization": true, | ||
71 | "name": "observatorio-metropolitano-centrogeo", | 71 | "name": "observatorio-metropolitano-centrogeo", | ||
72 | "state": "active", | 72 | "state": "active", | ||
73 | "title": "Observatorio Metropolitano CentroGeo", | 73 | "title": "Observatorio Metropolitano CentroGeo", | ||
74 | "type": "organization" | 74 | "type": "organization" | ||
75 | }, | 75 | }, | ||
76 | "owner_org": "b3b3a79d-748a-4464-9471-732b6c74ec53", | 76 | "owner_org": "b3b3a79d-748a-4464-9471-732b6c74ec53", | ||
77 | "private": false, | 77 | "private": false, | ||
78 | "relationships_as_object": [], | 78 | "relationships_as_object": [], | ||
79 | "relationships_as_subject": [], | 79 | "relationships_as_subject": [], | ||
t | 80 | "resources": [], | t | 80 | "resources": [ |
81 | { | ||||
82 | "cache_last_updated": null, | ||||
83 | "cache_url": null, | ||||
84 | "created": "2025-10-11T01:23:59.911749", | ||||
85 | "datastore_active": false, | ||||
86 | "description": "The performance of deep neural networks, such as | ||||
87 | Deep Belief Networks formed by Restricted Boltzmann Machines (RBMs), | ||||
88 | strongly depends on their training, which is the process of adjusting | ||||
89 | their parameters. This process can be posed as an optimization problem | ||||
90 | over n dimensions. However, typical networks contain tens of thousands | ||||
91 | of parameters, making this a High-Dimensional Problem (HDP). Although | ||||
92 | different optimization methods have been employed for this goal, the | ||||
93 | use of most of the Evolutionary Algorithms (EAs) becomes prohibitive | ||||
94 | due to their inability to deal with HDPs. For instance, the Covariance | ||||
95 | Matrix Adaptation Evolutionary Strategy (CMA-ES) which is regarded as | ||||
96 | one of the most effective EAs, exhibits the enormous disadvantage of | ||||
97 | requiring $O(n^2)$ memory and operations, making it unpractical for | ||||
98 | problems with more than a few hundred variables. In this paper, we | ||||
99 | introduce a novel EA that requires $O(n)$ operations and memory, but | ||||
100 | delivers competitive solutions for the training stage of RBMs with | ||||
101 | over one million variables, when compared against CMA-ES and the | ||||
102 | Contrastive Divergence algorithm, which is the standard method for | ||||
103 | training RBMs.", | ||||
104 | "format": "HTML", | ||||
105 | "hash": "", | ||||
106 | "id": "899f557c-824c-4eba-a676-23c0c105b20c", | ||||
107 | "last_modified": null, | ||||
108 | "metadata_modified": "2025-10-11T01:23:59.883047", | ||||
109 | "mimetype": null, | ||||
110 | "mimetype_inner": null, | ||||
111 | "name": "An evolutionary algorithm of linear complexity: | ||||
112 | application to training of deep neural networks", | ||||
113 | "package_id": "a5a7eb6d-e5e8-4e67-8260-b98accfbb5ec", | ||||
114 | "position": 0, | ||||
115 | "resource_type": null, | ||||
116 | "size": null, | ||||
117 | "state": "active", | ||||
118 | "url": "https://doi.org/10.1155/2019/2952304", | ||||
119 | "url_type": null | ||||
120 | } | ||||
121 | ], | ||||
81 | "state": "active", | 122 | "state": "active", | ||
82 | "tags": [ | 123 | "tags": [ | ||
83 | { | 124 | { | ||
84 | "display_name": "algorithm", | 125 | "display_name": "algorithm", | ||
85 | "id": "da224d7c-d794-48c8-8316-05d6b5fdc51c", | 126 | "id": "da224d7c-d794-48c8-8316-05d6b5fdc51c", | ||
86 | "name": "algorithm", | 127 | "name": "algorithm", | ||
87 | "state": "active", | 128 | "state": "active", | ||
88 | "vocabulary_id": null | 129 | "vocabulary_id": null | ||
89 | }, | 130 | }, | ||
90 | { | 131 | { | ||
91 | "display_name": "artificial-intelligence", | 132 | "display_name": "artificial-intelligence", | ||
92 | "id": "47cb06d3-c2a4-42af-b633-03eea6181083", | 133 | "id": "47cb06d3-c2a4-42af-b633-03eea6181083", | ||
93 | "name": "artificial-intelligence", | 134 | "name": "artificial-intelligence", | ||
94 | "state": "active", | 135 | "state": "active", | ||
95 | "vocabulary_id": null | 136 | "vocabulary_id": null | ||
96 | }, | 137 | }, | ||
97 | { | 138 | { | ||
98 | "display_name": "artificial-neural-network", | 139 | "display_name": "artificial-neural-network", | ||
99 | "id": "d63d29ea-7b70-4ed5-83cc-66f3b9586c1c", | 140 | "id": "d63d29ea-7b70-4ed5-83cc-66f3b9586c1c", | ||
100 | "name": "artificial-neural-network", | 141 | "name": "artificial-neural-network", | ||
101 | "state": "active", | 142 | "state": "active", | ||
102 | "vocabulary_id": null | 143 | "vocabulary_id": null | ||
103 | }, | 144 | }, | ||
104 | { | 145 | { | ||
105 | "display_name": "boltzmann-machine", | 146 | "display_name": "boltzmann-machine", | ||
106 | "id": "5eee20bd-596c-4b7f-a187-efb2a9478358", | 147 | "id": "5eee20bd-596c-4b7f-a187-efb2a9478358", | ||
107 | "name": "boltzmann-machine", | 148 | "name": "boltzmann-machine", | ||
108 | "state": "active", | 149 | "state": "active", | ||
109 | "vocabulary_id": null | 150 | "vocabulary_id": null | ||
110 | }, | 151 | }, | ||
111 | { | 152 | { | ||
112 | "display_name": "cma-es", | 153 | "display_name": "cma-es", | ||
113 | "id": "352ed64b-c0c4-4780-a4f1-871f7df7691f", | 154 | "id": "352ed64b-c0c4-4780-a4f1-871f7df7691f", | ||
114 | "name": "cma-es", | 155 | "name": "cma-es", | ||
115 | "state": "active", | 156 | "state": "active", | ||
116 | "vocabulary_id": null | 157 | "vocabulary_id": null | ||
117 | }, | 158 | }, | ||
118 | { | 159 | { | ||
119 | "display_name": "computer-science", | 160 | "display_name": "computer-science", | ||
120 | "id": "29cae056-cd7e-43f7-be5b-b25869a3fbf2", | 161 | "id": "29cae056-cd7e-43f7-be5b-b25869a3fbf2", | ||
121 | "name": "computer-science", | 162 | "name": "computer-science", | ||
122 | "state": "active", | 163 | "state": "active", | ||
123 | "vocabulary_id": null | 164 | "vocabulary_id": null | ||
124 | }, | 165 | }, | ||
125 | { | 166 | { | ||
126 | "display_name": "deep-neural-networks", | 167 | "display_name": "deep-neural-networks", | ||
127 | "id": "4f23ca30-4b12-401d-a93c-958cf047df6e", | 168 | "id": "4f23ca30-4b12-401d-a93c-958cf047df6e", | ||
128 | "name": "deep-neural-networks", | 169 | "name": "deep-neural-networks", | ||
129 | "state": "active", | 170 | "state": "active", | ||
130 | "vocabulary_id": null | 171 | "vocabulary_id": null | ||
131 | }, | 172 | }, | ||
132 | { | 173 | { | ||
133 | "display_name": "divergence-linguistics", | 174 | "display_name": "divergence-linguistics", | ||
134 | "id": "34d5602a-4b6e-464f-bc6e-c6636fc7cc58", | 175 | "id": "34d5602a-4b6e-464f-bc6e-c6636fc7cc58", | ||
135 | "name": "divergence-linguistics", | 176 | "name": "divergence-linguistics", | ||
136 | "state": "active", | 177 | "state": "active", | ||
137 | "vocabulary_id": null | 178 | "vocabulary_id": null | ||
138 | }, | 179 | }, | ||
139 | { | 180 | { | ||
140 | "display_name": "evolution-strategy", | 181 | "display_name": "evolution-strategy", | ||
141 | "id": "836dbf9c-a84f-424f-8df2-441a41df838c", | 182 | "id": "836dbf9c-a84f-424f-8df2-441a41df838c", | ||
142 | "name": "evolution-strategy", | 183 | "name": "evolution-strategy", | ||
143 | "state": "active", | 184 | "state": "active", | ||
144 | "vocabulary_id": null | 185 | "vocabulary_id": null | ||
145 | }, | 186 | }, | ||
146 | { | 187 | { | ||
147 | "display_name": "evolutionary-algorithm", | 188 | "display_name": "evolutionary-algorithm", | ||
148 | "id": "22db3110-8a27-435b-8d9a-c910fe7201ca", | 189 | "id": "22db3110-8a27-435b-8d9a-c910fe7201ca", | ||
149 | "name": "evolutionary-algorithm", | 190 | "name": "evolutionary-algorithm", | ||
150 | "state": "active", | 191 | "state": "active", | ||
151 | "vocabulary_id": null | 192 | "vocabulary_id": null | ||
152 | }, | 193 | }, | ||
153 | { | 194 | { | ||
154 | "display_name": "machine-learning", | 195 | "display_name": "machine-learning", | ||
155 | "id": "99557dd8-2914-4e28-8278-7a042cfb23d0", | 196 | "id": "99557dd8-2914-4e28-8278-7a042cfb23d0", | ||
156 | "name": "machine-learning", | 197 | "name": "machine-learning", | ||
157 | "state": "active", | 198 | "state": "active", | ||
158 | "vocabulary_id": null | 199 | "vocabulary_id": null | ||
159 | }, | 200 | }, | ||
160 | { | 201 | { | ||
161 | "display_name": "mathematical-optimization", | 202 | "display_name": "mathematical-optimization", | ||
162 | "id": "b33a86ee-df39-4619-8165-c8d73afbf5a4", | 203 | "id": "b33a86ee-df39-4619-8165-c8d73afbf5a4", | ||
163 | "name": "mathematical-optimization", | 204 | "name": "mathematical-optimization", | ||
164 | "state": "active", | 205 | "state": "active", | ||
165 | "vocabulary_id": null | 206 | "vocabulary_id": null | ||
166 | }, | 207 | }, | ||
167 | { | 208 | { | ||
168 | "display_name": "process-computing", | 209 | "display_name": "process-computing", | ||
169 | "id": "d07d93c6-53f7-48a4-bded-e205ec0bc2c2", | 210 | "id": "d07d93c6-53f7-48a4-bded-e205ec0bc2c2", | ||
170 | "name": "process-computing", | 211 | "name": "process-computing", | ||
171 | "state": "active", | 212 | "state": "active", | ||
172 | "vocabulary_id": null | 213 | "vocabulary_id": null | ||
173 | } | 214 | } | ||
174 | ], | 215 | ], | ||
175 | "title": "An evolutionary algorithm of linear complexity: | 216 | "title": "An evolutionary algorithm of linear complexity: | ||
176 | application to training of deep neural networks", | 217 | application to training of deep neural networks", | ||
177 | "type": "dataset", | 218 | "type": "dataset", | ||
178 | "url": "https://doi.org/10.1155/2019/2952304", | 219 | "url": "https://doi.org/10.1155/2019/2952304", | ||
179 | "version": null | 220 | "version": null | ||
180 | } | 221 | } |