Cambios
En el instante 11 de octubre de 2025, 1:24:03 UTC,
-
Añadido recurso Evolutionary Training of Deep Belief Networks for Handwritten Digit Recognition a Evolutionary Training of Deep Belief Networks for Handwritten Digit Recognition
f | 1 | { | f | 1 | { |
2 | "author": "S Espinoza-P\u00e9rez, A Rojas-Dom\u00ednguez, SI Valdez, | 2 | "author": "S Espinoza-P\u00e9rez, A Rojas-Dom\u00ednguez, SI Valdez, | ||
3 | LE Mancilla-Espinoza", | 3 | LE Mancilla-Espinoza", | ||
4 | "author_email": null, | 4 | "author_email": null, | ||
5 | "creator_user_id": "a3da3ec9-3fd4-47a4-8d04-0a90b09614e0", | 5 | "creator_user_id": "a3da3ec9-3fd4-47a4-8d04-0a90b09614e0", | ||
6 | "extras": [ | 6 | "extras": [ | ||
7 | { | 7 | { | ||
8 | "key": "Publicaci\u00f3n", | 8 | "key": "Publicaci\u00f3n", | ||
9 | "value": "Revista" | 9 | "value": "Revista" | ||
10 | }, | 10 | }, | ||
11 | { | 11 | { | ||
12 | "key": "Tipo", | 12 | "key": "Tipo", | ||
13 | "value": "Publicaci\u00f3n" | 13 | "value": "Publicaci\u00f3n" | ||
14 | } | 14 | } | ||
15 | ], | 15 | ], | ||
16 | "groups": [ | 16 | "groups": [ | ||
17 | { | 17 | { | ||
18 | "description": "Este grupo integra las publicaciones | 18 | "description": "Este grupo integra las publicaciones | ||
19 | acad\u00e9micas derivadas de los proyectos de investigaci\u00f3n del | 19 | acad\u00e9micas derivadas de los proyectos de investigaci\u00f3n del | ||
20 | Observatorio Metropolitano CentroGeo. Incluye art\u00edculos | 20 | Observatorio Metropolitano CentroGeo. Incluye art\u00edculos | ||
21 | presentados en congresos nacionales e internacionales, manuscritos en | 21 | presentados en congresos nacionales e internacionales, manuscritos en | ||
22 | formato preprint, cap\u00edtulos de libro y trabajos publicados en | 22 | formato preprint, cap\u00edtulos de libro y trabajos publicados en | ||
23 | revistas cient\u00edficas especializadas. Estos materiales reflejan la | 23 | revistas cient\u00edficas especializadas. Estos materiales reflejan la | ||
24 | labor de investigaci\u00f3n, desarrollo metodol\u00f3gico y | 24 | labor de investigaci\u00f3n, desarrollo metodol\u00f3gico y | ||
25 | an\u00e1lisis territorial del observatorio, contribuyendo al avance | 25 | an\u00e1lisis territorial del observatorio, contribuyendo al avance | ||
26 | del conocimiento en temas urbanos, metropolitanos y geoespaciales.", | 26 | del conocimiento en temas urbanos, metropolitanos y geoespaciales.", | ||
27 | "display_name": "Publicaciones", | 27 | "display_name": "Publicaciones", | ||
28 | "id": "a15a6b77-ddf5-4594-acab-7e772938a5b0", | 28 | "id": "a15a6b77-ddf5-4594-acab-7e772938a5b0", | ||
29 | "image_display_url": "", | 29 | "image_display_url": "", | ||
30 | "name": "publicaciones", | 30 | "name": "publicaciones", | ||
31 | "title": "Publicaciones" | 31 | "title": "Publicaciones" | ||
32 | } | 32 | } | ||
33 | ], | 33 | ], | ||
34 | "id": "bc86979a-4217-4811-8be8-483e9532d36c", | 34 | "id": "bc86979a-4217-4811-8be8-483e9532d36c", | ||
35 | "isopen": false, | 35 | "isopen": false, | ||
36 | "license_id": null, | 36 | "license_id": null, | ||
37 | "license_title": null, | 37 | "license_title": null, | ||
38 | "maintainer": null, | 38 | "maintainer": null, | ||
39 | "maintainer_email": null, | 39 | "maintainer_email": null, | ||
40 | "metadata_created": "2025-10-11T01:24:03.008052", | 40 | "metadata_created": "2025-10-11T01:24:03.008052", | ||
n | 41 | "metadata_modified": "2025-10-11T01:24:03.008064", | n | 41 | "metadata_modified": "2025-10-11T01:24:03.629372", |
42 | "name": | 42 | "name": | ||
43 | -deep-belief-networks-for-handwritten-digit-recognition-53bb7da3ac24", | 43 | -deep-belief-networks-for-handwritten-digit-recognition-53bb7da3ac24", | ||
44 | "notes": "Two of the most representative deep architectures are Deep | 44 | "notes": "Two of the most representative deep architectures are Deep | ||
45 | Convolutional Neural Networks and Deep Belief Networks (DBNs).Both of | 45 | Convolutional Neural Networks and Deep Belief Networks (DBNs).Both of | ||
46 | these can be applied to the problem of pattern | 46 | these can be applied to the problem of pattern | ||
47 | classification.Nevertheless, they differ in the training method: while | 47 | classification.Nevertheless, they differ in the training method: while | ||
48 | the first is trained by backpropagation of the error through the whole | 48 | the first is trained by backpropagation of the error through the whole | ||
49 | network, the latter is typically pre-trained on a per-layer basis | 49 | network, the latter is typically pre-trained on a per-layer basis | ||
50 | using an unsupervised algorithm known as Contrastive Divergence (CD), | 50 | using an unsupervised algorithm known as Contrastive Divergence (CD), | ||
51 | and then it is fine-tuned with a gradient descent algorithm.Although | 51 | and then it is fine-tuned with a gradient descent algorithm.Although | ||
52 | metaheuristic algorithms have been widely applied for hyperparameter | 52 | metaheuristic algorithms have been widely applied for hyperparameter | ||
53 | tuning, little has been published regarding alternative methods to | 53 | tuning, little has been published regarding alternative methods to | ||
54 | pre-train DBNs.In this work, we substitute the conventional | 54 | pre-train DBNs.In this work, we substitute the conventional | ||
55 | pre-training method with an evolutionary optimization algorithm called | 55 | pre-training method with an evolutionary optimization algorithm called | ||
56 | the Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES).The | 56 | the Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES).The | ||
57 | pretraining is achieved by minimizing the so-called reconstruction | 57 | pretraining is achieved by minimizing the so-called reconstruction | ||
58 | error.This proposal is validated on the problem of MNIST digit | 58 | error.This proposal is validated on the problem of MNIST digit | ||
59 | recognition by training a Deep Belief Network, following the | 59 | recognition by training a Deep Belief Network, following the | ||
60 | methodology described by Hinton and Salakhutdinov (Science, 2006).It | 60 | methodology described by Hinton and Salakhutdinov (Science, 2006).It | ||
61 | is also compared against the well known Genetic Algorithm (GA).We | 61 | is also compared against the well known Genetic Algorithm (GA).We | ||
62 | provide evidence to show that, although the computational cost is | 62 | provide evidence to show that, although the computational cost is | ||
63 | significantly highern than CD, the use of CMA-ES leads to a | 63 | significantly highern than CD, the use of CMA-ES leads to a | ||
64 | significantly smaller reconstruction error than CD and the GA.", | 64 | significantly smaller reconstruction error than CD and the GA.", | ||
n | 65 | "num_resources": 0, | n | 65 | "num_resources": 1, |
66 | "num_tags": 6, | 66 | "num_tags": 6, | ||
67 | "organization": { | 67 | "organization": { | ||
68 | "approval_status": "approved", | 68 | "approval_status": "approved", | ||
69 | "created": "2022-05-19T00:10:30.480393", | 69 | "created": "2022-05-19T00:10:30.480393", | ||
70 | "description": "Observatorio Metropolitano CentroGeo", | 70 | "description": "Observatorio Metropolitano CentroGeo", | ||
71 | "id": "b3b3a79d-748a-4464-9471-732b6c74ec53", | 71 | "id": "b3b3a79d-748a-4464-9471-732b6c74ec53", | ||
72 | "image_url": | 72 | "image_url": | ||
73 | "2022-05-19-001030.456616FullColor1280x1024LogoOnly.png", | 73 | "2022-05-19-001030.456616FullColor1280x1024LogoOnly.png", | ||
74 | "is_organization": true, | 74 | "is_organization": true, | ||
75 | "name": "observatorio-metropolitano-centrogeo", | 75 | "name": "observatorio-metropolitano-centrogeo", | ||
76 | "state": "active", | 76 | "state": "active", | ||
77 | "title": "Observatorio Metropolitano CentroGeo", | 77 | "title": "Observatorio Metropolitano CentroGeo", | ||
78 | "type": "organization" | 78 | "type": "organization" | ||
79 | }, | 79 | }, | ||
80 | "owner_org": "b3b3a79d-748a-4464-9471-732b6c74ec53", | 80 | "owner_org": "b3b3a79d-748a-4464-9471-732b6c74ec53", | ||
81 | "private": false, | 81 | "private": false, | ||
82 | "relationships_as_object": [], | 82 | "relationships_as_object": [], | ||
83 | "relationships_as_subject": [], | 83 | "relationships_as_subject": [], | ||
t | 84 | "resources": [], | t | 84 | "resources": [ |
85 | { | ||||
86 | "cache_last_updated": null, | ||||
87 | "cache_url": null, | ||||
88 | "created": "2025-10-11T01:24:03.670887", | ||||
89 | "datastore_active": false, | ||||
90 | "description": "Two of the most representative deep | ||||
91 | architectures are Deep Convolutional Neural Networks and Deep Belief | ||||
92 | Networks (DBNs).Both of these can be applied to the problem of pattern | ||||
93 | classification.Nevertheless, they differ in the training method: while | ||||
94 | the first is trained by backpropagation of the error through the whole | ||||
95 | network, the latter is typically pre-trained on a per-layer basis | ||||
96 | using an unsupervised algorithm known as Contrastive Divergence (CD), | ||||
97 | and then it is fine-tuned with a gradient descent algorithm.Although | ||||
98 | metaheuristic algorithms have been widely applied for hyperparameter | ||||
99 | tuning, little has been published regarding alternative methods to | ||||
100 | pre-train DBNs.In this work, we substitute the conventional | ||||
101 | pre-training method with an evolutionary optimization algorithm called | ||||
102 | the Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES).The | ||||
103 | pretraining is achieved by minimizing the so-called reconstruction | ||||
104 | error.This proposal is validated on the problem of MNIST digit | ||||
105 | recognition by training a Deep Belief Network, following the | ||||
106 | methodology described by Hinton and Salakhutdinov (Science, 2006).It | ||||
107 | is also compared against the well known Genetic Algorithm (GA).We | ||||
108 | provide evidence to show that, although the computational cost is | ||||
109 | significantly highern than CD, the use of CMA-ES leads to a | ||||
110 | significantly smaller reconstruction error than CD and the GA.", | ||||
111 | "format": "HTML", | ||||
112 | "hash": "", | ||||
113 | "id": "c334ad49-6be6-4517-8b25-5bc83b1f1741", | ||||
114 | "last_modified": null, | ||||
115 | "metadata_modified": "2025-10-11T01:24:03.634517", | ||||
116 | "mimetype": null, | ||||
117 | "mimetype_inner": null, | ||||
118 | "name": "Evolutionary Training of Deep Belief Networks for | ||||
119 | Handwritten Digit Recognition", | ||||
120 | "package_id": "bc86979a-4217-4811-8be8-483e9532d36c", | ||||
121 | "position": 0, | ||||
122 | "resource_type": null, | ||||
123 | "size": null, | ||||
124 | "state": "active", | ||||
125 | "url": "https://doi.org/10.13053/rcs-148-3-10", | ||||
126 | "url_type": null | ||||
127 | } | ||||
128 | ], | ||||
85 | "state": "active", | 129 | "state": "active", | ||
86 | "tags": [ | 130 | "tags": [ | ||
87 | { | 131 | { | ||
88 | "display_name": "artificial-intelligence", | 132 | "display_name": "artificial-intelligence", | ||
89 | "id": "47cb06d3-c2a4-42af-b633-03eea6181083", | 133 | "id": "47cb06d3-c2a4-42af-b633-03eea6181083", | ||
90 | "name": "artificial-intelligence", | 134 | "name": "artificial-intelligence", | ||
91 | "state": "active", | 135 | "state": "active", | ||
92 | "vocabulary_id": null | 136 | "vocabulary_id": null | ||
93 | }, | 137 | }, | ||
94 | { | 138 | { | ||
95 | "display_name": "computer-science", | 139 | "display_name": "computer-science", | ||
96 | "id": "29cae056-cd7e-43f7-be5b-b25869a3fbf2", | 140 | "id": "29cae056-cd7e-43f7-be5b-b25869a3fbf2", | ||
97 | "name": "computer-science", | 141 | "name": "computer-science", | ||
98 | "state": "active", | 142 | "state": "active", | ||
99 | "vocabulary_id": null | 143 | "vocabulary_id": null | ||
100 | }, | 144 | }, | ||
101 | { | 145 | { | ||
102 | "display_name": "digit-recognition", | 146 | "display_name": "digit-recognition", | ||
103 | "id": "0d444652-f9e4-4197-b9a3-9097b67a9b57", | 147 | "id": "0d444652-f9e4-4197-b9a3-9097b67a9b57", | ||
104 | "name": "digit-recognition", | 148 | "name": "digit-recognition", | ||
105 | "state": "active", | 149 | "state": "active", | ||
106 | "vocabulary_id": null | 150 | "vocabulary_id": null | ||
107 | }, | 151 | }, | ||
108 | { | 152 | { | ||
109 | "display_name": "numerical-digit", | 153 | "display_name": "numerical-digit", | ||
110 | "id": "d1bf41fc-f39a-4c10-b534-a3c16ac8f367", | 154 | "id": "d1bf41fc-f39a-4c10-b534-a3c16ac8f367", | ||
111 | "name": "numerical-digit", | 155 | "name": "numerical-digit", | ||
112 | "state": "active", | 156 | "state": "active", | ||
113 | "vocabulary_id": null | 157 | "vocabulary_id": null | ||
114 | }, | 158 | }, | ||
115 | { | 159 | { | ||
116 | "display_name": "speech-recognition", | 160 | "display_name": "speech-recognition", | ||
117 | "id": "5181565a-95e4-4d7c-b459-d8e5425c1ee7", | 161 | "id": "5181565a-95e4-4d7c-b459-d8e5425c1ee7", | ||
118 | "name": "speech-recognition", | 162 | "name": "speech-recognition", | ||
119 | "state": "active", | 163 | "state": "active", | ||
120 | "vocabulary_id": null | 164 | "vocabulary_id": null | ||
121 | }, | 165 | }, | ||
122 | { | 166 | { | ||
123 | "display_name": "training-meteorology", | 167 | "display_name": "training-meteorology", | ||
124 | "id": "5278a683-25af-4654-9242-1566903f1da0", | 168 | "id": "5278a683-25af-4654-9242-1566903f1da0", | ||
125 | "name": "training-meteorology", | 169 | "name": "training-meteorology", | ||
126 | "state": "active", | 170 | "state": "active", | ||
127 | "vocabulary_id": null | 171 | "vocabulary_id": null | ||
128 | } | 172 | } | ||
129 | ], | 173 | ], | ||
130 | "title": "Evolutionary Training of Deep Belief Networks for | 174 | "title": "Evolutionary Training of Deep Belief Networks for | ||
131 | Handwritten Digit Recognition", | 175 | Handwritten Digit Recognition", | ||
132 | "type": "dataset", | 176 | "type": "dataset", | ||
133 | "url": "https://doi.org/10.13053/rcs-148-3-10", | 177 | "url": "https://doi.org/10.13053/rcs-148-3-10", | ||
134 | "version": null | 178 | "version": null | ||
135 | } | 179 | } |