\uc790\uae30 \uc870\uc9c1\ud654 \uc9c0\ub3c4(Self-Organizing Map, SOM)\uc740 \ube44\uc120\ud615 \ucc28\uc6d0 \ucd95\uc18c \ubc0f \ub370\uc774\ud130 \uad70\uc9d1\ud654\uc5d0 \uc0ac\uc6a9\ub418\ub294 \ube44\uc9c0\ub3c4 \ud559\uc2b5 \uc54c\uace0\ub9ac\uc998\uc785\ub2c8\ub2e4. \uc774 \uac15\uc88c\uc5d0\uc11c\ub294 SOM\uc758 \uae30\ubcf8 \uac1c\ub150, \uc791\ub3d9 \uc6d0\ub9ac, \uadf8\ub9ac\uace0 \ud30c\uc774\ud1a0\uce58<\/strong>\ub97c \uc0ac\uc6a9\ud55c \uad6c\ud604 \ubc29\ubc95\uc5d0 \ub300\ud574 \uc124\uba85\ud558\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n \uc790\uae30 \uc870\uc9c1\ud654 \uc9c0\ub3c4\ub294 \uc6d0\ub798 Teuvo Kohonen<\/strong>\uc5d0 \uc758\ud574 \uac1c\ubc1c\ub41c \uc2e0\uacbd\ub9dd\uc785\ub2c8\ub2e4. SOM\uc740 \uace0\ucc28\uc6d0 \ub370\uc774\ud130\ub97c \uc800\ucc28\uc6d0 \uacf5\uac04(\uc77c\ubc18\uc801\uc73c\ub85c 2D \uaca9\uc790)\uc73c\ub85c \ub9e4\ud551\ud558\ub294 \ub370 \uc0ac\uc6a9\ub429\ub2c8\ub2e4. \uc774 \uacfc\uc815\uc5d0\uc11c \ub370\uc774\ud130\uac00 \uc720\uc0ac\ud55c \ud2b9\uc131\uc744 \uac00\uc9c4 \uc774\uc6c3 \ub178\ub4dc\ub4e4\ub85c \uad6c\uc131\ub41c \uc9c0\ub3c4 \ud615\ud0dc\ub85c \uc870\uc9c1\ub429\ub2c8\ub2e4.<\/p>\n SOM\uc740 \uc785\ub825 \ubca1\ud130\uc640 \ub178\ub4dc \ubca1\ud130 \uac04\uc758 \uac70\ub9ac\ub97c \uacc4\uc0b0\ud558\uc5ec \ud559\uc2b5\uc744 \uc9c4\ud589\ud569\ub2c8\ub2e4. \ub2e4\uc74c\uc740 SOM\uc758 \uc77c\ubc18\uc801\uc778 \ud559\uc2b5 \ub2e8\uacc4\uc785\ub2c8\ub2e4:<\/p>\n \ubaa8\ub4e0 \ub178\ub4dc\ub294 \ub79c\ub364\ud558\uac8c \ucd08\uae30\ud654\ub429\ub2c8\ub2e4. \uac01 \ub178\ub4dc\ub294 \uc785\ub825 \ub370\uc774\ud130\uc640 \ub3d9\uc77c\ud55c \ucc28\uc6d0\uc758 \uac00\uc911\uce58 \ubca1\ud130\ub97c \uac00\uc9d1\ub2c8\ub2e4.<\/p>\n \ud6c8\ub828 \uc0d8\ud50c\uc744 \ub79c\ub364\ud558\uac8c \uc120\ud0dd\ud569\ub2c8\ub2e4. \uac01 \uc0d8\ud50c\uc740 SOM\uc758 \uc785\ub825\uc774 \ub429\ub2c8\ub2e4.<\/p>\n \uc120\ud0dd\ub41c \uc785\ub825 \ub370\uc774\ud130\uc640 \uac00\uc7a5 \uc720\uc0ac\ud55c \ub178\ub4dc\ub97c \ucc3e\uc2b5\ub2c8\ub2e4. \uc774 \ub178\ub4dc\ub97c \ubca0\uc2a4\ud2b8 \ub9e4\uce6d \uc720\ub2db(BMU)<\/strong>\ub77c\uace0 \ud569\ub2c8\ub2e4.<\/p>\n BMU\uc640 \uadf8\uc758 \uc774\uc6c3 \ub178\ub4dc\ub4e4\uc758 \uac00\uc911\uce58\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\uc5ec \uc785\ub825 \ub370\uc774\ud130\uc5d0 \ub354 \uac00\uae4c\uc6cc\uc9c0\ub3c4\ub85d \uc870\uc815\ud569\ub2c8\ub2e4. \uc774 \uacfc\uc815\uc740 \ub2e4\uc74c\uacfc \uac19\uc2b5\ub2c8\ub2e4:<\/p>\n \uc5ec\uae30\uc11c:<\/p>\n 2-4\ub2e8\uacc4\ub97c \ucda9\ubd84\ud55c \uc218\uc758 \uc5d0\ud3ec\ud06c(epoch) \ub3d9\uc548 \ubc18\ubcf5\ud558\uc5ec \uac00\uc911\uce58\ub97c \uc810\uc9c4\uc801\uc73c\ub85c \uc5c5\ub370\uc774\ud2b8\ud569\ub2c8\ub2e4.<\/p>\n \uc774\uc81c \ud30c\uc774\ud1a0\uce58\ub97c \uc0ac\uc6a9\ud558\uc5ec SOM\uc744 \uad6c\ud604\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4. \uc5ec\uae30\uc5d0\uc11c\ub294 \uae30\ubcf8\uc801\uc778 SOM\uc744 \uad6c\ucd95\ud558\uace0 \uc2dc\uac01\ud654\ud558\ub294 \uacfc\uc815\uc744 \ubcf4\uc5ec\ub4dc\ub9bd\ub2c8\ub2e4.<\/p>\n \uba3c\uc800 \ud544\uc694\ud55c \ub77c\uc774\ube0c\ub7ec\ub9ac\ub97c \uc124\uce58\ud569\ub2c8\ub2e4.<\/p>\n \ub2e4\uc74c\uc73c\ub85c, SOM \ud074\ub798\uc2a4\ub97c \uc815\uc758\ud569\ub2c8\ub2e4. \uc774 \ud074\ub798\uc2a4\ub294 \uac00\uc911\uce58 \ucd08\uae30\ud654, BMU \ucc3e\uae30, \uac00\uc911\uce58 \uc5c5\ub370\uc774\ud2b8 \uae30\ub2a5\uc744 \ud3ec\ud568\ud569\ub2c8\ub2e4.<\/p>\n \uc801\uc808\ud55c \ub370\uc774\ud130\ub97c \uc900\ube44\ud558\uace0 SOM \ubaa8\ub378\uc744 \ud6c8\ub828\uc2dc\ud0a4\uaca0\uc2b5\ub2c8\ub2e4. \uc5ec\uae30\uc11c\ub294 \ub79c\ub364\ud558\uac8c \uc0dd\uc131\ud55c \ub370\uc774\ud130\ub97c \uc0ac\uc6a9\ud560 \uac83\uc785\ub2c8\ub2e4.<\/p>\n \ud6c8\ub828\ub41c SOM\uc758 \uac00\uc911\uce58\ub97c \uc2dc\uac01\ud654\ud558\uc5ec \ub370\uc774\ud130\uc758 \ubd84\ud3ec\ub97c \ud655\uc778\ud574\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n \uc774 \uac15\uc88c\uc5d0\uc11c\ub294 \uc790\uae30 \uc870\uc9c1\ud654 \uc9c0\ub3c4(SOM)\uc758 \uae30\ubcf8 \uc6d0\ub9ac\uc640 \ud30c\uc774\ud1a0\uce58\ub97c \ud1b5\ud574 SOM\uc744 \uad6c\ud604\ud558\ub294 \ubc29\ubc95\uc744 \uc0b4\ud3b4\ubcf4\uc558\uc2b5\ub2c8\ub2e4. SOM\uc740 \ud6a8\uacfc\uc801\uc778 \ube44\uc9c0\ub3c4 \ud559\uc2b5 \uae30\ubc95\uc73c\ub85c, \ub370\uc774\ud130\uc758 \ud328\ud134\uc744 \ud30c\uc545\ud558\uace0 \uad70\uc9d1\ud654\ub97c \uc218\ud589\ud558\ub294 \ub370 \uc720\uc6a9\ud569\ub2c8\ub2e4. \uc55e\uc73c\ub85c\ub294 \ubcf4\ub2e4 \ubcf5\uc7a1\ud55c \ub370\uc774\ud130\uc14b\uc5d0\uc11c SOM\uc758 \ud65c\uc6a9\uc744 \uc2e4\ud5d8\ud574\ubcf4\uac70\ub098, \ucd5c\uc801\ud654 \uae30\ubc95\uc744 \uc801\uc6a9\ud558\uc5ec \ud559\uc2b5 \uc131\ub2a5\uc744 \ud5a5\uc0c1\uc2dc\ud0ac \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n \uc774 \uae00\uc774 \ub525\ub7ec\ub2dd\uc758 \uc138\uacc4\ub97c \ud0d0\ud5d8\ud558\ub294 \ub370 \ub3c4\uc6c0\uc774 \ub418\uc5c8\uae30\ub97c \ubc14\ub78d\ub2c8\ub2e4! \uc9c8\ubb38\uc774\ub098 \ud53c\ub4dc\ubc31\uc774 \uc788\ub2e4\uba74 \ub313\uae00\ub85c \ub0a8\uaca8\uc8fc\uc138\uc694.<\/p>\n <\/body><\/p>\n","protected":false},"excerpt":{"rendered":"\uc790\uae30 \uc870\uc9c1\ud654 \uc9c0\ub3c4(SOM)\ub780?<\/h2>\n
SOM\uc758 \uc8fc\uc694 \ud2b9\uc9d5<\/h3>\n
\n
SOM\uc758 \uc791\ub3d9 \uc6d0\ub9ac<\/h2>\n
1. \ucd08\uae30\ud654<\/h3>\n
2. \uc785\ub825 \ub370\uc774\ud130 \uc120\ud0dd<\/h3>\n
3. \uac00\uc7a5 \uac00\uae4c\uc6b4 \ub178\ub4dc \ucc3e\uae30<\/h3>\n
4. \uac00\uc911\uce58 \uc5c5\ub370\uc774\ud2b8<\/h3>\n
\nw_{i}(t+1) = w_{i}(t) + \u03b1(t) * h_{i,j}(t) * (x(t) - w_{i}(t))\n<\/code><\/pre>\n\n
w_{i}<\/code>: \ub178\ub4dc\uc758 \uac00\uc911\uce58 \ubca1\ud130<\/li>\n\u03b1(t)<\/code>: \ud559\uc2b5\ub960<\/li>\nh_{i,j}(t)<\/code>: BMU\uc5d0 \ub300\ud55c \ub178\ub4dc i<\/code>\uc758 \uc774\uc6c3 \ud568\uc218<\/li>\nx(t)<\/code>: \uc785\ub825 \ubca1\ud130<\/li>\n<\/ul>\n5. \ubc18\ubcf5<\/h3>\n
\ud30c\uc774\ud1a0\uce58\ub97c \uc774\uc6a9\ud55c SOM \uad6c\ud604<\/h2>\n
\ud544\uc694\ud55c \ub77c\uc774\ube0c\ub7ec\ub9ac \uc124\uce58<\/h3>\n
!pip install torch numpy matplotlib<\/code><\/pre>\n\ubaa8\ub378 \ud074\ub798\uc2a4 \uc815\uc758<\/h3>\n
\nimport numpy as np\nimport torch\n\nclass SelfOrganizingMap:\n def __init__(self, m, n, input_dim, learning_rate=0.5, sigma=None):\n self.m = m # grid rows\n self.n = n # grid columns\n self.input_dim = input_dim\n self.learning_rate = learning_rate\n self.sigma = sigma if sigma else max(m, n) \/ 2\n\n # Initialize weight vectors\n self.weights = torch.rand(m, n, input_dim)\n\n def find_bmu(self, x):\n distances = torch.sqrt(torch.sum((self.weights - x) ** 2, dim=2))\n bmu_index = torch.argmin(distances)\n return bmu_index \/\/ self.n, bmu_index % self.n # return row, column\n\n def update_weights(self, x, bmu, iteration):\n learning_rate = self.learning_rate * np.exp(-iteration \/ 100)\n sigma = self.sigma * np.exp(-iteration \/ 100)\n\n for i in range(self.m):\n for j in range(self.n):\n h = self.neighbourhood(bmu, (i, j), sigma)\n self.weights[i, j] += learning_rate * h * (x - self.weights[i, j])\n\n def neighbourhood(self, bmu, point, sigma):\n distance = np.sqrt((bmu[0] - point[0]) ** 2 + (bmu[1] - point[1]) ** 2)\n return np.exp(-distance ** 2 \/ (2 * sigma ** 2))\n\n def train(self, data, num_iterations):\n for i in range(num_iterations):\n for x in data:\n bmu = self.find_bmu(x)\n self.update_weights(x, bmu, i)\n<\/code><\/pre>\n\ub370\uc774\ud130 \uc900\ube44 \ubc0f \ubaa8\ub378 \ud6c8\ub828<\/h3>\n
\n# Generate random data\ndata = torch.rand(200, 3) # 200 samples, 3 dimensions\n\n# Create and train SOM\nsom = SelfOrganizingMap(10, 10, 3)\nsom.train(data, 100)\n<\/code><\/pre>\n\uacb0\uacfc \uc2dc\uac01\ud654<\/h3>\n
\nimport matplotlib.pyplot as plt\n\ndef plot_som(som):\n plt.figure(figsize=(8, 8))\n for i in range(som.m):\n for j in range(som.n):\n plt.scatter(som.weights[i, j, 0].item(), som.weights[i, j, 1].item(), c='blue')\n plt.title('Self Organizing Map')\n plt.xlabel('Dimension 1')\n plt.ylabel('Dimension 2')\n plt.show()\n\nplot_som(som)\n<\/code><\/pre>\n\uacb0\ub860<\/h2>\n