ニューラルãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã§æ™‚系列データã®äºˆæ¸¬ã‚’行ㆠ- Qiita
Hopfield networkã¯ã€ä¸€èˆ¬çš„ãªã‚¯ãƒ©ã‚¹åˆ†é¡žä»¥å¤–ã«æœ€é©åŒ–å•é¡Œã¸ã®å¿œç”¨ãŒå¯èƒ½ãªãƒ¢ãƒ‡ãƒ«ã§ã™ã€‚ Elman/Jordanã¯ã€Simple recurrent networksã¨è¨€ã‚ã‚Œã¦ã„るよã†ã«ä¸€ç•ªã‚·ãƒ³ãƒ—ルãªå½¢ã¨ãªã£ã¦ã„ã¾ã™ã€‚RNNを利用ã—ãŸã„å ´åˆã¯ã¾ãšã©ã¡ã‚‰ã‹ã§ã‚„ã£ã¦ã¿ã¦ã€ç²¾åº¦çš„ãªå•é¡ŒãŒã‚ã‚‹ã®ãªã‚‰ä»–ã®æ‰‹æ³•ã«åˆ‡ã‚Šæ›¿ãˆã¦ã¿ã‚‹ã€ã¨ã„ã†ã®ãŒã‚ˆã„ã®ã§ã¯ãªã„ã‹ã¨æ€ã„ã¾ã™ã€‚ Elman/Jordanã®é•ã„ã¯ä¸Šè¨˜ã®ã¨ãŠã‚Šã§ã™ãŒ(å‰å›žãƒ‡ãƒ¼ã‚¿ã®åæ˜ ãŒéš れ層ã‹ã‚‰è¡Œã‚れるã‹ã€å‡ºåŠ›å±¤ã‹ã‚‰è¡Œã‚れるã‹)ã€ã“ã¡ã‚‰ã«ã‚‚詳ã—ã書ã‹ã‚Œã¦ã„ã¾ã™ã€‚精度的ãªå„ªåŠ£ã¯ã‚ã‚Šã¾ã›ã‚“ãŒã€éš れ層ã®æ•°ã«ã‚ˆã£ã¦æ¬¡ã«ä¼æ’ã™ã‚‹é‡ã‚’変化ã•ã›ã‚‰ã‚Œã‚‹Elmanã®æ–¹ãŒæŸ”軟ã¨è¨€ãˆã‚‹ã¨æ€ã„ã¾ã™ã€‚ Echo state networkã¯æ¯›è‰²ãŒé•ã£ãŸãƒ¢ãƒ‡ãƒ«ã§ã€ãƒŽãƒ¼ãƒ‰ã‚’事å‰ã«çµåˆã›ãšReservoir(è²¯æ°´æ± ãªã©ã®æ„味)ã¨å‘¼ã°ã‚Œã‚‹ãƒ—ールã«è²¯ã‚ã¦ãŠãã€å…¥åŠ›ãŒä¸Žãˆã‚‰ã‚Œ
Keras Documentation
Keras is now available for JAX, TensorFlow, and PyTorch! Read the Keras 3.0 release announcement "Keras is one of the key building blocks in YouTube Discovery's new modeling infrastructure. It brings a clear, consistent API and a common way of expressing modeling ideas to 8 teams across the major surfaces of YouTube recommendations." Maciej Kula Staff Software Engineer - Google "Keras has tremendo
実践多クラス分類 Kaggle Ottoã‹ã‚‰å¦ã‚“ã ã“ã¨
3. kaggle Otto Group Product Classification Challenge IN:93個ã®ç‰¹å¾´é‡ã€OUT:9クラスã«åˆ†é¡ž train.csv 61878è¡Œ test.csv 144368è¡Œ å„クラスã§ã‚る確率をæ出ã—ã€log lossã§è©•ä¾¡ 3 5. Stratified K Fold train.csvã¯å„クラスã”ã¨ã«ãƒ‡ãƒ¼ã‚¿ãŒä¸¦ã‚“ã§ã„㟠→å˜ç´”ã«åˆ†å‰²ã—ã¦ãƒ†ã‚¹ãƒˆãƒ‡ãƒ¼ã‚¿ã«ã—ã¦ã¯ã„ã‘ãªã„ →Scikit-learnãŒè‰²ã€…ãªæ–¹æ³•ã‚’æä¾›ã—ã¦ã„ã‚‹ 例: StratifiedKFold：クラス比率をä¿ã¤ã‚ˆã†ã«åˆ†å‰² 5 def do_cross_validation(): model = make_model() cv = cross_validation.StratifiedKFold(ys) scores = cross_validation.cross_val_sc
Scikit-learnã§ãƒã‚¤ãƒ‘ーパラメータã®ã‚°ãƒªãƒƒãƒ‰ã‚µãƒ¼ãƒ - Qiita
# -*- coding: utf-8 -*- from sklearn import datasets from sklearn.cross_validation import train_test_split from sklearn.grid_search import GridSearchCV from sklearn.metrics import classification_report, confusion_matrix from sklearn.svm import SVC ## データã®èªã¿è¾¼ã¿ digits = datasets.load_digits() X = digits.data y = digits.target ## トレーニングデータã¨ãƒ†ã‚¹ãƒˆãƒ‡ãƒ¼ã‚¿ã«åˆ†å‰²ï¼Ž X_train, X_test, y_train, y_test = train_test_spli
Tatsu's Blog
Co-Founder, Ex-CTO of Connehito, Inc.Angel investorSoftware Engineer, Ph.D.Please see more details here PostsStartup in LATAM ⑤ : Mexico, the largest Spanish-speaking market in LATAM, shaping ambitious startupsMay 03, 2021 Read post →Startup in LATAM ④ : Unicorn from Colombia, fueled by the economic crisis in neighboring VenezuelaFeb 23, 2021 Read post →My view on the Good and Ugly of Impact Inves
勾é…ブースティングã«ã¤ã„ã¦ã–ã£ãã‚Šã¨èª¬æ˜Žã™ã‚‹ - About connecting the dots.
最近xgboostãŒã ã„ã¶æµè¡Œã£ã¦ã„ã‚‹ã‚ã‘ã§ã™ã‘ã©ï¼Œã“ã‚Œã¯Gradient Boosting（勾é…ブースティング）ã®é«˜é€ŸãªC++実装ã§ã™ï¼Žå¾“æ¥ä½¿ã‚ã‚Œã¦ãŸgbtより10å€é«˜é€Ÿã‚‰ã—ã„ã§ã™ï¼Žãã‚“ãªxgboostを使ã†ã«ã‚ãŸã£ã¦ï¼Œã¯ã¦ã©ã†ã„ã†ç†å±ˆã§å‹•ã„ã¦ã„ã‚‹ã‚‚ã®ã ã‚ã†ã¨æ€ã£ã¦ã„ã‚ã„ã‚文献をèªã‚“ã ã®ã§ã™ãŒï¼Œæ—¥æœ¬èªžã¯ãŠã‚ã‹ï¼Œè‹±èªžæ–‡çŒ®ã§ã‚‚ãã‚“ãªã«è³‡æ–™ãŒãªã‹ã£ãŸã®ã§ï¼Œã‚る程度概è¦ã‚’把æ¡ã™ã‚‹ã®ã«çµæ§‹æ™‚間を食ã„ã¾ã—ãŸï¼Ž ãã‚“ãªã‚ã‘ã§ï¼Œä»Šå›žã¯è‡ªåˆ†ã®ç†è§£ã®ç¢ºèªã‚‚å…¼ãã¦ï¼Œå‹¾é…ブースティングã«ã¤ã„ã¦ã–ã£ãã‚Šã¨èª¬æ˜Žã—ã¦ã¿ã‚ˆã†ã‹ã¨æ€ã„ã¾ã™ï¼Žã¨ã¯ã„ãˆç”Ÿç†è§£ãªéƒ¨åˆ†ã‚‚çµæ§‹ã‚ã‚‹ã®ã§ï¼Œãƒžã‚µã‚«ãƒªãŒé£›ã‚“ã§ããã†æ„ŸãŒå¤§ã„ã«ã—ã¦ã„ã¾ã™ï¼Žè…‘ã«è½ã¡ã‚‹çš„ãªç†è§£ã‚’優先ã—ã¦ã„ã‚‹ãŸã‚，数å¼ã¯ä¸€åˆ‡å‡ºã¦ãã¾ã›ã‚“． 勾é…ブースティングã¨ã¯ 複数ã®å¼±å¦ç¿’器を組ã¿åˆã‚ã›ã‚‹ã‚¢ãƒ³ã‚µãƒ³ãƒ–ルå¦ç¿’ã«ã¯ï¼Œã„ãã¤ã‹ã®æ‰‹æ³•ãŒã‚ã‚Šã¾ã™ãŒï¼Œãƒ–ースティングã¯é€æ¬¡çš„ã«å¼±å¦ç¿’器を構築ã—
パッケージユーザーã®ãŸã‚ã®æ©Ÿæ¢°å¦ç¿’(12)：Xgboost (eXtreme Gradient Boosting) - å…本木ã§åƒãデータサイエンティストã®ãƒ–ãƒã‚°
今やKaggleã‚„KDD cup以下åã ãŸã‚‹æ©Ÿæ¢°å¦ç¿’コンペã§çµ¶å¤§ãªäººæ°—を誇る分類器ã€Xgboost (eXtreme Gradient Boosting)。特ã«Kaggleã®Higgs Boson Machine Learning Challengeã®å„ªå‹ãƒãƒ¼ãƒ ãŒé§†ä½¿ã—ãŸã“ã¨ã§æœ‰åã«ãªã£ãŸæ„ŸãŒã‚るよã†ã§ã€‚ ãã®å®Ÿè£…ã§ã™ãŒã€C++ベースã§é«˜é€ŸåŒ–ã—ãŸã‚‚ã®ã«Rã¨Pythonã®ãƒ©ãƒƒãƒ‘ーをã¤ã‘ãŸã‚‚ã®ãŒæ—¢ã«GitHubã§å…¬é–‹ã•ã‚Œã¦ã„ã¾ã™ã€‚ Rパッケージã§ã‚ã‚‹{xgboost}ã®ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã«ã¤ã„ã¦ã®æ³¨æ„äº‹é …ã¯å‰å›žã®è¨˜äº‹ã«æ›¸ã„ã¦ã„ã¾ã™ã®ã§ã€ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã®éš›ã¯ã”å‚考ã«ã—ã¦ã„ãŸã ã‘ã‚Œã°ã¨ã€‚ ã•ã¦ã€‚ã“ã‚Œã ã‘大人気ã®åˆ†é¡žå™¨ã¨ãªã‚‹ã¨å›½å†…外å•ã‚ãšè‰²ã€…ãªè§£èª¬è¨˜äº‹ã«æº¢ã‚Œã¦ãŠã‚Šã¾ã—ã¦ã€ä¾‹ãˆã°æ—¥æœ¬èªžãƒ–ãƒã‚°ã§ã‚‚ã“ã‚Œã ã‘ã®è¨˜äº‹ãŒæ—¢ã«å‡ºã¦ã¾ã™ã€‚ 勾é…ブースティングã«ã¤ã„ã¦ã–ã£ãã‚Šã¨èª¬æ˜Žã™ã‚‹ - About connecting
Caffe | Deep Learning Framework
Caffe Caffe is a deep learning framework made with expression, speed, and modularity in mind. It is developed by Berkeley AI Research (BAIR) and by community contributors. Yangqing Jia created the project during his PhD at UC Berkeley. Caffe is released under the BSD 2-Clause license. Check out our web image classification demo! Why Caffe? Expressive architecture encourages application and innovat
ＢＥＳＯï¼ï¼ˆãƒ“ーソム）ブãƒã‚° LSTM ã€ã¡ã‚‡ã£ã¨ã ã‘ç†è§£ã‚’試ã¿ã‚‹
ＢＥＳＯï¼ï¼ˆãƒ“ーソム）ã¯ã€ãƒ™ã‚¤ã‚¸ã‚¢ãƒ³ãƒãƒƒãƒˆã‚’用ã„ã¦å¤§è„³çš®è³ªã®æƒ…å ±å‡¦ç†ã‚’å†ç¾ã™ã‚‹æ©Ÿæ¢°å¦ç¿’アルゴリズムã§ã™ã€‚ＢＥＳＯï¼ã«é–¢é€£ã™ã‚‹ç ”究者å‘ã‘ã®æƒ…å ±ç™ºä¿¡ã‚’ã—ã¦ã„ãã¾ã™ã€‚（内容ã¯å€‹äººã®æ„見ã§ã‚り所属ã™ã‚‹çµ„ç¹”ã®è¦‹è§£ã§ã¯ã‚ã‚Šã¾ã›ã‚“。） 時系列データをå¦ç¿’ã™ã‚‹ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã®ä¸€ç¨®ã€ LSTM ã«ã¤ã„ã¦ã€ ã¡ã‚‡ã£ã¨ã ã‘ç†è§£ã—よã†ã¨è©¦ã¿ãŸã®ã§ã€è‡ªåˆ†ç”¨ã®ãƒ¡ãƒ¢ã‚’ã“ã“ã«æ›¸ã„ã¦ãŠãã¾ã™ã€‚ 「Long Short-Term Memory in Recurrent Neural Networks〠http://www.felixgers.de/papers/phd.pdf ドイツ Hannover 大〠FELIX GERS ã® Thesis, 2001 。 p.11 ã«ã€Œå¾“æ¥ã®ï¼ˆåˆæœŸã®ï¼‰ï¼¬ï¼³ï¼´ï¼ã€ã®å…¨ä½“åƒãŒæ›¸ã‹ã‚Œã¦ã„る。 ãŒã€ãƒ¡ãƒ¢ãƒªãƒ–ãƒãƒƒã‚¯ãŒï¼‘ã¤ã«çœç•¥ã•ã‚Œã¦ã„ã‚‹ã®ã§å…¨ä½“åƒãŒåˆ†ã‹ã‚Šã«ãã„。 （ＲＮＮã®ä¸é–“層ã®å„ユニッ
Torch | Getting started with Torch
Getting started with Torch Edit on GitHub Installing Torch We provide a simple installation process for Torch on Mac OS X and Ubuntu 12+: Torch can be installed to your home folder in ~/torch by running these three commands: # in a terminal, run the commands WITHOUT sudo git clone https://github.com/torch/distro.git ~/torch --recursive cd ~/torch; bash install-deps; ./install.sh The first script i
Torch7ã®åˆ†ã‹ã‚Šã«ãã„話 - デー
FacebookãŒtorchã®æ‹¡å¼µã‚’公開ã—ãŸã“ã¨ã‚‚ã‚ã‚Šã€torchを使ã£ã¦ã¿ã¦ã„る方も出始ã‚ãŸã‚ˆã†ãªã®ã§ã€è‡ªåˆ†ãŒtorchã§ä½¿ã†ä¸Šã§åˆ†ã‹ã‚Šã«ãã‹ã£ãŸã“ã¨ã‚„ãƒãƒžã£ãŸã“ã¨ãªã©ã‚’上ã’ã¦ã„ãã¾ã™ã€‚ ã‚る程度使ã£ã¦ã„る方をå‰æã¨ã—ã¦ã„ã¾ã™ã€‚ ã²ã©ã„ã“ã¨ã‚’ãŸãã•ã‚“書ãã¾ã™ãŒã€torchãŒå«Œã„ãªã‚ã‘ã§ã¯ã‚ã‚Šã¾ã›ã‚“。torchã¯ã‚µã‚¤ã‚³ãƒ¼ã§ã™ã€‚ リリースやãƒãƒ¼ã‚¸ãƒ§ãƒ³ã«ã¤ã„㦠ezinstallã¯gitã®ãƒªãƒã‚¸ãƒˆãƒªã‚’cloneã—ã¦ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã—ã¾ã™ã€‚torchã¯ä»Šã®ã¨ã“ã‚リリースã®ç®¡ç†ãªã©ã—ã¦ã„ãªãã¦ã€æ—¥ã€…æ›´æ–°ã•ã‚Œã‚‹é–‹ç™ºãƒªãƒã‚¸ãƒˆãƒªãŒã‚ã‚‹ã ã‘ãªã®ã§ã€ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã—ãŸæ—¥ã«ã‚ˆã£ã¦ç•°ãªã‚‹ãƒãƒ¼ã‚¸ãƒ§ãƒ³ãŒã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã•ã‚Œã¾ã™ã€‚ ãŸã¾ã«ä»¥ä¸‹ã®ã‚ˆã†ãªã‚³ãƒžãƒ³ãƒ‰ã‚’実行ã—ã¦æ›´æ–°ã—ãŸã»ã†ãŒã„ã„ã§ã™ã€‚ #!/bin/sh sudo luarocks install cwrap sudo luarocks install torch su
Torch7ã§è¤‡é›‘ãªãƒ¢ãƒ‡ãƒ«ã‚’書ãã¨ãã«ä¾¿åˆ©ãªæŠ€ - デー
Torch7ã¯nnã«ç”¨æ„ã•ã‚Œã¦ã„る部å“を組ã¿åˆã‚ã›ã‚‹ã“ã¨ã§è¤‡é›‘ãªãƒ¢ãƒ‡ãƒ«ã‚’作れã¦ä¾¿åˆ©ãªã®ã§ã™ãŒã€è¤‡é›‘ãªãƒ¢ãƒ‡ãƒ«ã¯å…¥åŠ›ãƒ™ã‚¯ãƒˆãƒ«(orテンソル)を様々形ã«å¤‰æ›ã—ãªãŒã‚‰å®Ÿè¡Œã™ã‚‹ã®ã§ã€ã©ã®æ™‚点ã§ã©ã®ã‚µã‚¤ã‚ºã«ãªã£ã¦ã„ã‚‹ã‹åˆ†ã‹ã‚‰ãªãã¦æ›¸ãã®ãŒé›£ã—ã„ã¨ã„ã†ã®ãŒã‚ã‚Šã¾ã™ã€‚ ãŸã¨ãˆã°ã€ä¼çµ±çš„ãªCNNã¯ã€Torch7ã§ä»¥ä¸‹ã®ã‚ˆã†ã«æ›¸ãã®ã§ã™ãŒ require 'nn' function cnn_model() local model = nn.Sequential() -- convolution layers model:add(nn.SpatialConvolutionMM(3, 128, 5, 5, 1, 1)) model:add(nn.ReLU()) model:add(nn.SpatialMaxPooling(2, 2, 2, 2)) model:add(nn.SpatialConvolutionMM
[MLAC 2013 7日目] Torch7ã§ãŠæ‰‹è»½ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ - a lonely miner
ã¯ã˜ã‚ã« ã“ã®è¨˜äº‹ã¯ Machine Learning Advent Calendar 2013 ã® 7日目ã®è¨˜äº‹ã§ã™ï¼Ž 2013年，Deep Learning もアカデミックレベルã§ã¯ã•ã¾ã–ã¾ãªåˆ†é‡Žã¸ã®æµ¸é€ãŒé€²ã¿ï¼Œãƒã‚ºãƒ¯ãƒ¼ãƒ‰ã®åŸŸã‚’脱ã—ãŸã‚ˆã†ã«æ€ãˆã¾ã™ï¼Ž ã“ã‚Œã¾ã§ã¯ï¼Œæ©Ÿæ¢°å¦ç¿’ã¨ã„ã†ã¨ï¼Œå¿œç”¨åˆ†é‡Žã«ãŠã„ã¦ã¯(分類/回帰ã¨ã„ã£ãŸ)タスクを決ã‚ã¦ã—ã¾ã£ãŸä¸Šã§ãƒ–ラックボックスã¨ã—㦠扱ã†ã‚‚ã®ï¼Œã¨ã„ã†ç©ºæ°—ãŒã‚ã‚Šã¾ã—ãŸãŒï¼Œ Deep Learning ã«ä»£è¡¨ã•ã‚Œã‚‹æŸ”軟ãªï¼Œã„ã‹ã‚ˆã†ã«ã‚‚組ã¿ä¸Šã’られるモデルã¯ï¼Œå•é¡Œã®æ€§è³ªã‚’ç©æ¥µçš„ã«(特徴é‡ã¨ã—ã¦ã§ã¯ãªã，モデル自体ã«)組ã¿è¾¼ã‚€ã“ã¨ã‚’容易ã«ã™ã‚‹ï¼Œå¤§ããªæ¦å™¨ã«ãªã‚‹ã®ã§ã¯ãªã„ã‹ã¨æ„Ÿã˜ã¦ã„ã¾ã™ï¼Ž ç´ æ€§ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ãƒªãƒ³ã‚° vs モデルパラメータエンジニアリング ã®ä¸æ¯›ãªæˆ¦ã„ãŒå¹•ã‚’上ã’ãŸï¼Œã¨ã„ã†è¦‹æ–¹ã‚‚ã§ãã¡ã‚ƒã„ãã†ã§ã™ãŒãƒ»ãƒ»ãƒ»ï¼Žï¼Ž ã•ã¦ä»Šå›žã¯ï¼Œ Torch7 ã¨ã„ã†ï¼ŒNe
![[MLAC 2013 7日目] Torch7ã§ãŠæ‰‹è»½ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ - a lonely miner](https://cdn-ak-scissors.b.st-hatena.com/image/square/be0f4e8c7e927775c6d1d3ce704d32b08e1915a3/height=288;version=1;width=512/http%3A%2F%2Fconditional.github.io%2Fimages%2Ftorch%2Ffeatures.png)
1
リリースã€éšœå®³æƒ…å ±ãªã©ã®ã‚µãƒ¼ãƒ“スã®ãŠçŸ¥ã‚‰ã›
最新ã®äººæ°—エントリーã®é…ä¿¡
j次ã®ãƒ–ックマーク
kå‰ã®ãƒ–ックマーク
lã‚ã¨ã§èªã‚€
eコメント一覧を開ã
oページを開ã
https://dl.acm.org/doi/10.1145/1143844.1143891
{{#tags}}- {{label}}
{{/tags}}