8. • Artificial NN (Hinton, et. al 1990)
• LTSM (Schmidhuber, et. al 1997)
• CNNs (LeCun, et. al 1998)
• Deep NN (Hinton, et. al 2006)
• Autoencoders (DVincent, et. al 2010)
• Reading Digits in Unsupervised Networks (Ng, et. al 2011)
• Random Search HPO (Bergstra, et. al 2012)
• Distributed Deep Networks (Dean, et. al 2012)
• ImageNet (Krizhevsky, et. al 2012)
• Drift Adaptation (Gama, et. al 2013)
• Deep CNNs (Gong, et. al 2013)
• RNNs (Sutskever, et. al 2013)
• Dropouts (Hinton, et. al 2014)
• Video Class. with CNN (Fei-Fei, et. al 2014)
• Scene Recognition (Zhou, et. al 2014)
• GANs (Goodfellow, et. al 2014)
• High Speed Classifiers (Henriques, et. al 2014)
• Multi-Label Learning (Zhang, et. al 2014)
• Transfer Learning (Yosinski, et. al 2014)
• FLANN (Muja, et. al 2014)
• SDS (Hariharan et. al 2014)
• 1ms Face Alignment (Kazemi, et. al 2014)
• GloVe (Pennington, et. al 2014)
• Neural Machine Trans. (Bahdanau, et. al 2014)
• Neural Turning Machines (Graves, et. al 2014)
• Batch Normalization (Ioffe, et. al 2015)
• COCO (Lin, et. al 2015)
• Deep RL (Wang, et. al 2015)
• Inception v4 (Szegedy et. al 2016)
• One-shot Generalization (Rezende, et. al 2016)
• Capsule Networks (Hinton, et. al 2017)
17. Azure
Machine Learning
Develop Your Own Model
20
https://docs.microsoft.com/ja-jp/azure/architecture/data-guide/technology-choices/data-science-and-machine-learning
https://medium.com/microsoftazure/9-advanced-tips-for-production-machine-learning-6bbdebf49a6f
Use Pre-trained Model
80
Azure
Cognitive Services
27. import
'--batch-size' type int
'batch_size' 'mini
batch size for training'
'--epoch' type int
'epoch' 'epoch size
for training’
from import
'--data-folder' 'mnist'
'--batch-size' 50
'--epoch' 20
'--first-layer-neurons' 300
'--second-layer-neurons' 100
'--learning-rate' 0.001
'--activation'
'--optimizer'
'--loss'
'--dropout' 0.2
'--gpu'
'keras' 'matplotlib'
‘train.py'
True
1800
#2. Script Folder
'./keras-mnist'
True
import
‘./train.py'
'./utils.py'
Docker Image
Data Store
28. from import
# start an Azure ML run
class LogRunMetrics
# callback at the end of every epoch
def on_epoch_end
# log a value repeated which creates a list
'Loss' 'loss'
'Accuracy' 'acc'
2
Experiment
37. Mileage
Condition
Car brand
Year of make
Regulations
…
Parameter 1
Parameter 2
Parameter 3
Parameter 4
…
Gradient Boosted
Nearest Neighbors
SVM
Bayesian Regression
LGBM
…
Mileage Gradient Boosted Criterion
Loss
Min Samples Split
Min Samples Leaf
Others Model
Which algorithm? Which parameters?Which features?
Car brand
Year of make
試行錯誤
38. Criterion
Loss
Min Samples Split
Min Samples Leaf
Others
N Neighbors
Weights
Metric
P
Others
Which algorithm? Which parameters?Which features?
Mileage
Condition
Car brand
Year of make
Regulations
…
Gradient Boosted
Nearest Neighbors
SVM
Bayesian Regression
LGBM
…
Nearest Neighbors
Model
繰り返し
Gradient BoostedMileage
Car brand
Year of make
Car brand
Year of make
Condition
39. Mileage
Condition
Car brand
Year of make
Regulations
…
Gradient Boosted
Nearest Neighbors
SVM
Bayesian Regression
LGBM
…
Gradient Boosted
SVM
Bayesian Regression
LGBM
Nearest Neighbors
Which algorithm? Which parameters?Which features?
繰り返し
Regulations
Condition
Mileage
Car brand
Year of make
71. • AI や 機械学習の最新の
トレーニング
• 概要・基礎・チュートリアル
• 自分に適した、トレーニングコースの作成
• AI Business School
• Conversational AI
• AI Services
• Machine Learning
• Autonomous System
• Responsible AI
aischool.microsoft.com