14. A car is running A man is cutting a
piece of meat
A man is performing on
a stage
A man is riding a bike
A man is singing A panda is walking A woman is riding a horse A man is flying in a field
32. Facial Emotion Prediction
https://github.com/Microsoft/FERPlus https://arxiv.org/abs/1608.01041
Data Sources Model Train with Cloud AI Deploy Consume
ACTION
INTELLIGENCEDATA
Azure Machine
Learning
WindowsML
Model: VGG-13
Code:
CNTK
Face Emotion Recognition
12 emotion labels
28,709 training images
3,589 test images
3,589 validation images
Visual Studio
Tools for AI
Manage Models
Azure
Deep Learning GPU VM
34. Intelligent Disease Prediction
Data Sources Model Train with Cloud AI Deploy Consume
ACTION
INTELLIGENCEDATA
Azure Machine
Learning
IoT Hub
WindowsML
IOT Edge
Model:
DenseNet-121
Code:
Keras +
TensorFlow
National Institute
of Health
Chest Xray Data
112,120 images
14 pathology labels
30,805 unique patients
Visual Studio
Tools for AI
Manage Models
Azure
Deep Learning GPU VM
VSTS +
CI/CD
CosmosDB +
Azure Functions
NuGet
37. Pre-built ML Models (Azure Cognitive Services)
Easy / Less Control Full Control / Harder
Vision Speech Language
Knowledge SearchLabs
TextAnalyticsAPI client = new TextAnalyticsAPI();
client.AzureRegion = AzureRegions.Westus;
client.SubscriptionKey = "1bf33391DeadFish";
client.Sentiment(
new MultiLanguageBatchInput(
new List<MultiLanguageInput>()
{
new MultiLanguageInput("en","0",
"This is a great vacuum cleaner")
}));
例:Azure Cognitive Servicesによる感情分析
96% 好意的
38. Pre-built ML Models (Azure Cognitive Services)
Easy / Less Control Full Control / Harder
Vision Speech Language
Knowledge SearchLabs
TextAnalyticsAPI client = new TextAnalyticsAPI();
client.AzureRegion = AzureRegions.Westus;
client.SubscriptionKey = "1bf33391DeadFish";
client.Sentiment(
new MultiLanguageBatchInput(
new List<MultiLanguageInput>()
{
new MultiLanguageInput("en","0",
"This vacuum cleaner sucks so much dirt")
}));
9% 好意的
例:Azure Cognitive Servicesによる感情分析
39. Easy / Less Control
Full Control / Harder
データの準備 学習 & 評価 実行
56. DNN Processing Units
効率性柔軟性
Soft DPU
(FPGA)
Contr
ol
Unit
(CU)
Register
s
Arithmet
ic Logic
Unit
(ALU)
CPUs GPUs
ASICsHard
DPU
Cerebras
Google TPU
Graphcore
Groq
Intel Nervana
Movidius
Wave Computing
Etc.
BrainWave
Baidu SDA
Deephi Tech
ESE
Teradeep
Etc.
59. f f f
l0
l1
f f f
l0
Pretrained DNN モデル
CNTK などで
スケーラブルな DNN
ハードウェア マイクロサービス
BrainWave
Soft DPU
Instruction
Decoder & Ctrl
Neural FU
ネットワークスイッチ
FPGA
60. Web search
ranking
Traditional software (CPU) server plane
QPICPU
QSFP
40Gb/s ToR
FPGA
CPU
40Gb/s
QSFP QSFP
Hardware acceleration plane
Web search
ranking
Deep neural
networks
SDN offload
SQL
CPUs
FPGAs
Routers