次世代のデータ活用としてのIntelligent Data Platform －Microsoft Azure と AI (Artificial Intelligence)
•Download as PPTX, PDF•
2 likes•928 views
AIが何をどこまで出来るようになっているのか? そして、AIのエンジンである、機械学習部分を作るために、現在どんな道具があるのか。2017年9月時点の最新情報をお伝えします。
Recommended
More Related Content
What's hot
What's hot (20)
Similar to 次世代のデータ活用としてのIntelligent Data Platform －Microsoft Azure と AI (Artificial Intelligence)
Similar to 次世代のデータ活用としてのIntelligent Data Platform －Microsoft Azure と AI (Artificial Intelligence) (20)
More from Daiyu Hatakeyama
More from Daiyu Hatakeyama (20)
Recently uploaded
Recently uploaded (14)
次世代のデータ活用としてのIntelligent Data Platform －Microsoft Azure と AI (Artificial Intelligence)
- 12. with cntk.layers.default_options(activation=cntk.ops.relu, pad=False): conv1 = cntk.layers.Convolution2D((5,5), 32, pad=True)(scaled_input) pool1 = cntk.layers.MaxPooling((3,3), (2,2))(conv1) conv2 = cntk.layers.Convolution2D((3,3), 48)(pool1) pool2 = cntk.layers.MaxPooling((3,3), (2,2))(conv2) conv3 = cntk.layers.Convolution2D((3,3), 64)(pool2) f4 = cntk.layers.Dense(96)(conv3) drop4 = cntk.layers.Dropout(0.5)(f4) z = cntk.layers.Dense(num_output_classes, activation=None)(drop4) 例：手書き文字認識 (MNIST)
- 13. Microsoft Azure
- 15. Featuring / Training Prediction Data / Users Scalable Performance Throughput + Efficiency 1度のトレーニングで、 数十億 TFLOPS Xeon CPU だと、 1年がかり GPU だと1日 処理毎に数十億 FLOPS Xeon CPU だと数秒 GPU だと即座
- 17. ND6s ND12s ND24s ND24rs Cores 6 12 24 24 GPU 1 x P40 2 x P40 4 x P40 4 x P40 Memory 112 GB 224 GB 448 GB 448 GB Disk ~700 GB SSD ~1.4 TB SSD ~3 TB SSD ~3 TB SSD Network Azure Network Azure Network Azure Network InfiniBand
- 19. R User Workstation R Server for Hadoop RDDsHDInsight HDFS RDDs HDFS RDDs HDFS
- 22. Video Indexer Cognitive Services Labs
- 23. Video Indexer Cognitive Services Labs Custom Custom CustomCustom Custom
- 24. 従来の Database + App Intelligence Database + App Application + Intelligence Database Application Intelligence + Database VS
- 25. #azurejp Azure Data Lake Azure Media Service SQL Server 2017
- 26. Azure Data Lake service Big Data as a Services 無限にデータをストア・管理 Row Data を保存 高スループット、低いレイテンシの分析ジョ ブ セキュリティ、アクセスコントロール Azure Data Lake store HDInsight & Azure Data Lake Analytics
- 27. 1. 顔の解析 2. 画像への タグ 付け 3. 顔の感情分析 4. OCR 1. テキストからの重要語句の抽 出 2. テキストの感情分析 Azure Data Lake の 6つの Cognitive 機 能
- 28. • オブジェクト認識 (タグ) • 顔認識、感情認識 • JOIN処理 – 幸せな人は誰なのか? REFERENCE ASSEMBLY ImageCommon; REFERENCE ASSEMBLY FaceSdk; REFERENCE ASSEMBLY ImageEmotion; REFERENCE ASSEMBLY ImageTagging; @objects = PROCESS MegaFaceView PRODUCE FileName, NumObjects int, Tags string READONLY FileName USING new Cognition.Vision.ImageTagger(); @tags = SELECT FileName, T.Tag FROM @objects CROSS APPLY EXPLODE(SqlArray.Create(Tags.Split(';'))) AS T(Tag) WHERE T.Tag.ToString().Contains("dog") OR T.Tag.ToString().Contains("cat"); @emotion_raw = PROCESS MegaFaceView PRODUCE FileName string, NumFaces int, Emotion string READONLY FileName USING new Cognition.Vision.EmotionAnalyzer(); @emotion = SELECT FileName, T.Emotion FROM @emotion_raw CROSS APPLY EXPLODE(SqlArray.Create(Emotion.Split(';'))) AS T(Emotion); @correlation = SELECT T.FileName, Emotion, Tag FROM @emotion AS E INNER JOIN @tags AS T ON E.FileName == T.FileName; Images Objects Emotions filter join aggregate
- 29. 付加価値を提供する 多くのパートナー ソリューション Streaming / CDN コンテンツ 保護 Processing取り込み と保管 メディア 配信の コア 機能を API として提供 Azure Media Services PlayerAI (Artificial Intelligence)
- 31. Azure Media Service Azure Functions 翻訳 感情 キー ワード
- 33. R & Python ベースの AI の ストアドプロシージャ MicrosoftML Library の組み込み Graph モデルのサポートによる より複雑な関係を分析 クエリ 処理の最適化による 比類なき パフォーマンス AI を組み込んだ最初の商用データベース SQL Server 2017
- 34. • SQL Server, CNTK & R/Python – それぞれの強い部分を連携させた エンタープライズ グレードの AI アプリケーション • データを移動させる必要がないため、セキュリティと効率化が向上 • GPU による、処理能力の向上 Featurization classifier model での スコアリング Web App Diagnosis: 35% certainty Python / R で実装した ストアドプロシージャ ストアド プロシージャ の 呼び出し Model table, Features table, New Images table SQL Server
- 35. execute sp_execute_external_script @language = N'R' , @script = N' x <- as.matrix(InputDataSet); y <- array(dim1:dim2); OutputDataSet <- as.data.frame(x %*% y);' , @input_data_1 = N'SELECT [Col1] from MyData;' , @params = N'@dim1 int, @dim2 int' , @dim1 = 12, @dim2 = 15 WITH RESULT SETS (([Col1] int, [Col2] int, [Col3] int, [Col4] int)); ‘R’ もしくは ‘Python’ R file や Python file の 読み込みも 入力データ。 T-SQL SELECT も使 えるスクリプト用のパラメーター。”OUTPUT” もサポー ト。 トレーニング済みのモデルには varbinary(max) を 使用 Result set のバインド(Optional) STDOUT や STDERR と一緒に メッセージ文字列も R dataframe もしくは Python Pandas dataframe
- 36. launchpad.exe sp_execute_external_script sqlservr.exe Named pipe それぞれの SQL インスタ ンスがlaumchpad.exe を持 つ SQLOS XEvent MSSQLSERVER Service MSSQLLAUNCHPAD Service “何を” そして “どう” “実行” するか “launcher” Bxlserver.exe sqlsatellite.dll Bxlserver.exe sqlsatellite.dll Windows Satellite Process sqlsatellite.dll Run your “query”
- 37. In-memory OLTPColumn Store 大量データの 書き込み処理 PowerBI Dashboardスケールする Database 内の分析 R 業務ユーザー 分析の用意 Stored Predictions Visualize SQL-R を使って、秒間100万件の機械学習の処理も
- 38. Gathering Data Store Data Store Data Store Data Store Data Store Training Dataset Model Data Process Training Deploy Prediction Model Model Lang. Image Model Video Dataset Model Dataset Model Model Service
- 43. Azure IoT Edge IoT Hub Devices Local Storage Azure Machine Learning (Container) Functions RuntimeContainer Management Device Twin Device Twin Azure Stream Analytics (Container) Azure Functions (Container) Cognitive Services (Container) Custom Code (Container) Module Twin Module Twin Module Twin Module Twin Module Twin Module Twin Module Twin Module Twin Module Twin Module Twin
- 44. #azurejp
- 47. © 2017 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
Editor's Notes
- AI is not about creepy robots creating assembly-line healthcare. It is about systems that assist and support the wisdom and experience of well-trained clinicians in making better data-driven decisions and taking actions that best support the needs of those they serve. It does this by gathering and crunching massive amounts of data quickly and intelligently to identify patterns often overlooked or undiscovered in the traditional practice of care. The opportunity for AI in healthcare isn’t just about making doctors and healthcare providers more efficient in their work; it’s about making the lives of the patients better and saving lives. <click> Today’s health organizations need solutions that can help them engage patients in their health <click>, empower care teams <click>, optimize clinical operations <click>, transform the care continuum <click> and enhance security and compliance. Transition: AI can help you proactively engage your patients in their health. <click>
- Improve the patient experience throughout the care journey by empowering care teams with streamlined, 360-degree interactions. Through modern, patient-centered communications and efficient processes, health organizations can increase satisfaction and save time for patients and staff alike, while lowering costs. <click> Enable intelligent tools to recommend next best actions for individual care plans Reaching an accurate diagnosis can sometimes be a challenge even for the most skilled providers. However, the mass digitization of patient data via EMR and EHR systems opens the door for AI and machine learning to help augment the diagnostic process to build effective individual care plans. Through analysis of the patient’s digital health footprint, such as lab results, history and reported symptoms, advanced analytics tools can surface potential diagnoses and recommend the next best actions for care. Customer Story: https://customers.microsoft.com/en-US/story/delivering-data-to-doctors-to-enable-the-best-decisions-at-the-point-of-care Demonstration: (link) Products: Text Analysis API, Translator Text API, Recommendations API, ML Server <click> Engage digital assistants to record and transcribe patient history and chart notations Dictation tools have come a long way since the Dictaphone, and it’s not uncommon to find some type of dictation tool packaged into EMR or EHR systems. However, voice-to-text powered by AI is transforming the way providers can transcribe patient histories and chart notations. With blazingly fast voice-to-text and instant translation capabilities, AI can provide vital time-saving assistance to busy providers on-the-go, and across any device. Customer Story: (link) Demonstration: (link) Products: Translator Text API, Language Understanding Intelligent Service (LUIS) <click> Provide remote patient monitoring while applying analytics to generate care team alerts Remote monitoring via connected devices has changed the way that patients interact with their care teams. The ability to track biometrics with sensors allows patients to rest in the comfort of their own home, yet remain in a provider’s constant care. AI tools can further enhance the remote monitoring process by engaging chatbots to push notifications directly to care teams based on defined biometric triggers, enabling them to respond in real-time to changes in a patient’s status. Customer Story: https://customers.microsoft.com/en-US/story/zionchina Demonstration: (link) Products: Bot Framework, Text Analytics API, Translator Text API, Azure IoT Suite Transition: With AI solutions, you take steps to further optimize your organization for effectiveness. <click>
- Today’s AI technology has made massive leaps forward, and few industries have felt that more keenly than healthcare. Continued investment in AI technology is enabling modern health organizations to transform the care continuum in ways they never thought possible. <click> Enable virtual nursing assistants to remotely assess patient symptoms and decrease unnecessary visits When seeking medical care, patients are often required to see their provider in person for even minor concerns. A digital health organization can help lessen the number of unnecessary patient visits through AI. Using natural language, nursing assistant chatbots can interact with the patient and help assess their symptoms by asking decision-tree questions such as “Do you have a fever?” or “Are you coughing?” The chatbot can then route the conversation directly to a care provider or create a live telehealth session in order for the provider to make a determination on next steps for the patient. Cognitive capabilities can further expand virtual nursing services by providing real-time translation localized to where the patient lives. Customer Story: (link) Demonstration: (link) Products: Bot Framework, Text Analysis API, Translator Text API, Bing Speech API, Computer Vision API, Recommendations API, ML Server <click> Employ advanced learning models to expedite the medical imaging workflow and identify potential findings Today, one of the biggest problems facing physicians and clinicians in general is the overload of too much patient information to sift through. This is where AI can play a key role. AI will not be diagnosing patients and replacing doctors — it will be augmenting their ability to find the key, relevant data they need to care for a patient and present it in a concise, easily digestible format. When a radiologist calls up a scan to read, the AI will review the image and identify potential findings immediately—from the image and also by combing through the patient history related to the particular anatomy scanned—in order to suggest possible findings. Customer Story: (link) Demonstration: (link) Products: Text Analysis API, Computer Vision API, Recommendations API, ML Server <click> Combine cognitive robotic capabilities with medical records to guide and enhance physician instrument precision Cognitive robotics is still a newer field, but its advancing rapidly, and the impact to surgical teams could be significant. Cognitive robots use machine learning and other forms of AI, and can integrate information from pre-op medical records with real-time operating metrics to physically guide and enhance the physician’s instrument precision. The technology incorporates data from actual surgical experiences to inform new, improved techniques and insights. Using cognitive robotics can result not only in better outcomes but also reduce the length of patient hospital stays for health organizations. Customer Story: (link) Demonstration: (link) Products: Vision API, Text Analytics API, Entity Linking API, ML Server Transition: AI solutions can strengthen trust in security and compliance across the care continuum <click>