Machine Learning with Microsoft by Nalin Mujumdar
•Download as PPTX, PDF•
1 like•328 views
Sitecore User Group Bangalore meetup on 18th Nov 2017 at Indegene Bangalore. Machine Learning with Microsoft by Nalin Mujumdar
Recommended
More Related Content
What's hot
What's hot (16)
Similar to Machine Learning with Microsoft by Nalin Mujumdar
Similar to Machine Learning with Microsoft by Nalin Mujumdar (20)
More from Anindita Bhattacharya
More from Anindita Bhattacharya (20)
Recently uploaded
Recently uploaded (20)
Machine Learning with Microsoft by Nalin Mujumdar
- 3. Demo • Litware Insurance Bot
- 14. Define Objective Access and Understand the data Pre-processing
- 15. ML | AI Consume Build Machine Learning Consume Cortana Intelligence Gallery Machine Learning APIs Build Cloud Azure Machine Learning Studio R / Python / Drag & Drop Azure HD Insight SparkLib Mahout On Premise R SQL Server R Server Microsoft ML + AI Landscape
- 16. AI (Deep Learning) Consume Cognitive Services Vision Speech Language Knowledge Search Labs Build Cloud Custom Cognitive Services Cognitive Services Tool kit ML | AI Consume Build Microsoft ML + AI Landscape
- 19. Microsoft Cognitive Services Give your apps a human side Vision Speech KnowledgeLanguage LabsSearch Computer Vision Content Moderator Emotion Face Video Video Indexer Project Prague (gesture) Project Cuzco (events) Project Johannesburg (routing) Project Nanjing (isochrones) Project Abu Dhabi (distance matrix) Project Wollongong (location) Bing Autosuggest Bing Image Search Bing News Search Bing Video Search Bing Web Search Bing Entity Search Academic Knowledge Entity Linking Knowledge Exploration Recommendations QnA Maker Bing Spell Check Linguistic Analysis Text Analytics Translator Text & Speech Web Language Model Bing Speech Speaker Recognition
- 24. Roll your own with REST APIs Simple to add: just a few lines of code required Integrate into the language and platform of your choice Breadth of offerings helps you find the right API for your app Built by experts in their field from Microsoft Research, Bing, and Azure Machine Learning Quality documentation, sample code, and community support Easy Flexible Tested GET A KEY
- 27. MNIST Handwritten Digits (OCR) • Data set of hand written digits with 60,000 training images 10,000 test images • Each image is: 28 x 28 pixels • Performance with different classifiers (error rate): Neural nets (2-layers): 1.6 % Deep nets (6-layers): 0.35 % Conv nets (different): 0.21% - 0.31% Handwritten Digits 1 5 4 3 5 3 5 3 5 9 0 6 Corresponding Labels
- 28. 28 pix 28pix . 784 pixels (x) S S weights S = Sum (weights x pixels) 784 784 = map to (0-1) range Activation function 0.1 0.1 0.3 0.9 0.4 0.2 0.1 0.1 0.6 0.3 Probabilities (p) . 784 10 Model Logistic Regression 0.08 0.08 0.10 0.17 0.11 0.09 0.08 0.08 0.13 0.01 𝑒 𝑝i 𝑗=0 9 𝑒 𝑝j softmax SBias (b) + bias 0 1 9
- 29. 28 pix 28pix . 28 x 28 pix (p) Error or Loss Function Label One-hot encoded (Y) 0 0 0 1 0 0 0 0 0 0 0.1 0.1 0.3 0.9 0.4 0.2 0.1 0.1 0.6 0.3 Predicted Probabilities (p) Model (w, b) Loss function se = 𝑗= 0 9 𝑦𝑗 − 𝑝𝑗 2 Squared error ce = 𝑗=0 9 𝑦𝑗 𝑙𝑜𝑔 𝑝𝑗 Cross entropy error
- 30. Workflow MNIST Train .... 3 7 8 0 Input feature (X: 100 x 784) 100samples (mini-batch) Loss cross-entropy(softmax(out)) cross_entropy_with_softmax(out,Y) One-hot encoded Label (Y: 100 x 10) 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 . . . weights(W) (Dim-10x784) 784 10 bias(b) (Dim-10) Model S z = times(X , W) + b = Sum ( X . W) + b out = sigmoid(z) Trainer(model, loss, learner) Trainer.train_minibatch({X, Y}) Learner sgd, adagrad etc, are solvers to estimate – W & b
- 31. Deep learning at Microsoft • Microsoft Cognitive Services • Skype Translator • Cortana • Bing • Bing Ads • Augmented Reality • Microsoft Research
- 32. • CNTK is Microsoft’s open-source, cross-platform toolkit for learning and evaluating deep neural networks • CNTK expresses (nearly) arbitrary neural networks by composing simple building blocks into complex computational networks, supporting relevant network types and applications. • CNTK is production-ready: State-of-the-art accuracy, efficient, and scales to multi-GPU/multi-server. The Microsoft Cognitive Toolkit (CNTK)
- 38. Q & A
Editor's Notes
- Artificial Intelligence is term widely used nowadays but it lacks a common definition.
- Wikipedia: “Machines showing capabilities that are typically associated with human intelligence”.
- AI is not new. We’ve been in a journey for decades to augment human capabilities in one of those three areas. Think about the first calculators, which could do better than humans even with mechanical options like the Grant calculator in 1877. Electronics then opened a new frontier for AI. Natural experiences like Voder, a machine able to mimic human voice. Or first attempts on reasoning machines. Then computers, bringing graphical user interfaces, voice recognition or self driving cars as early as 1994. Then why are we talking about AI now? The progress all these decades has been linear, with even some setbacks. However, three things are dramatically changing the speed of innovation: Advances in AI algorithms, in particular deep learning with neural networks. Huge compute power available in the cloud. Big data, able to train these algorithms with huge sets of data.
- Bank of Kochi in Japan used Bot Framework and Cognitive Services to create an agent providing hospitality services at their bank branches, improving customer satisfaction.
- https://customers.microsoft.com/en-us/story/rollsroycestory https://customers.microsoft.com/en-us/story/microsoftfinance
- Why choose these APIs? They work, and it’s easy. Easy: The APIs are easy to implement because of the simple REST calls. Being REST APIs, there’s a common way to implement and you can get started with all of them for free simply by going to one place, one website, www.microsoft.com/cognitive. (You don’t have to hunt around to different places.) Flexible: We’ve got a breadth of intelligence and knowledge APIs so developers will be able to find what intelligence feature they need; and importantly, they all work on whatever language, framework, or platform developers choose. So, devs can integrated into their apps—iOS, Android, Windows—using their own tools they know and love (such as python or node.js, etc.). Tested: Tap into an ever-growing collection of powerful AI algorithms developed by experts. Developers can trust the quality and expertise build into each API by experts in their field from Microsoft’s Research organization, Bing, and Azure machine learning and these capabilities are used across many Microsoft first party products such as Cortana, Bing and Skype.
- Microsoft cognitive services uses CNTK for several tasks in speech, vision and text based deep learning applications. Microsoft customer service bot, face detection framework to name a few. Machine translation problems are very amenable to solutions based on deep learning. Skype translator efficiently uses such technologies. Cortana learns acoustic models using CNTK for scalable training on very large datasets. Bing uses the toolkit for relevance and ranking. Bing ads uses the toolkit for recognition of products in advertiser provided images in addition to other prediction and relevance problems. CNTK is used for object recognition in Augmented reality applications. Within Microsoft research CNTK has been pioneered and continues to extend the deep learning technology frontier.
- The toolkit is tested in production setting for accuracy, efficiency and scalability in a multi-GPU/multi-server environment. The toolkit is designed to bring lego like extensibility allowing for expression of arbitrary NN ……. Several use cases have been built using the toolkit for speech, image, text and combinations of the data types.
- [Optional slide for technical audiences]
- Cortana Intelligence delivers an end-to-end platform with an integrated and comprehensive set of tools and services to help you build intelligent applications that let you easily take advantage of Advanced Analytics and intelligence capabilities. First, Cortana Intelligence provides services to bring data in, so that you can analyze it. It provides information management capabilities like Azure Data Factory so that you can pull data from any source (relational DB like SQL or non-relational ones like your Hadoop cluster) in an automated and scheduled way, while performing the necessary data transforms (like setting certain data columns as dates vs. currency etc). Think ETL (Extract, Transform, Load) in the cloud. Event Hubs does the same for IoT type ingestion of data that streams in from lots of end points. The data brought in then can be persisted in flexible big data storage services like Data Lake Store and Azure SQL Data Warehouse. You can then use a wide range of analytics services from Machine Learning to Azure Data Lake Analytics to Azure HDInsight to Azure Stream Analytics to analyze the data stored in the big data storage. This means you can create analytics services and models specific to your business need (say real time demand forecasting). The resultant analytics services and models created by taking these steps can then be surfaced as interactive dashboards and visualizations via Power BI. These same analytics services and models created can also be integrated into various different UI (web apps or mobile apps or rich client apps), or with Cortana, so end users can naturally interact with them via speech etc., and so that end users can get proactively be notified by Cortana if the analytics model finds a new anomaly (unusual growth in certain product purchases- in the case of real time demand forecasting example given above) or whatever deserves the attention of the business users. Similar integration can occur with Cognitive Services or Bot Framework based applications. At a high level though, Cortana Intelligence capabilities are in three main areas: data, analytics and intelligence. <Transition>: We’re going to dive into each one, starting with data.