Anthill Inside 2019

On infrastructure for AI and ML: from managing training data to data storage, cloud strategy and costs of developing ML models

Explainable AI: Behind the Scenes

Submitted by Manjunath (@manjunath-123) on Tuesday, 30 April 2019

Section: Tutorials Technical level: Intermediate Session type: Tutorial

Abstract

With the Ai based systems proliferating in to all applications of the daily life, getting a better insight of their working mechanism is a much sought-after affair. Often the rationale behind the decision thrown out by an AI system is not well understood. Which part or feature of the input has influenced the decision to what extent is not known. This presentation provides insights to demystify the black box of the AI models
Research community and students: The tutorial provides better insights to the working of the deep learning models which otherwise work as black boxes. The different techniques for generating the heat map would be detailed with examples. Insights to the explanation text generation mechanism would be provided

AI & ML solution providers: They can pick up the steps in the workflow of the explanation generation and adopt the same in the design of the solution for deep learning based business problems.

Architects of products: The layer wise and neuron wise visualization of the activations in the network would provide information on importance of a neuron in decision making. Accordingly, the rest of the network may be pruned. The test cases to be designed are to activate the neurons under consideration. These concepts would be illustrated with the example of a case study involving misclassification identification in OCR.

Software vendors: The different components of explanation such as heat map, text explanation, misclassification identification etc. throw open multiple options for software vendor to provide them as a service These options would be illustrated with examples.

Design engineers: Provides a platform to crosscheck the design and debug the code while realizing the solution

Tool vendors: Helps to develop tools required to support the end to end workflow of explainable solution

Outline

Introduction: Background, positioning of the problem, examples

Relevance based explanation: Heat map generation, examples, Layer wise relevance propagation, Sensitivity analysis

Text explanation: Training of the LSTM models, vocabulary generation, sentence formation

Visualization of the layers: unwinding of the black box, layer wise activation generation, misclassification identification case study

Requirements

Note book, Pen

Speaker bio

Dr. Manjunath Ramachandra is working as Principal consultant at the AI research group of Wipro Limited. He has a blend of academic and industry experience over two decades. He has filed about 50 patents, chaired 33 conferences, conducted 17 tutorials and workshops, authored 180 Research papers in international conferences and journals and a book. He represented the industry in international standardization bodies such as Wi-Fi Alliance, served as the editor for the regional profiles standard in Digital living network alliance (DLNA) and as the industry liaison officer for the CE-Linux Forum. His areas of research include deep learning, NLP, AI applications etc.

Comments

  • Abhishek Balaji (@booleanbalaji) Reviewer a month ago

    Hi Manjunath,

    Thank you for submitting a proposal. For us to evaluate your proposal, we need to see detailed slides and a preview video. Your slides must take the following points into consideration:

    • Problem statement/context, which the audience can relate to and understand. The problem statement has to be a problem (based on this context) that can be generalized for all.
    • What were the tools/options available in the market to solve this problem? How did you evaluate these, and what metrics did you use for the evaluation? Why did you decide to build your own ML model?
    • Why did you pick the option that you did?
    • Explain how the situation was before the solution you picked/built and how was the fraud/ghosting after implementing the solution you picked and built? Show before-after scenario comparisons & metrics.
    • What compromises/trade-offs did you have to make in this process?
    • What are the privacy, regulatory and ethical considerations when building this solution?
    • What is the one takeaway that you want participants to go back with at the end of this talk? What is it that participants should learn/be cautious about when solving similar problems?

    As next steps, we’d need to see the detailed and/or updated slides by 21 May, in order to close the decision on your proposal. If we dont receive an update by 21 May, we’d have to move the proposal for consideration for a future conference.

  • manjunath ramachandra 29 days ago
    • Abhishek Balaji (@booleanbalaji) Reviewer 20 days ago

      Hi Manjunath,

      You’ll need to make both these links public, so reviewers can access them.

  • manjunath ramachandra 19 days ago
  • Abhishek Balaji (@booleanbalaji) Reviewer 17 days ago

    Hi Manjunath, these have to be accessible publicly. We’d recommend using Google Drive to upload these two and share public view links over here for the reviews.

Login with Twitter or Google to leave a comment