Pre-Conf Workshop

Wed, Aug 7
09:30

    Registration - 30 mins

10:00
  • Added to My Schedule
    keyboard_arrow_down
    Dipanjan Sarkar

    Dipanjan Sarkar / Anuj Gupta - A Hands-on Introduction to Natural Language Processing

    schedule  10:00 AM - 06:00 PM place Jupiter 1 people 2 Interested

    Data is the new oil and unstructured data, especially text, images and videos contain a wealth of information. However, due to the inherent complexity in processing and analyzing this data, people often refrain from spending extra time and effort in venturing out from structured datasets to analyze these unstructured sources of data, which can be a potential gold mine. Natural Language Processing (NLP) is all about leveraging tools, techniques and algorithms to process and understand natural language-based data, which is usually unstructured like text, speech and so on. In this workshop, we will be looking at tried and tested strategies, techniques and workflows which can be leveraged by practitioners and data scientists to extract useful insights from text data.

    Being specialized in domains like computer vision and natural language processing is no longer a luxury but a necessity which is expected of any data scientist in today’s fast-paced world! With a hands-on and interactive approach, we will understand essential concepts in NLP along with extensive case- studies and hands-on examples to master state-of-the-art tools, techniques and frameworks for actually applying NLP to solve real- world problems. We leverage Python 3 and the latest and best state-of- the-art frameworks including NLTK, Gensim, SpaCy, Scikit-Learn, TextBlob, Keras and TensorFlow to showcase our examples.

    In my journey in this field so far, I have struggled with various problems, faced many challenges, and learned various lessons over time. This workshop will contain a major chunk of the knowledge I’ve gained in the world of text analytics and natural language processing, where building a fancy word cloud from a bunch of text documents is not enough anymore. Perhaps the biggest problem with regard to learning text analytics is not a lack of information but too much information, often called information overload. There are so many resources, documentation, papers, books, and journals containing so much content that they often overwhelm someone new to the field. You might have had questions like ‘What is the right technique to solve a problem?’, ‘How does text summarization really work?’ and ‘Which are the best frameworks to solve multi-class text categorization?’ among many other questions! Based on my prior knowledge and learnings from publishing a couple of books in this domain, this workshop should help readers avoid the pressing issues I’ve faced in my journey so far and learn the strategies to master NLP.

    This workshop follows a comprehensive and structured approach. First it tackles the basics of natural language understanding and Python for handling text data in the initial chapters. Once you’re familiar with the basics, we cover text processing, parsing and understanding. Then, we address interesting problems in text analytics in each of the remaining chapters, including text classification, clustering and similarity analysis, text summarization and topic models, semantic analysis and named entity recognition, sentiment analysis and model interpretation. The last chapter is an interesting chapter on the recent advancements made in NLP thanks to deep learning and transfer learning and we cover an example of text classification with universal sentence embeddings.

  • Added to My Schedule
    keyboard_arrow_down
    Viral B. Shah

    Viral B. Shah / Abhijith - Computational Machine Learning

    schedule  10:00 AM - 06:00 PM place Jupiter 2

    You have been hearing about machine learning (ML) and artificial intelligence (AI) everywhere. You have heard about computers recognizing images, generating speech, natural language, and beating humans at Chess and Go.

    The objectives of the workshop:

    1. Learn machine learning, deep learning and AI concepts

    2. Provide hands-on training so that students can write applications in AI

    3. Provide ability to run real machine learning production examples

    4. Understand programming techniques that underlie the production software

    The concepts will be taught in Julia, a modern language for numerical computing and machine learning - but they can be applied in any language the audience are familiar with.

    Workshop will be structured as “reverse classroom” based laboratory exercises that have proven to be engaging and effective learning devices. Knowledgeable facilitators will help students learn the material and extrapolate to custom real world situations.

ODSC India Day 1

Thu, Aug 8
08:30

    Registration - 30 mins

09:00
  • schedule  09:00 - 09:45 AM place Grand Ball Room people 1 Interested

    Since we originally proposed the need for a first-class language, compiler and ecosystem for machine learning (ML) - a view that is increasingly shared by many, there have been plenty of interesting developments in the field. Not only have the tradeoffs in existing systems, such as TensorFlow and PyTorch, not been resolved, but they are clearer than ever now that both frameworks contain distinct "static graph" and "eager execution" interfaces. Meanwhile, the idea of ML models fundamentally being differentiable algorithms – often called differentiable programming – has caught on.

    Where current frameworks fall short, several exciting new projects have sprung up that dispense with graphs entirely, to bring differentiable programming to the mainstream. Myia, by the Theano team, differentiates and compiles a subset of Python to high-performance GPU code. Swift for TensorFlow extends Swift so that compatible functions can be compiled to TensorFlow graphs. And finally, the Flux ecosystem is extending Julia’s compiler with a number of ML-focused tools, including first-class gradients, just-in-time CUDA kernel compilation, automatic batching and support for new hardware such as TPUs.

    This talk will demonstrate how Julia is increasingly becoming a natural language for machine learning, the kind of libraries and applications the Julia community is building, the contributions from India (there are many!), and our plans going forward.

10:00

    Welcome Address & Conference Overview - 30 mins

10:30

    Coffee/Tea Break - 30 mins

11:00
  • schedule  11:00 - 11:45 AM place Grand Ball Room 1 people 1 Interested

    In recent years, there has been a lot of research in the area of sequence to sequence learning with neural network models. These models are widely used for applications such as language modeling, translation, part of speech tagging, and automatic speech recognition. In this talk, we will give an overview of sequence to sequence learning, starting with a description of recurrent neural networks (RNNs) for language modeling. We will then explain some of the drawbacks of RNNs, such as their inability to handle input and output sequences of different lengths, and describe how encoder-decoder networks, and attention mechanisms solve these problems. We will close with some real-world examples, including how encoder-decoder networks are used at LinkedIn.

  • Added to My Schedule
    keyboard_arrow_down
    Nicolas Dupuis

    Nicolas Dupuis - Using Deep-Learning to Accurately Diagnose Your Broadband Connection

    schedule  11:00 - 11:45 AM place Grand Ball Room 2

    Within Nokia Software Digital Experience, we build products that increase customer satisfaction and reduce churn through proactive identification of the user problems and that enable service providers to resolve problems faster. To achieve such tasks, ML and DL techniques are now contributing a lot to these successes. However, there is usually a long journey between building a first model up-to delivering a field-proven product. Besides providing highlights on how machine and deep learning are used today to boost the broadband connection, this talk will reveal some challenges encountered and best-practices involved to reach the expected quality level.

  • Added to My Schedule
    keyboard_arrow_down
    Favio Vázquez

    Favio Vázquez - Complete Data Science Workflows with Open Source Tools

    schedule  11:00 AM - 12:30 PM place Jupiter

    Cleaning, preparing , transforming, exploring data and modeling it's what we hear all the time about data science, and these steps maybe the most important ones. But that's not the only thing about data science, in this talk you will learn how the combination of Apache Spark, Optimus, the Python ecosystem and Data Operations can form a whole framework for data science that will allow you and your company to go further, and beyond common sense and intuition to solve complex business problems.

12:00
  • Added to My Schedule
    keyboard_arrow_down
    Dat Tran

    Dat Tran - Image ATM - Image Classification for Everyone

    schedule  12:00 - 12:45 PM place Grand Ball Room 1

    At idealo.de we store and display millions of images. Our gallery contains pictures of all sorts. You’ll find there vacuum cleaners, bike helmets as well as hotel rooms. Working with huge volume of images brings some challenges: How to organize the galleries? What exactly is in there? Do we actually need all of it?

    To tackle these problems you first need to label all the pictures. In 2018 our Data Science team completed four projects in the area of image classification. In 2019 there were many more to come. Therefore, we decided to automate this process by creating a software we called Image ATM (Automated Tagging Machine). With the help of transfer learning, Image ATM enables the user to train a Deep Learning model without knowledge or experience in the area of Machine Learning. All you need is data and spare couple of minutes!

    In this talk we will discuss the state-of-art technologies available for image classification and present Image ATM in the context of these technologies. We will then give a crash course of our product where we will guide you through different ways of using it - in shell, on Jupyter Notebook and on the Cloud. We will also talk about our roadmap for Image ATM.

  • Added to My Schedule
    keyboard_arrow_down
    Badri Narayanan Gopalakrishnan

    Badri Narayanan Gopalakrishnan / Shalini Sinha / Usha Rengaraju - Lifting Up: Deep Learning for implementing anti-hunger and anti-poverty programs

    schedule  12:00 - 12:45 PM place Grand Ball Room 2 people 1 Interested

    Ending poverty and zero hunger are top two goals United Nations aims to achieve by 2030 under its sustainable development program. Hunger and poverty are byproducts of multiple factors and fighting them require multi-fold effort from all stakeholders. Artificial Intelligence and Machine learning has transformed the way we live, work and interact. However economics of business has limited its application to few segments of the society. A much conscious effort is needed to bring the power of AI to the benefits of the ones who actually need it the most – people below the poverty line. Here we present our thoughts on how deep learning and big data analytics can be combined to enable effective implementation of anti-poverty programs. The advancements in deep learning , micro diagnostics combined with effective technology policy is the right recipe for a progressive growth of a nation. Deep learning can help identify poverty zones across the globe based on night time images where the level of light correlates to higher economic growth. Once the areas of lower economic growth are identified, geographic and demographic data can be combined to establish micro level diagnostics of these underdeveloped area. The insights from the data can help plan an effective intervention program. Machine Learning can be further used to identify potential donors, investors and contributors across the globe based on their skill-set, interest, history, ethnicity, purchasing power and their native connect to the location of the proposed program. Adequate resource allocation and efficient design of the program will also not guarantee success of a program unless the project execution is supervised at grass-root level. Data Analytics can be used to monitor project progress, effectiveness and detect anomaly in case of any fraud or mismanagement of funds.

12:45

    Lunch - 60 mins

01:45
  • Added to My Schedule
    keyboard_arrow_down
    Dr. Dakshinamurthy V Kolluru

    Dr. Dakshinamurthy V Kolluru - Understanding Text: An exciting journey from Probabilistic Models to Neural Networks

    schedule  01:45 - 02:30 PM place Grand Ball Room 1 people 1 Interested

    We will trace the journey of NLP over the past 50 odd years. We will cover chronologically Hidden Markov Models, Elman networks, Conditional Random Fields, LSTMs, Word2Vec, Encoder-Decoder models, Attention models, transfer learning in text and finally transformer architectures. Our emphasis is going to be on how the models became powerful and simple to implement simultaneously. To demonstrate this, we take a few case studies solved at INSOFE with a primary goal of retaining accuracy while simplifying engineering. Traditional methods will be compared and contrasted against modern models and show how the latest models actually are becoming easier to implement by the business. We also explain how this enhanced comfort with text data is paving way for state of the art inclusive architectures

  • Added to My Schedule
    keyboard_arrow_down
    Dipanjan Sarkar

    Dipanjan Sarkar - Explainable Artificial Intelligence - Demystifying the Hype

    schedule  01:45 - 02:30 PM place Grand Ball Room 2

    The field of Artificial Intelligence powered by Machine Learning and Deep Learning has gone through some phenomenal changes over the last decade. Starting off as just a pure academic and research-oriented domain, we have seen widespread industry adoption across diverse domains including retail, technology, healthcare, science and many more. More than often, the standard toolbox of machine learning, statistical or deep learning models remain the same. New models do come into existence like Capsule Networks, but industry adoption of the same usually takes several years. Hence, in the industry, the main focus of data science or machine learning is more ‘applied’ rather than theoretical and effective application of these models on the right data to solve complex real-world problems is of paramount importance.

    A machine learning or deep learning model by itself consists of an algorithm which tries to learn latent patterns and relationships from data without hard-coding fixed rules. Hence, explaining how a model works to the business always poses its own set of challenges. There are some domains in the industry especially in the world of finance like insurance or banking where data scientists often end up having to use more traditional machine learning models (linear or tree-based). The reason being that model interpretability is very important for the business to explain each and every decision being taken by the model.However, this often leads to a sacrifice in performance. This is where complex models like ensembles and neural networks typically give us better and more accurate performance (since true relationships are rarely linear in nature).We, however, end up being unable to have proper interpretations for model decisions.

    To address and talk about these gaps, I will take a conceptual yet hands-on approach where we will explore some of these challenges in-depth about explainable artificial intelligence (XAI) and human interpretable machine learning and even showcase with some examples using state-of-the-art model interpretation frameworks in Python!

  • Added to My Schedule
    keyboard_arrow_down
    Amar Lalwani

    Amar Lalwani - AI in Education: Transforming Education using Personalised Adaptive Learning

    schedule  01:45 - 02:30 PM place Jupiter

    There has been a significant rise in the gross enrolment ratio of the students in public schools over the past few decades. However, there is a decline in their learning outcomes, which results from staff crunch, crowded classrooms and insufficient infrastructure. Moreover, students are learning less as they move to higher classes. National Achievement Survey – 2017 shows that the national average score of a grade 8 student was barely 40% in Maths, Science and Social Studies. The survey also highlights the fact the country is short of at least 10 lakhs qualified teachers. With the advent of technology and AI, Personalised Adaptive Learning solutions might solve the current education crisis.

    With the belief that every child is unique, funtoot, an Intelligent Tutoring System designs a personalised learning path for each child. Funtoot tailors the teaching instructions according to the knowledge states of each learner and leads the learner towards her unique learning trajectory. Funtoot is used by more than 1.5 lakh school students (Grades 2 to 9) across different states in India.

    In this talk, we will go deep into the architecture of an Intelligent Tutoring System. We will start with Domain Model which helps deconstruct the knowledge. We will then move to Student Model which is an overlay on Domain Model used to estimate the students' knowledge states. We will also touch upon the Tutor Model to understand how the student's cognitive and affective states are used to design the student's personalised learning path.

02:45
  • Added to My Schedule
    keyboard_arrow_down
    Ashay Tamhane

    Ashay Tamhane - Modeling Contextual Changes In User Behaviour In Fashion e-commerce

    schedule  02:45 - 03:05 PM place Grand Ball Room 1 people 1 Interested

    Impulse purchases are quite frequent in fashion e-commerce; browse patterns indicate fluid context changes across diverse product types probably due to the lack of a well-defined need at the consumer’s end. Data from fashion e-commerce portal indicate that the final product a person ends-up purchasing is often very different from the initial product he/she started the session with. We refer to this characteristic as a ‘context change’. This feature of fashion e-commerce makes understanding and predicting user behaviour quite challenging. Our work attempts to model this characteristic so as to both detect and preempt context changes. Our approach employs a deep Gated Recurrent Unit (GRU) over clickstream data. We show that this model captures context changes better than other non-sequential baseline models.

  • Added to My Schedule
    keyboard_arrow_down
    Johnu George

    Johnu George / Ramdoot Kumar P - A Scalable Hyperparameter Optimization framework for ML workloads

    schedule  02:45 - 03:05 PM place Grand Ball Room 2

    In machine learning, hyperparameters are parameters that governs the training process itself. For example, learning rate, number of hidden layers, number of nodes per layer are typical hyperparameters for neural networks. Hyperparameter Tuning is the process of searching the best hyper parameters to initialize the learning algorithm, thus improving training performance.

    We present Katib, a scalable and general hyper parameter tuning framework based on Kubernetes which is ML framework agnostic (Tensorflow, Pytorch, MXNet, XGboost etc). You will learn about Katib in Kubeflow, an open source ML toolkit for Kubernetes, as we demonstrate the advantages of hyperparameter optimization by running a sample classification problem. In addition, as we dive into the implementation details, you will learn how to contribute as we expand this platform to include autoML tools.

  • Added to My Schedule
    keyboard_arrow_down

    Experience Report

    schedule  02:45 - 03:05 PM place Jupiter
03:05

    Coffee/Tea Break - 25 mins

03:30
  • Added to My Schedule
    keyboard_arrow_down
    Govind Chada

    Govind Chada - Using 3D Convolutional Neural Networks with Visual Insights for Classification of Lung Nodules and Early Detection of Lung Cancer

    schedule  03:30 - 04:15 PM place Grand Ball Room 1

    Lung cancer is the leading cause of cancer death among both men and women in the U.S., with more than a hundred thousand deaths every year. The five-year survival rate is only 17%; however, early detection of malignant lung nodules significantly improves the chances of survival and prognosis.

    This study aims to show that 3D Convolutional Neural Networks (CNNs) which use the full 3D nature of the input data perform better in classifying lung nodules compared to previously used 2D CNNs. It also demonstrates an approach to develop an optimized 3D CNN that performs with state of art classification accuracies. CNNs, like other deep neural networks, have been black boxes giving users no understanding of why they predict what they predict. This study, for the first time, demonstrates that Gradient-weighted Class Activation Mapping (Grad-CAM) techniques can provide visual explanations for model decisions in lung nodule classification by highlighting discriminative regions. Several CNN architectures using Keras and TensorFlow were implemented as part of this study. The publicly available LUNA16 dataset, comprising 888 CT scans with candidate nodules manually annotated by radiologists, was used to train and test the models. The models were optimized by varying the hyperparameters, to reach accuracies exceeding 90%. Grad-CAM techniques were applied to the optimized 3D CNN to generate images that provide quality visual insights into the model decision making. The results demonstrate the promise of 3D CNNs as highly accurate and trustworthy classifiers for early lung cancer detection, leading to improved chances of survival and prognosis.

  • Added to My Schedule
    keyboard_arrow_down
    Anne Ogborn

    Anne Ogborn - Symbolic AI in a Machine Learning Age

    schedule  03:30 - 04:15 PM place Grand Ball Room 2 people 1 Interested

    Before machine learning took over, AI was done symbolically.

    Symbolic methods still have value, and merging of symbolic and statistical methods is an emerging research area.

    In particular, symbolic methods often have much greater explanatory power. Fusing symbolic methods with ML often creates a more explicable system.

    In this talk we will explore some areas of active work on hybrid applications of symbolic and machine learning.

  • schedule  03:30 - 04:15 PM place Jupiter

    With the big boom in Data Science and Analytics Industry in India, a lot of data scientists are keen on learning a variety of learning algorithms and data manipulation techniques. At the same time, there is this growing interest among data scientists to give back to the society, harness their acquired skills and help fix some of the major burning problems in the nation. But how does one go about finding meaningful datasets connecting to societal problems and plan data-for-good projects? This session will summarize our experience of working in Data-for-Good sector in last 5 years, sharing few interesting datasets and associated use-cases of employing machine learning and artificial intelligence in social sector. Indian social sector is replete with good volume of open data on attributes like annotated images, geospatial information, time-series, Indic languages, Satellite Imagery, etc. We will dive into understanding journey of a Data-for-Good project, getting essential open datasets and understand insights from certain data projects in development sector. Lastly, we will explore how we can work with various communities and scale our algorithmic experiments in meaningful contributions.

04:30
  • Added to My Schedule
    keyboard_arrow_down
    Dr. Vikas Agrawal

    Dr. Vikas Agrawal - Non-Stationary Time Series: Finding Relationships Between Changing Processes for Enterprise Prescriptive Systems

    schedule  04:30 - 05:15 PM place Grand Ball Room 1

    It is too tedious to keep on asking questions, seek explanations or set thresholds for trends or anomalies. Why not find problems before they happen, find explanations for the glitches and suggest shortest paths to fixing them? Businesses are always changing along with their competitive environment and processes. No static model can handle that. Using dynamic models that find time-delayed interactions between multiple time series, we need to make proactive forecasts of anomalous trends of risks and opportunities in operations, sales, revenue and personnel, based on multiple factors influencing each other over time. We need to know how to set what is “normal” and determine when the business processes from six months ago do not apply any more, or only applies to 35% of the cases today, while explaining the causes of risk and sources of opportunity, their relative directions and magnitude, in the context of the decision-making and transactional applications, using state-of-the-art techniques.

    Real world processes and businesses keeps changing, with one moving part changing another over time. Can we capture these changing relationships? Can we use multiple variables to find risks on key interesting ones? We will take a fun journey culminating in the most recent developments in the field. What methods work well and which break? What can we use in practice?

    For instance, we can show a CEO that they would miss their revenue target by over 6% for the quarter, and tell us why i.e. in what ways has their business changed over the last year. Then we provide the prioritized ordered lists of quickest, cheapest and least risky paths to help turn them over the tide, with estimates of relative costs and expected probability of success.

  • Added to My Schedule
    keyboard_arrow_down
    Anuj Gupta

    Anuj Gupta - Continuous Learning Systems: Building ML systems that keep learning from their mistakes

    schedule  04:30 - 05:15 PM place Grand Ball Room 2

    Won't it be great to have ML models that can update their “learning” as and when they make mistake and correction is provided in real time? In this talk we look at a concrete business use case which warrants such a system. We will take a deep dive to understand the use case and how we went about building a continuously learning system for text classification. The approaches we took, the results we got.

    For most machine learning systems, “train once, just predict thereafter” paradigm works well. However, there are scenarios when this paradigm does not suffice. The model needs to be updated often enough. Two of the most common cases are:

    1. When the distribution is non-stationary i.e. the distribution of the data changes. This implies that with time the test data will have very different distribution from the training data.
    2. The model needs to learn from its mistakes.

    While (1) is often addressed by retraining the model, (2) is often addressed using batch update. Batch updation requires collecting a sizeable number of feedback points. What if you have much fewer feedback points? You need model that can learn continuously - as and when model makes a mistake and feedback is provided. To best of our knowledge there is a very limited literature on this.

  • Added to My Schedule
    keyboard_arrow_down
    Yash Deo

    Yash Deo - Big Data to Big Intelligence - Generation of Actionable Insights from Open Source Data

    schedule  04:30 - 05:15 PM place Jupiter people 1 Interested

    As a data scientist i have been lucky enough to be a part of highly critical and cutting edge solutions for pristine organizations like Intel , Indian Army etc. While each of them was an amazing experience in its own right , the challenges i faced and the knowledge i gained from making an Open Source Intelligence gathering and Analytics/Prediction tool for the Indian Army are unmatched and i will be sharing some of the main challenges that i think will be encountered in .

    An OSINT tool can have some powerful capabalities like :-

    • Predict and estimate the location of an Twitter/Facebook user (who has disabled his location obviously!) through various metrics.
    • Predict occurrence of certain events (eg. Riot's) based of information gathered from various Open Sources.
    • Identify and Predict accounts of people who may be potential suspects (Security use Case) or potential Influencers (Commercial use case).
    • Contextual analysis of words to derive relevant insights.

    I worked on this project for over an year , and since then have been using my experience of OSINT in other sectors like Healthcare/Pharma . While working for other sectors i was appalled by the way OSINT information was being under-utilized .

    I would like to share my experience of working with some great mean on a very critical project for out military by discussing the problems i faced and how they can be overcome , at the same time i hope to give you a guideline on how you can efficiently utilize the power of OSINT information in your respective field be in consumer goods , healthcare or energy. I hope that the attendees will pick up some valuable insights from my experience which will help them in projects ranging from NLP to Time-Series analytics.

05:30

    Day 1 Evening Keynote - 45 mins

06:30

    Panel - 45 mins

07:30

    Reception Dinner & Networking - 150 mins

ODSC India Day 2

Fri, Aug 9
09:00
  • Added to My Schedule
    keyboard_arrow_down
    Grant Sanderson

    Grant Sanderson - Concrete before Abstract

    schedule  09:00 - 09:45 AM place Grand Ball Room

    This talk outlines a principle of technical communication which seems simple at first but is devilishly difficult to abide by. It's a principle I try to keep in mind when creating videos aimed at making math and related fields more accessible, and it stands to benefit anyone who regularly needs to describe mathematical ideas in their work. Put simply, it's to resist the temptation to open a topic by describing a general result or definition, and instead let examples precede generality. More than that, it's about finding the type of example which guides the audience to rediscover the general results for themselves. We'll look, aptly enough, at examples of what I mean by this, why it's deceptively difficult to follow, and why this ordering matters.

10:00

    Important Announcements - 30 mins

10:30

    Coffee/Tea Break - 30 mins

11:00
  • Added to My Schedule
    keyboard_arrow_down
    Juan Manuel Contreras

    Juan Manuel Contreras - Beyond Individual Contribution: How to Lead Data Science Teams

    schedule  11:00 - 11:45 AM place Grand Ball Room 1 people 1 Interested

    Despite the increasing number of data scientists who are being asked to take on managerial and leadership roles as they grow in their careers, there are still few resources on how to manage data scientists and lead data science teams. There is also scant practical advice on how to serve as head of a data science practice: how to set a vision and craft a strategy for an organization to use data science.

    In this talk, I will describe my experience as a data science leader both at a political party (the Democratic Party of the United States of America) and at a fintech startup (Even.com), share lessons learned from these experiences and conversations with other data science leaders, and offer a framework for how new data science leaders can better transition to both managing data scientists and heading a data science practice.

  • Added to My Schedule
    keyboard_arrow_down
    Vijay Gabale

    Vijay Gabale - Data Distribution Search: Deep Reinforcement Learning To Improvise Input Datasets

    schedule  11:00 - 11:45 AM place Grand Ball Room 2

    Beyond computer games and neural architecture search; practical applications of Deep Reinforcement Learning to improve classical classification or detection tasks are few and far between. In this talk, I will share a technique and our experiences of applying D-RL on improving the distribution input datasets to achieve state of the art performance, specifically on object detection tasks.

    Beyond open source datasets, when it comes to building neural networks for real-world problems, dataset matters, which is often small and skewed.The talk presents a few fresh perspectives on how to artificially increase the size of datasets while balancing the data distribution. We show that these ideas result in 2% to 3% increase in accuracy on popular object detection tasks, whereas small and skewed datasets yield up to 22% increase in model accuracies.

  • Added to My Schedule
    keyboard_arrow_down
    Jeetendra Kumar Sharma

    Jeetendra Kumar Sharma / Vikas Grover - Leveraging Video Analytics at United Airlines: Calculating Deplaning Times Using Deep Learning

    schedule  11:00 - 11:45 AM place Jupiter

    For United Airlines, running a Safe and Efficient airline is core to our business. And with such a complex operation, we need to constantly track key events that keep the airline running smoothly. While tracking these events can be time-intensive and laborious, we believe developments in deep learning and edge computing are going to help us simplify that process. Over the past few months, United’s Data Science team has been exploring how to leverage advances in computer vision to solve some of these problems. Our presentation will focus on solving one of these tasks: timing how long it takes for passengers to exit an aircraft. We’ll provide an overview of key concepts of video analytics, share how we leveraged open source technology to build a solution and provide a demonstration of our work.

12:00
  • schedule  12:00 - 12:45 PM place Grand Ball Room 1 people 1 Interested

    In todays world majority of information is generated by self sustaining systems like various kinds of bots, crawlers, servers, various online services, etc. This information is flowing on the axis of time and is generated by these actors under some complex logic. For example, a stream of buy/sell order requests by an Order Gateway in financial world, or a stream of web requests by a monitoring / crawling service in the web world, or may be a hacker's bot sitting on internet and attacking various computers. Although we may not be able to know the motive or intention behind these data sources. But via some unsupervised techniques we can try to infer the pattern or correlate the events based on their multiple occurrences on the axis of time. Thus we could automatically identify signatures of various actors and take appropriate actions.

    Sessionisation is one such unsupervised technique that tries to find the signal in a stream of events associated with a timestamp. In the ideal world it would resolve to finding periods with a mixture of sinusoidal waves. But for the real world this is a much complex activity, as even the systematic events generated by machines over the internet behave in a much erratic manner. So the notion of a period for a signal also changes in the real world. We can no longer associate it with a number, it has to be treated as a random variable, with expected values and associated variance. Hence we need to model "Stochastic periods" and learn their probability distributions in an unsupervised manner. This would be done via non-parametric Bayesian techniques with Gaussian prior.

    In this talk we will do a walk through of a real security use cases solved via Sessionisation for the SOC (Security Operations Centre) centre of an international firm with offices in 56 countries being monitored via a central SOC team.

    In this talk we will go through a Sessionisation technique based on stochastic periods. The journey would begin by extracting relevant data from a sequence of timestamped events. Then we would apply various techniques like FFT (Fast Fourier Transform), kernel density estimation, optimal signal selection, Gaussian Mixture Models, etc. and eventually discover patterns in time stamped events.

    Key concepts explained in talk: Sessionisation, Bayesian techniques of Machine Learning, Gaussian Mixture Models, Kernel density estimation, FFT, stochastic periods, probabilistic modelling

  • Added to My Schedule
    keyboard_arrow_down
    Venkata Pingali

    Venkata Pingali - Accelerating ML using Production Feature Engineering Platform

    schedule  12:00 - 12:45 PM place Grand Ball Room 2

    Anecdotally only 2% of the models developed are productionized, i.e., used day to day to improve business outcomes. Part of the reason is the high cost and complexity of productionization of models. It is estimated to be anywhere from 40 to 80% of the overall work.

    In this talk, we will share Scribble Data’s insights into productionization of ML, and how to reduce the cost and complexity in organizations. It is based on the last two years of work at Scribble developing and deploying production ML Feature Engineering Platform, and study of platforms from major organizations such as Uber. This talk expands on a previous talk given in January.

    First, we discuss the complexity of production ML systems, and where time and effort goes. Second, we give an overview of feature engineering, which is an expensive ML task, and the associated challenges Third, we suggest an architecture for Production Feature Engineering platform. Last, we discuss how one could go about building one for your organization

  • Added to My Schedule
    keyboard_arrow_down
    Anant Jain

    Anant Jain - Adversarial Attacks on Neural Networks

    schedule  12:00 - 12:45 PM place Jupiter

    Since 2014, adversarial examples in Deep Neural Networks have come a long way. This talk aims to be a comprehensive introduction to adversarial attacks including various threat models (black box/white box), approaches to create adversarial examples and will include demos. The talk will dive deep into the intuition behind why adversarial examples exhibit the properties they do — in particular, transferability across models and training data, as well as high confidence of incorrect labels. Finally, we will go over various approaches to mitigate these attacks (Adversarial Training, Defensive Distillation, Gradient Masking, etc.) and discuss what seems to have worked best over the past year.

12:45

    Lunch - 60 mins

01:45
  • Added to My Schedule
    keyboard_arrow_down
    Kabir Rustogi

    Kabir Rustogi - Generation of Locality Polygons using Open Source Road Network Data and Non-Linear Multi-classification Techniques

    schedule  01:45 - 02:30 PM place Grand Ball Room 1 people 1 Interested

    One of the principal problems in the developing world is the poor localization of its addresses. This inhibits discoverability of local trade, reduces availability of amenities such as creation of bank accounts and delivery of goods and services (e.g., e-commerce) and delays emergency services such as fire brigades and ambulances. In general, people in the developing World identify an address based on neighbourhood/locality names and points of interest (POIs), which are neither standardized nor any official records exist that can help in locating them systematically. In this paper, we describe an approach to build accurate geographical boundaries (polygons) for such localities.

    As training data, we are provided with two pieces of information for millions of address records: (i) a geocode, which is captured by a human for the given address, (ii) set of localities present in that address. The latter is determined either by manual tagging or by using an algorithm which is able to take a raw address string as input and output meaningful locality information present in that address. For example, for the address, “A-161 Raheja Atlantis Sector 31 Gurgaon 122002”, its geocode is given as (28.452800, 77.045903), and the set of localities present in that address is given as (Raheja Atlantis, Sector 31, Gurgaon, Pin-code 122002). Development of this algorithm are part of any other project we are working on; details about the same can be found here.

    Many industries, such as the food-delivery industry, courier-delivery industry, KYC (know-your-customer) data-collection industry, are likely to have huge amounts of such data. Such crowdsourced data usually contain large a amount of noise, acquired either due to machine/human error in capturing the geocode, or due to error in identifying the correct set of localities from a poorly written address. For example, for the address, “Plot 1000, Sector 31 opposite Sector 40 road, Gurgaon 122002”, a machine may output the set of localities present in this address as (Sector 31, Sector 40, Gurgaon, Pin-code 122002), even though it is clear that the address does not lie in Sector 40.

    The solution described in this paper is expected to consume the provided data and output polygons for each of the localities identified in the address data. We assume that the localities for which we must build polygons are non-overlapping, e.g., this assumption is true for pin-codes. The problem is solved in two phases.

    In the first phase, we separate the noisy points from the points that lie within a locality. This is done by formulating the problem as a non-linear multi-classification problem. The latitudes and longitudes of all non-overlapping localities act as features, and their corresponding locality name acts as a label, in the training data. The classifier is expected to partition the 2D space containing the latitudes and longitudes of the union of all non-overlapping localities into disjoint regions corresponding to each locality. These partitions are defined as non-linear boundaries, which are obtained by optimizing for two objectives: (i) the area enclosed by the boundaries should maximize the number of points of the corresponding locality and minimize the number of points of other localities, (ii) the separation boundary should be smooth. We compare two algorithms, decision trees and neural networks for creating such partitions.

    In the second phase, we extract all the points that satisfy the partition constraints, i.e., lie within the boundary of a locality L, as candidate points, for generating the polygon for locality L. The resulting polygon must contain all candidate points and should have the minimum possible area while maintaining the smoothness of the polygon boundary. This objective can be achieved by algorithms such as concave hull. However, since localities are always bounded by roads, we have further enhanced our locality polygons by leveraging open source data of road networks. To achieve this, we solve a non-linear optimisation problem which decides the set of roads to be selected, so that the enclosed area is minimized, while ensuring that all the candidate points lie within the enclosed area. The output of this optimisation problem is a set of roads, which represents the boundary of a locality L.

  • Added to My Schedule
    keyboard_arrow_down
    Ishita Mathur

    Ishita Mathur - How GO-FOOD built a Query Semantics Engine to help you find the food you want to order

    schedule  01:45 - 02:30 PM place Grand Ball Room 2 people 1 Interested

    Context: The Search problem

    GOJEK is a SuperApp: 19+ apps within an umbrella app. One of these is GO-FOOD, the first food delivery service in Indonesia and the largest food delivery service in Southeast Asia. There are over 300 thousand restaurants on the platform with a total of over 16 million dishes between them.

    Over two-thirds of those who order food online using GO-FOOD do so by utilising text search. Search engines are so essential to our everyday digital experience that we don’t think twice when using them anymore. Search engines involve two primary tasks: retrieval of documents and ranking them in order of relevance. While improving that ranking is an extremely important part of improving the search experience, actually understanding that query helps give the searcher exactly what they’re looking for. This talk will show you what we are doing to make it easy for users to find what they want.

    GO-FOOD uses the ElasticSearch stack with restaurant and dish indexes to search for what the user types. However, this results in only exact text matches and at most, fuzzy matches. We wanted to create a holistic search experience that not only personalised search results, but also retrieved restaurants and dishes that were more relevant to what the user was looking for. This is being done by not only taking advantage of ElasticSearch features, but also developing a Query semantics engine.

    Query Understanding: What & Why

    This is where Query Understanding comes into the picture: it’s about using NLP to correctly identify the search intent behind the query and return more relevant search results, it’s about the interpretation process even before the results are even retrieved and ranked. The semantic neighbours of the query itself become the focus of the search process: after all, if I don’t understand what you’re trying to ask for, how will I give you what you want?

    In the duration of this talk, you will learn about how we are taking advantage of word embeddings to build a Query Understanding Engine that is holistically designed to make the customer’s experience as smooth as possible. I will go over the techniques we used to build each component of the engine, the data and algorithmic challenges we faced and how we solved each problem we came across.

  • Added to My Schedule
    keyboard_arrow_down
    Subhasish Misra

    Subhasish Misra - Causal data science: Answering the crucial ‘why’ in your analysis.

    schedule  01:45 - 02:30 PM place Jupiter people 1 Interested

    Causal questions are ubiquitous in data science. For e.g. questions such as, did changing a feature in a website lead to more traffic or if digital ad exposure led to incremental purchase are deeply rooted in causality.

    Randomized tests are considered to be the gold standard when it comes to getting to causal effects. However, experiments in many cases are unfeasible or unethical. In such cases one has to rely on observational (non-experimental) data to derive causal insights. The crucial difference between randomized experiments and observational data is that in the former, test subjects (e.g. customers) are randomly assigned a treatment (e.g. digital advertisement exposure). This helps curb the possibility that user response (e.g. clicking on a link in the ad and purchasing the product) across the two groups of treated and non-treated subjects is different owing to pre-existing differences in user characteristic (e.g. demographics, geo-location etc.). In essence, we can then attribute divergences observed post-treatment in key outcomes (e.g. purchase rate), as the causal impact of the treatment.

    This treatment assignment mechanism that makes causal attribution possible via randomization is absent though when using observational data. Thankfully, there are scientific (statistical and beyond) techniques available to ensure that we are able to circumvent this shortcoming and get to causal reads.

    The aim of this talk, will be to offer a practical overview of the above aspects of causal inference -which in turn as a discipline lies at the fascinating confluence of statistics, philosophy, computer science, psychology, economics, and medicine, among others. Topics include:

    • The fundamental tenets of causality and measuring causal effects.
    • Challenges involved in measuring causal effects in real world situations.
    • Distinguishing between randomized and observational approaches to measuring the same.
    • Provide an introduction to measuring causal effects using observational data using matching and its extension of propensity score based matching with a focus on the a) the intuition and statistics behind it b) Tips from the trenches, basis the speakers experience in these techniques and c) Practical limitations of such approaches
    • Walk through an example of how matching was applied to get to causal insights regarding effectiveness of a digital product for a major retailer.
    • Finally conclude with why understanding having a nuanced understanding of causality is all the more important in the big data era we are into.
02:45
  • Added to My Schedule
    keyboard_arrow_down
    Dr. Vijay Srinivas Agneeswaran

    Dr. Vijay Srinivas Agneeswaran / Abhishek Kumar - Industrialized Capsule Networks for Text Analytics

    schedule  02:45 - 03:30 PM place Grand Ball Room 1 people 1 Interested

    Multi-label text classification is an interesting problem where multiple tags or categories may have to be associated with the given text/documents. Multi-label text classification occurs in numerous real-world scenarios, for instance, in news categorization and in bioinformatics (gene classification problem, see [Zafer Barutcuoglu et. al 2006]). Kaggle data set is representative of the problem: https://www.kaggle.com/jhoward/nb-svm-strong-linear-baseline/data.

    Several other interesting problem in text analytics exist, such as abstractive summarization [Chen, Yen-Chun 2018], sentiment analysis, search and information retrieval, entity resolution, document categorization, document clustering, machine translation etc. Deep learning has been applied to solve many of the above problems – for instance, the paper [Rie Johnson et. al 2015] gives an early approach to applying a convolutional network to make effective use of word order in text categorization. Recurrent Neural Networks (RNNs) have been effective in various tasks in text analytics, as explained here. Significant progress has been achieved in language translation by modelling machine translation using an encoder-decoder approach with the encoder formed by a neural network [Dzmitry Bahdanau et. al 2014].

    However, as shown in [Dan Rosa de Jesus et. al 2018] , certain cases require modelling the hierarchical relationship in text data and is difficult to achieve with traditional deep learning networks because linguistic knowledge may have to be incorporated in these networks to achieve high accuracy. Moreover, deep learning networks do not consider hierarchical relationships between local features as pooling operation of CNNs lose information about the hierarchical relationships.

    We show one industrial scale use case of capsule networks which we have implemented for our client in the realm of text analytics – news categorization. We explain how traditional deep learning methods may not be useful in the case when single-label data is only available for training (as in many real-life cases), while the test data set is multi-labelled – this is the sweet spot for capsule networks. We also discuss the key challenges faced industrialization of capsule networks – starting from providing a scalable implementation of capsule networks in TensorFlow, we show how capsule networks can be industrialized by providing an implementation on top of KubeFlow, which helps in productionization.

    1. History of impact of machine learning and deep learning on NLP.

    2. Motivation for capsule networks and how they can be used in text analytics.

    3. Implementation of capsule networks in TensorFlow.

    4. Industrialization of capsule nets with KubeFlow.

    References:

    [Zafer Barutcuoglu et. al 2006] Zafer Barutcuoglu, Robert E. Schapire, and Olga G. Troyanskaya. 2006. Hierarchical multi-label prediction of gene function. Bioinformatics 22, 7 (April 2006), 830-836. DOI=http://dx.doi.org/10.1093/bioinformatics/btk048

    [Rie Johnson et. al 2015] Rie Johnson, Tong Zhang: Effective Use of Word Order for Text Categorization with Convolutional Neural Networks. HLT-NAACL 2015: 103-112.

    [Dzmitry Bahdanau et. al 2014] Bahdanau, Dzmitry et al. “Neural Machine Translation by Jointly Learning to Align and Translate.” CoRR abs/1409.0473 (2014).

    [Dan Rosa de Jesus et. al 2018] Dan Rosa de Jesus, Julian Cuevas, Wilson Rivera, Silvia Crivelli (2018). “Capsule Networks for Protein Structure Classification and Prediction”,

    available at https://arxiv.org/abs/1808.07475.

    [Yequan Wang et. al 2018] Yequan Wang, Aixin Sun, Jialong Han, Ying Liu, and Xiaoyan Zhu. 2018. Sentiment Analysis by Capsules. In Proceedings of the 2018 World Wide Web Conference (WWW '18). International World Wide Web Conferences Steering Committee, Republic and Canton of Geneva, Switzerland, 1165-1174. DOI: https://doi.org/10.1145/3178876.3186015

    Chen, Yen-Chun and Bansal, Mohit (2018), “Fast Abstractive Summarization with Reinforce-Selected Sentence Rewriting”, eprint arXiv:1805.11080.

  • Added to My Schedule
    keyboard_arrow_down
    Akshay Bahadur

    Akshay Bahadur - Minimizing CPU utilization for deep networks

    schedule  02:45 - 03:30 PM place Grand Ball Room 2

    The advent of machine learning along with its integration with computer vision has enabled users to efficiently to develop image-based solutions for innumerable use cases. A machine learning model consists of an algorithm which draws some meaningful correlation between the data without being tightly coupled to a specific set of rules. It's crucial to explain the subtle nuances of the network along with the use-case we are trying to solve. With the advent of technology, the quality of the images has increased which in turn has increased the need for resources to process the images for building a model. The main question, however, is to discuss the need to develop lightweight models keeping the performance of the system intact.
    To connect the dots, we will talk about the development of these applications specifically aimed to provide equally accurate results without using much of the resources. This is achieved by using image processing techniques along with optimizing the network architecture.
    These applications will range from recognizing digits, alphabets which the user can 'draw' at runtime; developing state of the art facial recognition system; predicting hand emojis, developing a self-driving system, detecting Malaria and brain tumor, along with Google's project of 'Quick, Draw' of hand doodles.
    In this presentation, we will discuss the development of such applications with minimization of CPU usage.

  • Added to My Schedule
    keyboard_arrow_down
    Samiran Roy

    Samiran Roy / Shibsankar Das - Algorithms that learn to solve tasks by watching (one) Youtube video

    schedule  02:45 - 03:30 PM place Jupiter

    Two branches of AI - Deep Learning, and Reinforcement Learning are now responsible for many real-world applications. Machine Translation, Speech Recognition, Object Detection, Robot Control, and Drug Discovery - are some of the numerous examples.

    Both approaches are data hungry - DL requires many examples of each class, and RL needs to play through many episodes to learn a policy. Contrast this to human intelligence. A small child can typically see an image just once, and instantly recognize it in other contexts and environments. We seem to possess an innate model/representation of how the world works, which helps us grasp new concepts and adapt to new situations fast. Humans are excellent one/few shot learners. We are able to learn complex tasks by observing and imitating other humans (eg: cooking, dancing or playing soccer) - despite having a different point of view, sense modalities, body structure, mental facility.

    Humans may be very good at picking up novel tasks, but Deep RL agents surpass us in performance. Once a Deep RL has learned a good representation [1], it is easy to surpass human performance in complex tasks like Go[2], Dota 2[3], and Starcraft[4]. We are biologically limited by time, memory and computation (A computer can be made to simulate thousands of plays in a minute).

    RL struggles with tasks that have sparse rewards. Take an example of a soccer playing robot - controlled by applying a torque to each one of its joints. The environment rewards you when it scores a goal. If the policy is initialized randomly (we apply a random torque to each joint, every few milliseconds) the probability of the robot scoring a goal is negligible - it won't even be able to learn how to stand up. In tasks requiring long term planning or low-level skills, getting to that initial reward can prove impossible. These situations have the potential to greatly benefit from a demonstration - in this case showing the robot how to walk and kick - and then letting it figure out how to score a goal.

    We have an abundance of visual data on humans performing various tasks, in the public domain, in the form of videos from sources like YouTube. In Youtube alone, 400 hours of videos are uploaded every minute, and it is easy to find demonstration videos for any skill imaginable. What if we could harness this by designing agents that could learn how to perform tasks - just by watching a video clip?

    Imitation Learning, also known as apprenticeship learning, teaches an agent a sequence of decisions through demonstration, often by a human expert. It has been used in many applications such as teaching drones how to fly[5] and autonomous cars how to drive[6] - It relies on domain engineered features - or extremely precise representations such as mocap [7]. Directly applying imitation learning to learn from videos proves challenging, there is a misalignment of representation between the demonstrations and the agent’s environment. For example: How can a robot sensing its world through a 3d point cloud - learn from a noisy 2d video clip of a soccer player dribbling?

    Leveraging recent advances in Reinforcement Learning, Self Supervised Learning and Imitation Learning [8] [9] [10], We present a technical deep dive into an end to end framework which:

    1) Has prior knowledge about the world intelligence through Self-Supervised Learning - A relatively new area which seeks to build efficient deep learning representations from unlabelled data but training on a surrogate task. The surrogate task can be rotating an image and predicting the rotation angle or cropping two patches of the image, and predicting their relative tasks - or a combination of several such objectives.

    2) Has the ability to align the representation of how it senses the world, with that of the video - allowing it to learn diverse tasks from video clips.

    3) Has the ability to reproduce a skill, from only a single demonstration - using applied techniques from imitation learning

    [1] https://www.cse.iitb.ac.in/~shivaram/papers/ks_adprl_2011.pdf

    [2] https://ai.google/research/pubs/pub44806

    [3] https://openai.com/five/

    [4] https://deepmind.com/blog/alphastar-mastering-real-time-strategy-game-starcraft-ii/

    [5] http://cs231n.stanford.edu/reports/2017/pdfs/614.pdf

    [6] https://arxiv.org/pdf/1709.07174.pdf

    [7] https://en.wikipedia.org/wiki/Motion_capture

    [8] https://arxiv.org/pdf/1704.06888v3.pdf

    [9] https://bair.berkeley.edu/blog/2018/06/28/daml/

    [10] https://arxiv.org/pdf/1805.11592v2.pdf

03:30

    Coffee/Tea Break - 30 mins

04:00
  • Added to My Schedule
    keyboard_arrow_down
    Yogesh H. Kulkarni

    Yogesh H. Kulkarni - MidcurveNN: Encoder-Decoder Neural Network for Computing Midcurve of a Thin Polygon

    schedule  04:00 - 04:45 PM place Grand Ball Room 1

    Various applications need lower dimensional representation of shapes. Midcurve is one- dimensional(1D) representation of a two-dimensional (2D) planar shape. It is used in applications such as animation, shape matching, retrieval, finite element analysis, etc. Methods available to compute midcurves vary based on the type of the input shape (images, sketches, etc.) and processing approaches such as Thinning, Medial Axis Transform (MAT), Chordal Axis Transform (CAT), Straight Skeletons, etc., all of which are rule-based.

    This presentation talks about a novel method called MidcurveNN which uses Encoder-Decoder neural network for computing midcurve from images of 2D thin polygons in supervised learning manner. This dimension reduction transformation from input 2D thin polygon image to output 1D midcurve image is learnt by the neural network, which can then be used to compute midcurve of an unseen 2D thin polygonal shape.

  • Added to My Schedule
    keyboard_arrow_down

    Case Study 2

    schedule  04:00 - 04:45 PM place Grand Ball Room 2
  • Added to My Schedule
    keyboard_arrow_down

    Case Study 3

    schedule  04:00 - 04:45 PM place Jupiter
05:00

    Closing Keynote - 45 mins

05:45

    Closing Talk - 15 mins

Post-Conf Workshop

Sat, Aug 10
09:30

    Registration - 30 mins

10:00
  • Added to My Schedule
    keyboard_arrow_down
    Rahee Walambe

    Rahee Walambe / Vishal Gokhale - Processing Sequential Data using RNNs

    schedule  10:00 AM - 06:00 PM place Jupiter 1

    Data that forms the basis of many of our daily activities like speech, text, videos has sequential/temporal dependencies. Traditional deep learning models, being inadequate to model this connectivity needed to be made recurrent before they brought technologies such as voice assistants (Alexa, Siri) or video based speech translation (Google Translate) to a practically usable form by reducing the Word Error Rate (WER) significantly. RNNs solve this problem by adding internal memory. The capacities of traditional neural networks are bolstered with this addition and the results outperform the conventional ML techniques wherever the temporal dynamics are more important.
    In this full-day immersive workshop, participants will develop an intuition for sequence models through hands-on learning along with the mathematical premise of RNNs.

  • Added to My Schedule
    keyboard_arrow_down

    Full-Day Workshop

    schedule  10:00 AM - 06:00 PM place Jupiter 2