DOWNLOAD TENTATIVE SCHEDULE
Over the last two years, the Natural Language Processing (NLP) community has witnessed an acceleration in progress on a wide range of different tasks and applications. However, it still has its limitations such as Lexical ambiguity, referential ambiguity, slangs and so on. Mathangi from her vast experience will shed light on the current state as well as limitations of NLP
Power Talk: Why is Enterprise Problem-Solving Not Like Chess or Go or ImageNet or Kaggle challenges?
Vikas Agarwal, who comes with immense experience with Oracle talks about upcoming best practices when it comes to AI in the enterprise. How to detect that our chosen tools or methods are misbehaving or are not modelling the process? How to ensure that the key sources of variation are included in the data used for modelling? will be some of the questions that would be tackled head-on.
Jayatu and Abhishek will introduce American Express as a global organization, and provide insight on its journey to become an analytics-focused company. They will walk through a set of use cases showcasing how billions of data inputs, combined with the right AI/ML techniques and business insights, deliver a significant competitive edge for success.
Machine Learning Models are always used as means to an end which in other words is also called prescriptive analytics. Eric, in this talk would focus on this particular aspect wherein he will lay foundations on the impact aspect of Machine Learning and why model building is just a small part of the entire Data Science Lifecycle.
DJ Sarkar from his in-depth understanding of NLP will demonstrate an industry use case of applying transfer models. Learn how to leverage models such as BERT etc. to build high-performance NLP Models in identifying how to flag software vulnerabilities before it becomes a serious issue affecting all downstream applications using it.
Being able to learn a robust representation of a video is not an easy task as it is a complex mix of sequential data (like time series) and images (RGB tensor). Axel De Romblay who is the creator of automatic machine learning library MLBox will cover an overview of building modelling pipelines for video classification and video encoding with State of the art computer vision algorithms and transfer learning.
Apple’s Swift is creating ripples in the data science world and we data scientists don’t miss out on things that have the potential to change the way we work. This session will help you understand why Swift ecosystem is increasingly being talked about with developments such as Swift in Tensorflow and why are leading researchers such as Jeremy Howard backing it.
Data that typically collected are the symptoms of real-world processes of interests and not really the underlying valuable hidden features. Listen to this power talk to help you understand the working of the not so important variables and hidden variables. Dr Sarabjot will help you understand the role of latent features in recommender systems, NLP and more.
Neural Networks are attempting to transform human society through machine-based representations that mimic patterns of biological neural activity. This power talk will help you understand more about biologically-realistic compartmental structure in the design of deep learning algorithms.
Standard machine learning approaches require centralizing the training data on one machine or in a datacenter. Federated Learning enables mobile phones to collaboratively learn a shared prediction model while keeping all the training data on device, decoupling the ability to do machine learning from the need to store the data in the cloud. Bargava & Tuhin will cover how to build Deep Learning models using federated learning in this exciting session.
Do you know that more than a million customer support requests in local languages are raised from the Jio network every month? Accurately detecting the intent of the user from natural language utterance is one of the fundamental problems to be solved in order to truly move from clicks to conversations. This session will help you understand multi-label text classification and ranking based on cosine similarity to build an intent identification system specifically for Indic languages.
Learn how to build high-speed inference translation models using quantized deep learning models along with its opportunities and challenges. Nishant from Intel in this exciting session will also talk about the best practices to run an inference with high efficiency on Intel CPUs.
Can we identify whether an employee is unhappy from his LinkedIn activity? HR Analytics poses interesting uses cases of employing ML & Data Science to reduce attrition rate, identify the next leaders or managers and more. Check out Kavita's talk, where she will cover some of these use cases using social media metrics to solve problems in HR Analytics.
Machine learning for sensors and signal data is becoming easier than ever: hardware is becoming smaller and sensors are getting cheaper, making IoT devices widely available for a variety of applications ranging from predictive maintenance to user behaviour monitoring. Anurag will cover predictive maintenance and also how autoencoders are used to get the state of sensors.
Out of 20 artificial intelligence projects, only three will succeed – 17 of them will fall short and not even go in production. With all the buzz and talk about AI technologies, that’s an interesting and surprising fact. This panel discussion involving top industry leaders in AI will touch upon the reasons for the failure of AI projects and way forward.
Envestnet | Yodlee
Supervised data is expensive/time-consuming to obtain – Semi Supervised Learning (SSL) Algorithms improve sample efficiency by leveraging a large amount of unlabelled data in conjunction with labelled data. Samiran in this hack session will do deep dive code walkthrough of Graph Convolutional Neural Networks (GCNs) which have showed special promise for Semi-Supervised Learning.
Most problems faced in industry is in some way an optimization problem which is nothing but finding the best solution from all feasible solutions. Be it optimizing profits, time, manpower and so on. This talk is all about solving such problems with approaches such as linear programming, genetic algorithm, simulated annealing, and reinforcement learning.
NLP's latest pre-trained language models like BERT, GPT2, TransformerXL, XLM, etc. are achieving state of the art results in a wide range of NLP tasks. In this hack session, community favourite and Kaggle Grandmaster SRK will compare the performance of these different pre-trained models along with pre-trained word vector models on classification tasks.
Deploying a model poses several challenges, such as model versioning, containerization of the model, etc. Web frameworks like Flask and Django can be used to wrap the model into a REST API and expose the API. But this solution requires developers to write and maintain code to handle requests to the model and support other deployment-related features as well. Tata in this hack session will do a deep dive into Tensorflow Serving which is a flexible, high-performance serving system for machine learning models, designed for production environments with a use case.
This power talk will be a detailed introduction to Quantum Computing, NISQ era,Quantum computation models, Quantum Machine Learning (QML) Algorithms and advantages of QML algorithms over their classical counterparts.
Symbiosis Centre for IT
For providing haptic feedback, users have been dependent on external devices including buttons, dials, stylus or even touch screens. The advent of machine learning along with its integration with computer vision has enabled users to efficiently provide inputs and feedback to the system. 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 along with Google’s project of ‘Quick, Draw’ of hand doodles. By the end of the session, the audience will have a clearer understanding of building vision-based optimized models which can take feedback from anatomical features.
Hyperparameter tuning is an essential part of any machine learning pipeline. With better compute we now have the power to explore more range of hyperparameters quickly but especially for more complex algorithms, the space for hyperparameters remain vast and techniques such as Bayesian Optimization might help in making the tuning process faster. Witness how you can use HyperOpt library uses Bayesian techniques to find the best hyperparameters with a real dataset.
Take a deep dive into data and formulations in traditional (e.g., retail) and non-traditional fields (e.g., genomics) to draw similarities and differences between different Deep Learning approaches with Vijay Gabale in this exciting power talk.
Artificial Intelligence based solutions are enabling data-based decision making in many domains. With growing confidence in AI algorithms, recent years have seen applications of AI in healthcare to improve healthcare operations, delivery and care. In this talk, Dr. Geetha will share her experience developing a novel AI-based solution for detecting breast cancer, and the story of the most awarded startup NIRAMAI, which is taking the tech solution to market.
Learn how to manage machine learning projects from the concept stage to the final completion stage. Various aspects of a typical ML project, including how to successfully manage one from scratch will be discussed along with tips on model building and last mile optimization.
Hack Session: Evaluating ML Models for Bias – Build an Interpretable Model using a Financial Dataset
Although machine learning, by its very nature, is always a form of statistical discrimination, the discrimination becomes objectionable when it places certain privileged groups at a systematic advantage and certain unprivileged groups at a systematic disadvantage. In this hack session, Rajesh and Prateek from IBM will discuss the concepts and capabilities of a model to test for biases and explanations. along with the wider spectrum of explainability methods, notably data explanations, metrics and persona specific explanations.
IBM Watson OpenScale
As a data engineer, Spark & Kafka are essential tools to work with streaming data and build robust pipelines for the same. Durgaraju in this hack session will give an overview of streaming analytics and then demonstrate the integration of Kafka and Spark Structured Streaming
Today’s communications service providers (CSPs) face increasing demands for higher quality services and better customer experience. Additionally, the advent of 5G and IoT is introducing new challenges for mobile CSPs, and integrating AI techniques into networks is one way the industry is addressing these complexities. Amit from leadership team of Ericsson in this talk will uncover how AI/ML is helping automate, evolve and transform the Telco world.
Can Computer Vision help build the next gen Yoga Instructor? Mohsin and Apurva will do a live demo of an in browser yoga trainer which uses deep learning models to score correctness of yoga poses. Then they will dive deeper into what goes behind creating such a system in-browser.
Can reinforcement learning be used to find the best route for cab aggregators for each cab? Join this session to get answers from Sayan Ranu who is a professor of Computer Science and Engineering at IIT Delhi and has numerous publications for transportation.
CSE, IIT Delhi
This talk will deep dive into the finance world and help you understand Automated Portfolio Management using Reinforcement learning.
This talk will be a guide on how to use Machine Learning and Deep Learning inside your healthcare projects with real life stories in ophthalmology, cytology and dental areas. Tarry who is the founder of DeepKapha.ai will also discuss the challenges and opportunities in this domain.
The financial service industry has been revolutionized by technological advancements. The way in which banks and fintech enterprises operate today has been fundamentally changed by technological advancements. Numerous advantages can be identified when discussing fintech big data. Ratnakar in this talk will cover the encapsulation of borrower’s behavior, computing technologies, statistical models and analytics in rich data sources for FinTech.
In this talk, Dat Tran will present how his team solves the problem of teaching a computer to recognize images as beautiful or not. In particular, he will share training approaches and the peculiarities of the models including the “little tricks” that were key to solving this problem.
The recent proliferation of digital games for commercial, social, and educational purposes has necessitated a new research direction towards game intelligence and knowledge discovery. The key quest is to enable end-to-end informatics around game dynamics, game platforms, and the players using the immense volume of multi-dimensional data. This talk will cover the fundamental building blocks in the domain of game intelligence and informatics.
It’s so easy for us, as human beings, to just have a glance at a picture and describe it in an appropriate language. Even a 5-year-old could do this with the utmost ease. But, can you write a computer program that takes an image as input and produces a relevant caption as output? Attend this hack session as Rajesh & Souradip tackle automatic image captioning using deep learning.
This panel discussion comprising of leading researchers and industry leaders will try to shed more light on adversarial attacks and robust AI which is need of the hour for AI in enterprise.
If streaming data is what interests you, this is the right place to get an insight on that. Here we discuss what streaming data is, its applications and how to harness the power of online learning to build streaming analytics systems
The existing techniques for time-series classification are very difficult to deploy on the tiny devices due to computation and memory bottleneck. In this talk, we will discuss two new methods: FastGRNN and EMI-RNN that can enable time-series inference on devices as small as Arduino Uno that have just 2KB of RAM!
In this talk Rishabh who is the founder of Locale.ai - a start-up that provides location intelligence solutions will talk about caveats of spatial data, need for spatial modelling and will also cover challenges faced by businesses to run spatial models at scale.
Contextualizing NLP and AI solutions in the healthcare industry, is of paramount importance, as they can have a direct impact on patent safety. The talk will focus on the business impact and successful application of such solutions in the healthcare industry via ATH Precision – ‘Powerful Ecosystem for Business Integrated Analytics’. ATH Precision is a domain agnostic analytics eco-system built on the 4 pillars of – Contextualization, Connection, Collaboration and Institutionalization, that drives the development and application of advanced AI solutions in the healthcare domain, to mitigate business and patient risk.
How do we assess what the representations and computations are that the network learns in natural language processing? The goal of this session is to learn to leverage attention-based models for interpretation of NLP models and also cover some model agnostic techniques for NLP interpretability.
This talk is about developing a Machine Learning model that determines which loan applicants are credible and developing a monitoring framework and feedback loop for the Credit Risk Model.
In this talk, we discuss reinforcement learning and its internal working, sticking to mathematical intuition rather than rigorous equations and derivations. We will also discuss the challenges faced when moving RL from computer simulations and games into the real world and solutions to these problems, with practical examples.
PyTorch continues to gain momentum because of its focus on meeting the needs of researchers, its streamlined workflow for production use, and most of all because of the enthusiastic support it has received from the AI community. While Tensorflow remains ahead when it comes to deploying deep learning models, recent updates in PyTorch has taken it head to head with the former. Vishnu in this code walk through session would take us through deployment of models with PyTorch using flask
Genetic algorithms simulate the process of natural selection which means those species who can adapt to changes in their environment are able to survive and reproduce and go to next generation. Sahil Verma who is well known for winning machine learning competitions will cover how genetic algorithms can be leveraged for tasks such as feature selection, hyperparameter tuning etc.
This session will cover various techniques for extracting valuable features from images that can be used to train simple machine learning models such as logistic regression. Further, the duo will use these techniques to classify images and compare the model performance.
If you’re thinking “I don’t do Kaggle so this session is of no use to me”, don’t – these hacks are not only for Kaggle competitions, they apply to all data science projects. Pavel (Kaggle Rank 4) in this session would reveal the top hacks he has used efficiently to win multiple ML competitions from varied domains.
Hack Session: Reinforcement Learning in Real Life – Implementation Guide
Most content on Reinforcement learning revolves around gaming AI or playground problems. Rarely do we see real life implementation and deployment of RL models. Well look no more because Hardik & Richa from TCS research bring to you this exciting session with real world use cases with implementation in python for solving these cases such as port management.
This talk will discuss the state-of-art technologies available for image classification and present Image Automatic Tagging Machine (ATM) in the context of these technologies. A crash course of Dat's team's product will be given where he will walk through different ways of using it – in shell, on Jupyter Notebook and on the Cloud.
Graphs naturally represent data created in a host of real-world processes, including interactions between people on social or communication networks, relations between entities in a knowledge graph, links between content with its creators and consumers in content platforms, and many others. Sourabh who has won multiple competitions by creating effective graph-based features would do a deep dive on how to build graph-based features and further use them to build high-performance ML models.
This exciting hack session by Xander covers one of the greatest ideas in Deep Learning of the past couple of years: Generative Adversarial Networks. He will first explain how a generative adversarial network (GAN) really works. He will then dive into Nvidia's StyleGAN model and learn how one can manipulate it's latent space to morph arbitrary images of faces.
In this hack session, we will cover the motivations behind developing a robust pipeline for handling handwritten text. Learn about an interesting use case where Deep Learning (DL) techniques are being utilized to generate synthetic data for training along with some interesting architectures for the same.
This hack session presents an introduction to deep-learning based question-answer models. These models by virtue of the underlying transfer learning layer (using contextualized word embeddings such as BERT) can easily find exact answers to factoid questions from a corpus of documents on which they were not trained.
This hack session will walk participants through deploying machine learning models locally and on a cloud platform. An overview of relevant principles from software engineering and DataOps disciplines will be covered with focus on doing all of these at scale.
Here is a unique opportunity for the attendees to understand how the top hackers approach various types of problem statements and competitions at DataHack and Kaggle.
American Express, with its global footprint, has several hundred products catering to different types of customers – individuals, small and large businesses and merchants. They have over 100M Card Members and an ever-growing prospect base. Therefore, one of their biggest challenges is creating solutions to reach customers and prospects with the right product/offer through the right channels. Today, hundreds of response models in production determine their strategy for distributing various offers across channel combinations in different segments and markets. As a part of this hack session, the speakers will address the following in the response model development process:
- Procuring and curating the appropriate data
- Developing the best analytical solution that makes ‘sense’ to our colleagues and also regulators
- Implementing the solution, monitoring its performance, and updating the solution based on learnings
Combined with more traditional content-based recommendation systems, image-based recommendations can help to increase robustness and performance, for example, by better matching a particular customer style. In this hack session, learn how to build content-based recommender systems using image data.
With the ever-growing volume of online information, recommender systems have been an effective strategy to overcome such information overload. This session will be a detailed study of deep learning techniques for recommendation tasks. Further, Janu Verma who works as ML Scientist with CureFit will explore deep learning methods like auto-encoders, graph embeddings as extensions of the matrix factorization.
Improving search results offers huge payoffs for retailers. Machine learning can improve ecommerce search results every time a customer shops on the website, taking into account personal preferences and purchase history. Instead of using traditional search methods like keyword matching, machine learning can generate a search ranking based on relevance for that particular user. Atul and Sonu in this hack session would cover code walkthrough of deep learning search systems along with the deployment aspects for the same.
Just like in our everyday lives, AI and robotics are increasingly a part of our healthcare ecosystem. Currently, healthcare is broken; there’s a shortage of doctors; poor quality of care. There is a dire need to provide assistance to the whole medical industry to improve healthcare. Abhishek in this hack session would cover how one can build automatic medical imaging computer vision models using PyTorch.
Different data sources mean different schema, extraction logic, de-duplication and being in sync with changing data sources, in addition to a number of other challenges. This session will walk the participants through integrating data sources and associated best practices. We’ll also cover the need for data engineering at a broad level.
Solving multiple time series (more than 100 million time series) in a single shot has always been a challenging task for traditional machine learning models. LSTMs are capable of solving multi-time series problems with a capability to learn embeddings of categorical features for each object (time series). This hack session will involve end-to-end Neural Network architecture walkthrough and code running session in PyTorch which includes data loader creation, efficient batching, Categorical Embeddings, Multilayer Perceptron for static features and LSTM for temporal features.
Generating images from natural language is one of the primary applications of recent conditional generative models. Besides testing our ability to model conditional, highly dimensional distributions, text to image synthesis has many exciting and practical applications such as photo editing or computer-aided content creation. Shibshankar will do a live code walkthrough to build a GAN that can create images from just the text description.
Tarry in this one of its kind workshop for industry leaders will give an introduction to the advancements of AI into business areas and how companies are increasingly tackling the three Core challenges inside their organizations today, namely:
1) AI strategy and defining the operating model
2) Creating awareness by establishing strong pillars for AI capabilities and finally
3) Identifying latent potential within organization and training them to become AI experts.
Chatbots are everywhere today, from booking your flight tickets to ordering food, chances are that you have already interacted with one. But do you think making an intelligent chatbot is hard? Do you want to be able to quickly build a reasonably good chatbot for your business? In this workshop, you will build multiple intelligent Chatbots using concepts of Machine Learning, Natural Language Processing and the Open Source RASA framework. You will also learn how to deploy them to various chat applications for example - the Slack messenger!
Trying to master time series but finding it too complex? We have designed this comprehensive workshop just for you! Learn the core components and techniques for time series analysis, how to build time series models in Python, and much more!
Recommender systems find patterns in user behaviour to improve personalized experiences and understand the environment that they are acting in. They are ubiquitous and are most often used to recommend items to users (for example, books and movies on Amazon and Netflix, relevant documentation in large software projects, or papers of interest to scientists). In this workshop, Anand will take recommender system case studies from different industries, refine the business problem to a Machine Learning problem and show how it is scaled up to work for millions of users. He has a knack of building everything from scratch to have an in-depth understanding of the workings of such systems.
Do you want to jumpstart in the Deep Learning? Are you still stuck on grasping basic concepts such as how backpropagation works? Then this workshop is for you! The goal of this workshop is to get you up to speed with the advancements in Deep Learning with a practical perspective. Along with this, you will also get to know about Keras, a tool which has been getting heavy attention from the community as it provides simple higher-level interface to build Deep Learning models. In this workshop, our objective is to make you comfortable to build deep learning based image and text classification models using Keras. And, you will also learn the building blocks of deep learning: MLP (Multi Layer Perceptron), CNN(Convolutional Neural Network) and RNN (Recurrent Neural Network).
Interacting with artificial intelligent systems seems a bit simulated at times. This is because the way we converse as humans to one another is completely different from that we do usually with AI systems. Thankfully, research has been rampant in the area to bridge the gap in conversational AI systems. In this 8-hour workshop, you will get to know about natural language processing, creating word embeddings and developing learners to perform NLP tasks like sentiment analysis, auto-correction and much more.
Have you struggled to improve your model beyond a particular score in a data science competition? Do you wonder what goes inside the minds of top data scientists? Then this workshop is what you are looking for. This is a highly interactive experience with Kaggle Grandmaster Pavel Pleskov & Ankit Choudhary who leads the hackathon category at Analytics Vidhya. The goal of this workshop is to provide a forum for exchanging ideas and new approaches to become masters of data science challenges
Have you ever wondered how to apply machine learning to business problems? The focus of this workshop will be on the machine learning pipeline including data cleaning, feature engineering, model building and evaluation. You will also learn how to structure a business problem as an ML problem, and then go on to build, select and evaluate the model.