• Where

  • When

[email protected] +91-95389 89028, +91-96325 33477

Schedules

Event Schedule

We are in the process of finalizing the sessions. Expect more than 30 talks at the summit. Please check back this page again.
Here is our schedule from last year.

Expand All +

  • Day 1

    February 11, 2021

  • Will talk about use of Machine learning in risk management. E.g., What kind of techniques can be used in risk analytics across risk categories (credit/market/operational and PPNR), what are the pros and cons, regulatory view on machine learning models, Challenges for banks using machine learning.
    TECH TALKS

  • Quantum theory provides an unvarying framework for the construction of various modern physical theories. After more than 50 years from its inception, quantum theory married with computer science, another great intellectual triumph of the 20th century and the new subject of quantum computation was born. One of the fertile areas for quantum computing is AI (Artificial Intelligence), which relies on processing huge amounts of complex datasets. There is also a need to evolve algorithms to allow for better learning, reasoning and understanding.
    TECH TALKS

  • Collecting and curating data for AI is not without its challenges, this is even more difficult when it comes to healthcare. This talk introduces the audience to the difficulties in healthcare data collection and what can be done to work with small data sets. We will touch upon the modalities in orthopedics, considerations during data collection, current trends in DL with small datasets, and the ways to make the most of the available data with various augmentation techniques.
    TECH TALKS

  • In today’s fast paced world, businesses cannot shy away from leveraging Artificial Intelligence and Machine Learning to solve critical problems. However, dealing with the complexities in big data while also having to provide rapid results, can be a challenge. Making the journey towards building a ‘ML-savvy’ analytical framework overwhelming. Join us in this Tech-Talk as we uncover the evolution of machine learning and its applications. Through real-world use cases, learn how leading businesses apply machine learning to solve critical problems. The secret is ‘incremental improvements’, after all it’s elementary, my dear ML!
    TECH TALKS

  • Damage assessment is a key component of automobile insurance claim process which is currently manually executed by third parties in general and involves huge amount of cost and resources. There has been many attempts to automate the process in industry but inherent challenges of image capture in uncontrolled environment make it difficult. The proposed solution uses state-of-the-art deep learning architectures and data enhancement approaches to generate millions of images to train the system. The solution is able to precisely segment out body parts and damaged areas at pixel level to generate robust insight for decision for total loss, repair-replace and cost estimates.
    TECH TALKS

  • Passenger Airlines industry is over 100 years old! The advent of many low cost airlines, in many parts of the world, in early 2000s, compelled quicker business transformation for the overall industry. Currently airline leapfrogging in adoption of data science/ ML based decision making, basis technology initiatives & digital transformations. It’s a significant & a massive change, when it is implemented across the core airline functions. At SpiceJet, our journey started with a focus on maximizing revenue. A collaborative effort with key stakeholders, and with the guidance of Executive Management, we focused on a multiple facets of Commercial initiatives, and help improvise decision making. Currently we are also looking into potential cost initiatives that could lead to significant savings for the internal value chain! In this interaction, we will have a quick overview of our experience of developing data science/ ML based solutions.
    TECH TALKS

  • ML-powered digital twins are affordable, scalable, self-sustaining, and, with the right user interface, are extremely useful in telling machine operators the exact condition of the equipment under their care. This talk by Arvind Mahishi and Vighnesh Subramanian will walk you through the journey on how ML powered digital twins can be leveraged to bring about a transformation in how maintenance is carried out in a manufacturing setup
    TECH TALKS

  • This talk underlines the necessity of a custom designed recommender system for a business case comprising of several complex underlying business rules & intertwined ways of working. It will help the audience to think out of the box when off-the-shelf machine learning algorithms don’t completely solve a business problem & warrant a custom-made solution. The services unit of Ericsson wanted to have a recommender system for their resource manager who routes an incoming project demand to an appropriate pool of resources today, based on his experience. The demand routing is considered successful only if the routed resource pool has the expected resources available who can support the delivery of the project. The standard algorithms like apriori, FP-growth etc. were not suitable for this problem because each recommendation needed to factor in the complex feature set of the input project demand & recommend out of a set of 700 pools. The input demand feature set was a mixture of several categorical, time based, text & few numeric features. The problem complexity was further compounded by the need to consider the composition of the resource pools before recommending the suitable resource pools. The AI & Data Innovation team devised a customized Machine Learning solution which was a combination of innovative feature engineering, clustering techniques, classification algorithms & usage of real time resource availability data via APIs. The automated offline training of the model using an Azure Databricks pipeline enables it for quick adaptability. The E2E machine learning solution was implemented as a web service using MLflow, Azure Kubernetes Services. The Machine Learning service’s enhancements are handled by the Azure CICD pipeline. The Machine Learning service also has a 3-tier support structure with Data Scientists at L3.
    TECH TALKS

  • Today, consumers are looking for real life experiences yet personalized, in the targeted content that reaches them. Hence, marketing efforts are heavily associated with the experience endeavors from marketers. The need of the hour for the marketer is to tease insights out of every available information at their disposal and make a sincere effort to reach each consumer at her viewpoint. This is not easy. To accumulate all available information an organization possesses for a particular consumer would mean breaking the silos to create an integrated sales, marketing and experience function. The solution looks at how we can unify and transform customer experience across the life-cycle.
    TECH TALKS

  • With the advent of technology, adversaries engage in intelligent combats, through interesting avenues. Financial institutions are often the targets of money laundering. Anti-Money Laundering Policies (AML) are employed by financial institutions to prevent money laundering done by criminals be they are white-collar, street-level criminals etc. Online banking and crypto-currencies are still susceptible to money laundering, where detection itself takes time and effort. Adversarial examples are inputs designed by adversaries to intentionally cause a mistake in machine learning models. Deep Neural Networks (DNN) are known for their accuracy and are known to outperform human beings. Thus, they are used to a greater extent in Image processing, Natural Language Processing, etc. The very complexity of DNN is utilized by adversaries to their benefit. This session is thus, going to be a primer on adversarial insights and the adversarial examples, in real-world scenarios
    TECH TALKS

  • One of the greatest paradoxes is that paradoxes abound yet go unnoticed. A close second is when they are noticed, their influence on outcomes are often disregarded. Paradoxes proliferate in all kinds of common and complex mathematical applications, including business analytics, machine learning, econometric modeling, statistical reporting, and business inferential and strategic development processes. Some paradoxes are comical, whiles others are odd, and still others are highly counter-intuitive. Yet all have tremendous implications. And in some (many) contexts, the paradoxes can lead to diametrically opposite outcomes than intended.
    TECH TALKS

  • Organizations need to design, hire and manage their analytical teams. All organizations want their analytics teams and function to not only exist but to perform in a highly effective manner. Analytics teams are not like development teams or typical project teams. We will discuss what makes them different and how to hire and manage for high performance and operational impact. In this session we’ll be discussing best practices relating to investing in, hiring and managing your analytics team, including the following topics: • We will discuss how to build an analytics teams using the Artisanal or Modular model. • Analytics team evolve and grow into Hybrid organizations. We will discuss how to manage this evolution. • High performance analytics teams are unique. We will discuss how to best manage them. • We will define how to build robust analytics processes that drive results in the real world • We will examine how best to move from robust analytical processes to integrating analytical models and processes into operational production processes
    TECH TALKS

  • Day 2

    February 12, 2021

  • Automatic identification and classification of weed species are essential in the agricultural field for controlling weed species. Weeds are an undesirable and unfortunate plant that meddles with the usage of land and water assets and along these lines unfavourably influence crop creation and human government assistance. So, identification and classification of weed are important for farmers to protect the crop field and to maintain the productivity and quality of the crop. But it takes a long time and huge human- effort to manually identify and classify weed species. Technology advancement has made complex problems to solve more efficiently and reduces manpower and lower the costs. With technological advancement, many methods have been introduced. Using deep learning methods such as neural networks on agricultural data has increased enormous consideration lately. The evolution of deep learning made it easy for identifying and classifying the weed type. This paper uses publicly available large multiclass image datasets of weed species obtained from Australian rangelands for classification. In this paper, we are using the Transfer Learning technique with a pretrained network called resnet18 to classify the type of weed from the images present in the dataset and also calculating performance metrics like accuracy, sensitivity, recall, precision, etc. This helps in controlling weed species in the crop field.
    Paper Presentations

  • The Agency for Healthcare Research & Quality estimates more than 2.5 million individuals in the United States develop bedsores annually that costs $9.1- $11.6 billion to the healthcare system. In this paper, we develop a low-cost solution to reduce the risk of bedsores for bed-ridden patients using machine learning. Elderly patients with mobility impairments are the highest risk population segment for bedsores (also known as pressure ulcers). Currently, smart beds that send alarms when patients have not changed their position on bed for long time are used to manage the risk of developing bedsores. However, such smart devices are cost prohibitive. The proposed affordable solution uses low-cost load-cells’ readings to accurately estimate the patient position with an accuracy of 98.8%. Specifically, the solution manages bedsore risk by deriving meaningful intuitive features that are used by machine learning model to generate alerts when a patient has been in the same position for a prolonged period of time.
    Paper Presentations

  • Complex pharmacy transactions, especially for patients who have to fill multiple medications periodically, is a challenge. Having to make multiple trips to the pharmacy to fill medication leads to non-adherence which has an impact on health outcomes and overall patient experience. Healthcare organizations which are focused on improving adherence and patient experience are exploring innovative online pharmacy distribution methods. However to be successful with these innovative strategies it is important to segment patients and study behaviors of members that have higher propensity to engage in online pharmacy distribution methods. Targeted intervention of converter personas drastically improves conversion rate. In addition to internal data sources a lot of external data sources like social determinants of health (SDoH), consumer behavior data are critical to building the personas.
    TECH TALKS

  • In recent years, Computer Aided Diagnosis (CAD) systems have been designed for the detection of lung space anomalies. Pneumothorax is an abnormal collection of air in the pleural space between the lung and the chest wall that can result in partial or complete lung collapse [1]. This is a medical emergency in which quick detection and timely intervention can be life-saving. Pneumothorax detection on chest radiographs is important & may be facilitated with the help of image processing and deep learning algorithms. In this study we aim to evaluate the performances of two artificial intelligence systems in detection of pneumothorax on chest radiographs. The AI system was trained on open source datasets obtained from from the US National Institutes of Health (NIH) [2], Society for Imaging Informatics in Medicine (SIIM) [3][4] & private datasets. Two unique approaches were used, one involved processing high-resolution complete images of size 1024x1024px and other involved feeding medium resolution images in portions (segments), each of size 448x448px. The trained AI systems was trained with binary mask as ground truth evaluated by a team of radiologists where, the segmental approach yielded a dice coefficient of 0.72, sensitivity of 0.986, specificity of 0.95, accuracy of 0.9683 and with Area Under Receiver Operating Characteristic Curve (AUROC) of 0.95, while the full image approach yielded an accuracy of 0.9417, dice coefficient of 0.865, sensitivity of 0.9084, specificity of 0.9510 with AUROC of 0.93.
    Paper Presentations

  • The banking industry is going through major digital transformation and at the same time generating much more data than it used to be. The question is how does a bank leverage on these data into actionable insights for betterment of customer servicing, drive operational efficiency and productivity to sustain business growth. The speaker will share their point-of-views and experience on harvesting data for building smart operations utilizing advanced analytics and transforming to data driven operating model.
    TECH TALKS

  • A lot of industries throughout the supply chain pipeline like manufacturing, retail, and even the services supply chain relies on demand forecasts to anticipate and plan. They use the demand forecasts to plan all activities like production speed, resource allocation, raw materials purchase, and even stock levels. They usually rely on standard demand forecasting models that considers historic demand over a period during different sales cycles to determine the future demand. But in certain unforeseen scenarios be it a catastrophic weather event, or a pandemic like the one we are going through, or even a new trend causing a sales spike, the previously forecasted demand might be inaccurate and causes disruptions in the supply chain due to shortage of supply. The paper discusses a method to continuously monitor the external events in real time, stream such events, cluster them and then arrive at an offset in the demand based on what was seen in the past and the context of occurrence of the events, like the point in the sales cycle when the event is occurring, domain attributes like weather and other sociopolitical climate. This helps to increase the reliability of the demand forecasts and helps supply chain planners react to such unforeseen events effectively and improve the resilience of the supply chain.
    Paper Presentations

  • A typical automobile insurance rating plan contains a plethora of risk factors, ranging from driver, vehicle, to policy characteristics. Including the geographical risk characteristics into the pricing has been a challenge owing to its high cardinality. The traditional approach groups the postal codes based on the historical loss experience, which suffers from two major drawbacks: a) For geographies with low exposure, the loss cost is almost always zero b) Low confidence as we lose information on the latent variables. In this paper, we demonstrate a case study of Greece automobile insurance product offered by a major US based P&C provider, where a Geo-score was developed at a postal code level to improve risk segmentation in own damage cover pricing. The base loss cost(loss/exposure) model was built using Tweedie Compound Poisson regression and geospatial attributes are added in the model without changing the existing rating structure. The external attributes like socio- demographic variables and highway/network data is sourced to create geographical clusters using partitioning around medoids (PAM). Further, various high cardinality feature reduction techniques were used to predict the residual loss cost. This paper illustrates the hybrid approach of the target-based encoding methods and XGBoost to create the geo-score.
    Paper Presentations

  • Enterprise Telecom services is an immensely competitive market the world over, with tailor-made tariff plans complimented with rebates makes reaching targeted profits to be an astronomical challenge. Consumers are the lifeline that firms have a huge challenge in retaining today. Customer Churn the standard industry buzz word could turn out to be a nightmare for any established service provider, so whilst there is always a renewed focus in bringing new customers that call for intense investments and resources, maintaining existing ones is relatively inexpensive. The advancements in predictive modelling are one step in the right direction that alleviates churn for telecom service providers. We propose a progressive analytical approach to predict customer churn one month ahead and the churn pool derived by means of consumer segmentation to identify high valued among potential churn customers, aided by advanced analytics that takes into account the churn se verity level, churn priority, etc., to provide a persona-based treatment plan for consumer retention. This will ultimately lead to increased revenue with an ever-growing customer base.
    Paper Presentations

  • How is AI impacting orthopedics? Is it a hype? What has been realized? How does it fit into the entire pipeline from diagnosis to treatment delivery in orthopedics? What happens pre-operatively, intraoperatively, and postoperatively. In this talk, we will try to find some answers to these questions. A software engineer's perspective into this interdisciplinary domain.
    TECH TALKS

  • Abstract—In this work, we present a Deep Reinforcement Learning based approach as a soluion to one of the popular optimization problems, namely “Capacitated Vehicle Routing Problem”. We have benchmarked the results againt genetic algorithm and have evaluated the performance using two KPIs- Travelling cost (distance covered) and Computational time. The comparison shows a 5X-20X reduction in cost and a 100X–1000X reduction in computational time. The Deep Reinforcement Learning based solution adheres to an adaptive learning framework where the system automatically thrives for optimality rather than being explicitly programmed.
    Paper Presentations

  • Cost per click is a common metric to judge digital advertising campaign performance. In this paper we discuss an approach that generates a feature targeting recommendation to optimise cost per click. We also discuss a technique to assign bid prices to features without compromising on the number of features recommended. Our approach utilises impression and click stream data sets corresponding to real time auctions that we have won. The data contains information about device type, website, RTB Exchange ID. We leverage data across all campaigns that we have access to while ensuring that recommendations are sensitive to both individual campaign level features and globally well performing features as well. We model Bid recommendation around the hypothesis that a click is a Bernoulli trial and click stream follows Binomial distribution which is then updated based on live performance ensuring week over week improvement. This approach has been live tested over 10 weeks across 5 campaigns. We see Cost per click gains of 16-60% and click through rate improvement of 42-137%. At the same time, the campaign delivery was competitive.
    Paper Presentations

  • Streamlit is an open-source app framework for Machine Learning and Data Science teams. It is used to create beautiful dashboards and web apps in a very short time. And that too, in pure Python. In this workshop, we will learn how to deploy an ML model on streamlit and let users explore the following functionalities- Upload datasets Play around while doing Exploratory Data Analysis (EDA) Run ML algorithm Explore output and visualize it aesthetically We will also learn how to "beautify" our apps by adding buttons, sliders, checkbox, etc. Prerequsites - Python (Basic to Intermediate) Streamlit installed in your system (pip install streamlit)
    TECH TALKS

  • 0110
    Paper Presentations

  • With the advancement of technology and globalization cause exponentially increase in unstructured data. As the data grows redundancy/duplication in data also rise and it’s becoming difficult to categorize similar kind of data. Unsupervised theme identification and duplication removal techniques provides us a way to deal with such kind of problems by understanding the structure of data, remove duplicate words and tag the sentences to its most suitable theme and subtheme by which identical data can be distinguished among unstructured data. It reduces the tedious task of labelling the data and give better understanding enormous data of any business with the use of machine learning and taxonomy-based techniques. The objective of this paper is to show the capability of this technique on the news article related to covid-19, classify articles to most suited area(pharma, bank, restaurants) and visualize the categorized data to get impact of covid-19 on these areas.
    Paper Presentations

  • Semantic text matching is the task of estimating semantic similarity between source and target text pieces and has applications in various problems like query-to-document matching, web search, question answering, conversational chatbots, recommendation system etc. We will talk about building machine learning system for large scale semantic text matching by decomposing the problem into candidate generation and re-ranking steps. We will also talk about how deep learning models like BERT can be used for each of these steps.
    TECH TALKS

  • Abstract—Millions of people have been infected by the coronavirus disease of 2019 (COVID-19) and lost their lives to it despite various measures to curb the same. To make the situation worse, a traditional observational method of in-person reporting cannot be used because it poses a risk for the observer to catch the infection. Social Distancing and Face Mask compliance, therefore, remain vital measures to curb the spread of COVID-19. We propose an end-to-end solution that can monitor different social distancing and face mask compliance metrics and be deployed efficiently in Python using open-source libraries. It is scalable and enables the users to implement the solution at a large scale, i.e., cover a broader area using multiple live camera feeds simultaneously. Our solution precisely calculates the distance between two people or objects by mapping the 2-dimensional pixel distances to 3-dimensional actual distances. These attributes make our solution unique, and it can be deployed for usage in various situations and locations such as shopping malls, supermarkets, large workspaces, manufacturing facilities, etc. which can help to dampen the effect of COVID-19 as early as possible
    Paper Presentations

  • The SEIR model is used to study the spread of epidemics. The parameters here are assumed to be specific to not only the disease but the country/region. We have applied ML techniques to solve the inverse problem for epidemics based on data, deviating from traditional approaches and stochastic models. Since real situations have far more variables, our new approach is amenable to extension by adding more aspects as needed. Our algorithm uses a multi-step SEIR model approach with the state-space search for parameters. The data is taken from publicly available data sources (primarily Kaggle). We present a Machine Learning based algorithm to estimate parameters with windowing. A custom accuracy score is used to choose the right window for estimating parameters. Time/Window dependent parameter (alpha, beta, and gamma) values present an understanding of measures taken by countries and are learning that can be shared in real-time. All the code in the literate programming paradigm is in the public domain
    TECH TALKS

  • Journey of a cognitive solution is meaningful when it's put to use or can actually solve business problems in real time through inference. Deep Learning model Inference is as important as model training and especially when it comes to deploying cognitive solutions on the edge, inference becomes a lot more critical as it also controls the performance and accuracy of the implemented solution. For a given computer vision application, once the deep learning model is trained, the next step would be to ensure it is deployment/production ready, which requires application and model to be efficient and reliable. It's very essential to maintain a healthy balance between model performance/accuracy and inference time. Inference time decides the running cost for “on the cloud” solutions and cost optimal “on the edge” solutions come with processing speed and memory constraints, so it's important to have memory optimal and real time (lower processing time) deep learning models. With the rising use of Augmented Reality, Facial Recognition, Facial Authentication and Voice assistants that require real time processing, developers are looking for newer and more effective ways of reducing the size/memory and amount of compute required for the application of neural networks.
    Paper Presentations

  • Building a product requires expertise to prepare a framework and then applying the model to arrive at a result and then visualizing the results. It also involves a considerable effort to put in for the data pipeline. To build a web application or on a Kotlin/ Flutter application requires the skill of a full stack developer. This requires some specialized skills and my session will help the people to understand how to develop this product.
    TECH TALKS

  • - Application of DS in fintech & at PayU – customer acquisition and retention, fraud detection, risk analysis, etc - Job industry scenario and career opportunities in the field - Skills that companies such as PayU look for in a candidate - Careers in DS at PayU
    TECH TALKS

  • The role of AI in healthcare will have an impact in all our lives owing to the change it brings to the Patientcare system, changing the traditional way of handling illness and diseases. Most of the AI based applications use Machine Learning algorithms that use data. Hence the source of Data and the nature of Data holds the key in developing effective AI based solutions for many health issues in the society. Though Data is available in all the hospitals and medical care facilities for many years now. They cannot be used directly to develop AI based applications until and unless they are transformed and made it into a format, on which machine learning algorithms can work. In this research paper, we discuss the process of developing an AI based application to predict Dengue, one of the vector borne diseases based on the symptoms. Our work was done on the data collected from the clinical notes of a 1500 bed hospital in the coastal district of Karnataka. We have implemented few of the machine algorithms like Logistic regression, Support vector machine and Decision Tree classifier. As the dataset is highly imbalanced (1:50), we applied over sampling techniques (Random over sampling, SMOTE) to overcome this problem. We compared the over sampling techniques and find that combination of SMOTE and Decision Tree classifier gave the best result (98% F1 micro score) compared to the other algorithms that we used in thisstudy.
    Paper Presentations

  • Productionalisation of ML models is a very complex process involving numerous stakeholders such as product managers, data engineers, data scientists, DevOps engineers, and MLOps engineers. The complexity of their interaction is compounded by the unique nature of ML projects which tend to have high levels of uncertainty and experimentation as well as a large arsenal of tools, languages, and algorithms at the disposal of data scientists. Today, estimates indicate that 50 to 80% of ML models never make it into production and for the ones that do, most are prone to numerous failures and operational issues. The reasons for this sad state of affairs range from lack of organizational support to unavailability of tried and tested frameworks. In this presentation, we will take a journey through the ML Lifecycle and share the challenges which happen throughout the cycle thereby increasing productionalization difficulties. The intention of the presentation is not to focus on how to do these steps but rather to share common errors and mistakes that I have seen (or made!) with the objective of raising the awareness of the practitioner so that they can be better equipped to build sustainable ML systems. The stages we will explore are Project Goal Definition, Data Collection and Preparation, Data Preprocessing, Feature Engineering, Model Training and Evaluation, Model Deployment and Monitoring.
    Workshops

  • In spite of showing unreasonable effectiveness in modalities like Text and Image, Deep Learning has always lagged Gradient Boosting in tabular data—both in popularity and performance. But recently there have been newer models created specifically for tabular data, which is pushing the performance bar. But popularity is still a challenge because there is no easy, ready-to-use library like Sci-Kit Learn for deep learning. PyTorch Tabular is a new deep learning library which makes working with Deep Learning and tabular data easy and fast. It is a library built on top of PyTorch and PyTorch Lightning and works on pandas dataframes directly. Many SOTA models like NODE and TabNet are already integrated and implemented in the library with a unified API. PyTorch Tabular is designed to be easily extensible for researchers, simple for practitioners, and robust in industrial deployments. The library is available at https://github.com/manujosephv/pytorch_tabular
    Paper Presentations

  • Today’s retail market is consumer driven and has become quite competitive. Shopper have wide range of options to choose in terms of products as well as in terms of stores. This has made retailers to re-look at their merchandising strategies more precisely to understand demand at a more granular level. Most retailers group stores into clusters and take strategic decisions on pricing, promotion, assortment & marketing specific to cluster behaviour. This paper provides a clustering solution by comparing two approaches i.e. product-based vs shopping mission based. Our objective to identify group of stores is to provide customer-centric assortment which in turn improves the customer shopping experience and cut down the stock cost by removing non-performing lines which eventually drives retailer growth.
    Paper Presentations

  • In retail, taxonomy is a hierarchal and logical arrangement of products such that customers can easily navigate and find what they need in the store or website. Taxonomists, information scientists, and linguistics experts all collaborate to build an effective taxonomy. Clearly this requires a lot of resources in terms of time and effort. It is not always feasible for companies to put these resources for all the products, especially newly launched products. In this research, we have developed a novel machine learning algorithm to predict a product’s taxonomy by leveraging N-gram Mixture Model, cross-entropy function, and Newton’s optimization method. A modified Naïve Bayes and up to 4-gram models are combined with general heuristics inspired by Jaccard Similarity. A One-vs-all classifier is trained with weights for combining different n-gram models and heuristics scores using cross-entropy loss function and Newton’s optimization method. This model is developed and tested on online retail data. The model predicts the correct product taxonomy in 84% of the cases using online retail data.
    Paper Presentations

  • Classification of different plant leaf diseases using multiple convolutional neural network and image pre- processing
    Paper Presentations

  • Day 2-3 | Hackathon


  • Welcome to MLDS 2021 Hackathon powered by Analyttica TreasureHunt LEAPS
    It is time for the global data science community to be challenged in creating solutions that drive businesses. The aim is to bring together some of the brightest minds in a “Machine Learning” theme-based event.
    This HACKATHON is a day long-event where people from all over the world would have chance to contribute to a real-world challenge and come up with valuable insights.
    The topic of the hackathon will be released 24 hours before the start date and time, along with all the details that will help you to understand the problem that you will be solving. This will give you a glimpse of the problems that the data science professionals solve on regular basis.
    Who should attend?
    • All Working Professionals: Data Analysts, Statisticians, Data Scientists, Engineers, Students and First Time Hackers
    • Anyone who wants to be a part of the global data science community.
    • Anyone who is looking for recognition at a global level.
    Practice:
    Do you have the mettle but do not know how to join data science hackathon community? Fear Not!
    Participate in any of our live or practice hackathons on the experiential learning platform Analyttica TreasureHunt LEAPS.
    Further, you can learn and practice with 20+ premium courses and 150+ Industry Use cases to build your understanding of the domain.
    To know more, visit https://leaps.analyttica.com/hackathons
    Important Dates:
    Case details announcement: 12 Feb 2021
    Start of Hackathon: 10:00 AM 12 Feb 2021
    End of Hackathon: - 10:00 AM 13 Feb 2021
    Winners announcement: 04:00 PM 13 Feb 2021
    Hackathon

  • Day 3 - Workshops

    February 13, 2021

  • When designing the model architecture, you are presented with a range of hyperparameters that you want to pick for maneuvering the solution of a problem. Ideally, we want the machine to explore these choices itself and select the optimal one. The process of exploring for the optimal model architecture is known as hyperparameter tuning. In this workshop, we will learn how to leverage hyperparameter tuning in machine learning algorithms for an added performance boost. This will be a hands-on session focused on finding the solution to a business problem, thereby enabling the audience to LEARN, APPLY and SOLVE. The session will be conducted through Analyttica’s LEAPS platform, to follow the session hands-on, for a step-by-step execution, do login to the platform here, https://leaps.analyttica.com
    Workshops

  • Covers the following: Tensor Layers, Tensor Ops Model types - sequential, functional, custom Build custom layers Activation functions and optimizers Wide, deep and residual models Comparing DNN models Brief into CNN \ RNN
    Workshops

  • All business problems require optimal decisions given different situations. Descriptive and predictive analytics becomes input in a complex solution to arrive at optimal decision making. This talk will help participants get a great overview of this process and also help them to understand steps to create optimization framework
    Workshops

  • Medical Image Processing is critical for surgical applications that require surgeons to see the bones or tissues in a patient’s body before even making incisions on the patient’s skin. Radiographs are used to visualize the bones and plan surgeries pre-operatively or navigate to planned location intra operatively. The first step in this image processing is to extract the area that contains the target object. Any application dependent on these images has to be able to be robust to the variability in the imaging devices from different vendors. The images can vary in illumination and contrast as well as they may have some obstructions and the shape of the view area might be different. One of the challenges is to extract the view area of the images in presence of noise. In this workshop, we look at problems simplified for the ease of learning that directly relates to this important medical imaging problem with its solutions in machine learning along with some use cases in supervised machine learning. Image Segmentation is one of the most prominently solved problems in the medical domain. In the recent years, one of the Deep Learning models that has been widely used to solve this problem is the U-Net architecture and its variations. We explore more about this architecture and important elements of a typical image segmentation pipeline to solve a real world image segmentation data set. Pre-requisite for the people attending the workshop - Presenters will conduct the session using Jupyter and/or Google colab. Problem 1 – Working knowledge of Numpy, Pandas, Scikit-learn on Jupyter or any platform. Problem 2 - Google colab account setup, working knowledge of Tensorflow 2.x.
    Workshops

  • The objective of “Video Action Recognition (VAR)” is to detect/localize an action or an activity in three dimensional spatio-temporal video data. Recently, advances in deep learning has significantly improved robustness of VAR and opened up wide range of applications in automatic task monitoring in domain such as medical, entertainment, visual surveillance etc. This technology also has application in content based video retrieval, video summarization. In this workshop, we shall cover basics of video action recognition using deep learning and through coding hands-on impart practical knowledge on implementation specifics.
    Workshops

Check Last Year's Schedule

Schedule 2019

Extraordinary Speakers

Meet the best Machine Learning Practitioners & Researchers from the country.

  • Virtual Pass

    Access the live sessions in online mode.
  • Access to all tracks & workshops in online format only
  • Access the recorded sessions later
  • Access to networking with online attendees only
  • ₹2000 + taxes
  • Book Now
Recommended

  • In-Person Pass

    ONLY 200 PASSES AVAILABLE
  • Access to all in-person tracks & workshops
  • Access the recorded sessions later
  • Access to in-person networking with attendees & speakers
  • Conference Lunch for 2 days
  • Group Discount available
  • ₹8000 + taxes
  • Book Now