Previous proposalContextual Autocomplete suggestions in Realtime
The session throws light on the Bayesian sampling technique. This is a much sought-after sampling technique when the data is highly complex and resembles a typical real-world scenario. A step-by-step explanation of transformations and techniques needed to yield a perfect sample along with evaluation metrics is covered. The classes of algorithms used to carry out the process are Auto-encoder, Bayesian modeling, and Monte Carlo Markov Chain.
The target audience range from ‘data scientists who intend to learn and implement probabilistic modeling on big data’ to ‘tech managers for the breadth and depth of probabilistic algorithms’.
Key takeaways: Bayesian modeling, Auto-encoders, Monte Carlo Markov Chain, Business use-cases of sampling.
The session aims to explain a much sought after sampling technique, called Bayesian sampling in detail. It first discusses the challenges associated with real-world data in the Advertisement Tech industry for sampling or any other form of analysis. To mitigate these challenges, data is first transformed into a latent space using Auto-encoders. A detailed explanation of the workings of Auto-encoder is also covered. This transformed data is then used for Bayesian sampling through a category of algorithms, called Monte Carlo Markov Chain. Before getting onto the details of the Monte Carlo Markov Chain, a couple of pre-requisites are discussed. These pre-requisites mainly include the understanding of Bayesian modeling and its inference. The session is concluded with the intricacies of Monte Carlo Markov Chain’s implementation, followed by a brief description of business use-cases.
I’ve been in the Data Science field for 5 years now. In the last 3 years, I’ve worked on a variety of Machine learning problems in the Adtech domain, ranging from Pricing models, Probabilistic models, Recommendation Systems, to Generative models. Prior to that, my projects were mainly in Computer Vision.
I have a special inclination towards Probabilistic modeling and Generating modeling. Sampling, Correction of sampling bias, Variational auto-encoders, Generative Adversarial networks, are some of the projects that I’ve extensively worked on.
I have also been associated with Springboard as an Artificial Intelligence course mentor and teach students on a weekly basis.
To sum it up, I’ve good working and theoretical knowledge of Machine Learning algorithms, especially Probabilistic modeling and I’m an effective communicator.