Transfer learning is a machine learning \ deep learning technique where knowledge gained during training in one set of machine learning problem can be used to train other similar types of problems. This is an extremely useful approach to leveraging pre-trained models to solve real-world problems having constraints and limitations of less data availability.

This talk will cover essentials around deep learning and transfer learning concepts. The various methodologies of transfer learning. We will then look at diverse ways of how transfer learning can be applied in the real-world on complex problems around the following areas.

  • Computer Vision
  • Natural Language Processing
  • Audio Categorization

We will briefly look at a multitude of real-world case studies and problems around the preceding areas like text classification, image classification, image captioning, style transfer and audio classification.


Outline/Structure of the Talk

  1. Brief into Machine Learning & Deep Learning
  2. What is Transfer Learning
  3. Strategies for Transfer Learning
  4. The power of Transfer Learning - solving a small data constraint problem
  5. Introduction to pre-trained models for computer vision - VGG, Inception etc
  6. Transfer Learning case-studies for computer vision
    1. Image Classification
    2. Style Transfer
  7. Transfer Learning case-studies for audio
    1. Audio Classification
  8. Transfer Learning case-studies for text
    1. Text Classification
    2. Image Captioning
  9. Conclusion & Wrap-up

Learning Outcome

  • Learn essential concepts pertaining to deep learning and transfer learning
  • Learn about effective strategies for transfer learning
  • Get to know real-world applications of transfer learning
  • Get an in-depth perspective of complex problems and possible solutions in computer vision, audio and text.

Target Audience

Data Scientists, Data Enthusiasts & Anyone interested in Deep Learning

Prerequisites for Attendees

Knowing basics around machine learning (types) and deep learning concepts (layers, architectures) helps. However the talk will cover these areas briefly before we get started, so anyone should still be able to follow.

schedule Submitted 3 years ago

  • Dipanjan Sarkar

    Dipanjan Sarkar - The Art of Effective Visualization of Multi-dimensional Data - A hands-on Approach

    45 Mins

    Descriptive Analytics is one of the core components of any analysis life-cycle pertaining to a data science project or even specific research. Data aggregation, summarization and visualization are some of the main pillars supporting this area of data analysis. However, dealing with multi-dimensional datasets with typically more than two attributes start causing problems, since our medium of data analysis and communication is typically restricted to two dimensions. We will explore some effective strategies of visualizing data in multiple dimensions (ranging from 1-D up to 6-D) using a hands-on approach with Python and popular open-source visualization libraries like matplotlib and seaborn. We will also do a brief coverage on excellent R visualization libraries like ggplot if we have time.

    BONUS: We will also look at ways to visualize unstructured data with several dimensions including text, images and audio!

  • 480 Mins

    Modern statistics has become almost synonymous with machine learning - a collection of techniques that utilize today's incredible computing power. Jared Lander walks you through the available methods for implementing machine learning algorithms in R and explores underlying theories such as the elastic net and boosted trees.

    • Building the design matrix
    • Penalized regression with the lasso and ridge methods
    • Fitting models with glmnet
    • Interactive visualization of the coefficient path
    • Use cross-validation to choose the optimal lambda
    • Visualize coefficients with coefplot
    • Perform binary classification with a single tree with xgboost
    • Train a boosted tree
    • Tune xgboost hyperparameters
    • Use validation data to understand performance
    • Visualize variable importance
    • Train a boosted random forest with xgboost
  • Dipanjan Sarkar

    Dipanjan Sarkar - Human Interpretable Machine Learning  — The Need and Importance of Model Interpretation (with hands-on examples)

    45 Mins

    The field of Machine Learning has gone through some phenomenal changes over the last decade. 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 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. In this talk, I will be covering the need and importance of human interpretable machine learning approaches, look at effective strategies for model interpretation and several hands-on examples. Detailed coverage of open-source frameworks for machine learning model interpretation will also be one of the major focus areas. Examples will be showcased in Python.