ThinkML

My thoughts on Machine Learning and Related.

IJCB 2024 Tutorial

Democratization of Deep Learning - #NeurIPS 2019

Deep learning is one of the most penetrating technologies into various domains, problem areas, and challenges. However, whether you are a student, or a practictioner, or a researcher, or a software engineer, or a business user, you must have asked one of the following questions, or faced one of the following challenges.

Read More >

What am I Doing in Research?

When I am having an unplanned holiday, I tend to have some time for myself - to think about my past, my present, and maybe my future. Especially, when I am having an unplanned holiday because I fell sick, I will reach my heights of introspection. The result of such a deep introspection is the …

Read More >

Automated Bug Triaging

The process of bug triaging consists of multiple steps where the first step primarily involves assigning the bug to one of the developers who has the expertise to solve the bug. In large scale systems, with high rates of incoming bugs, manually analyzing and triaging a bug report is laborious. Thus, can this process of bug triaging be automated?

Read More >

A Journey from Weekend Project to Watson Product

Everyone of us have hobby projects and weekend hack nights. Sometimes outside work we try to pursue our passions and keep our tech hunger satisfied. What if your hobby becomes your full time work, your passion earns you money, and you could do all of these without the need to do a startup? Here is a short journey of how one of my weekend project evolved slowly and steadily into a mature project and got integrated into Watson Studio.

Read More >

DLPaper2Code

Q1: How often do you read a research paper discussing a deep learning algorithm ?

Q2: How much time does it take for you to understand the research paper?

Q3: Assuming that you are successful in understanding the paper, how much time does it take for you to implement the paper in a single library?

Read More >

Human face recognition using DL-IDE

Given an image containing a face image, how to automatically recognize the identity of the person in the image. The face is captured in a highly unconstrained environment with any orientation, pose, expression, and illumination.

In this blog post, I share a cool way to visually design deep learning models for such tasks, in less than two minutes!

Read More >

Cats Vs. Dogs using DL-IDE

Does the given image contains a cat or a dog? Complex and deep learning models have tried to solve this daunting task for over a decade now! How do you develop and implement such complex deep learning models?

In this blog post, I share a cool way to visually design deep learning models for such tasks, in less than two minutes!

Read More >

Semantic Sentence Similarity Models using DL-IDE

Given two sentences, how semantically similar are the two sentences? This is a common task in natural language processing. Deep learning has produced some state-of-the-art results in the task of semantic sentence similarity.

In this blog post, I share a cool way to visually design deep learning models for such tasks, in less than two minutes!

Read More >

Natural Language Regression using DL-IDE

While working with natural langauge processing, we often come across the task of predicting a continuous valued integer output for a given input sentence. This includes tasks such as (i) sentiment prediction for a statement, (ii) impact prediction of a title, (iii) grammatical correctness of a sentence, (iv) importance of a tweet, and many more. This is commonly called as natural language regression.

In this blog post, I share a cool way to visually design deep learning models for such tasks, in less than two minutes!

Read More >

ML4Creativity - Workshop @ KDD-2017, Halifax, Canada

The first workshop on Machine Learning for Creativity (ML4Creativity) was co-hosted with KDD-2017, held at Halifax, Nova Scotia. I was one of the co-organizers of this exciting workshop along with my colleageus from IBM Research, Sony CSL, Google Brain, and UIUC.

In this blog post, I share my summary of this workshop, along with some interesting observations, and key take-ways.

Read More >

DL-IDE - A Visual IDE for Deep Learning

To make deep learning approachable and accessible to a larger developer audience, or even to non-experts, we at IBM Research, India have launched a new IDE for deep learning called IBM Deep Learning IDE (DL-IDE).

Read More >