ACM-W India Grad Cohort 2020: Day 1 Of The Women Virtual Workshop In Computing Opens With A Bang!
“When women do better, economies do better.” — Christine Lagarde, renowned politician, lawyer, President of the European Central Bank and ex-chair and managing director of the International Monetary Fund
With around 125 participants joining the event online, the ACM India Grad Cohort 2020 witnessed a booming start on 24th July 2020.
The third pan-India workshop for women in computing, this three-day event is co-organized by ACM-W India and the discipline of Computer Science and Engineering at the Indian Institute of Technology Gandhinagar.
The workshop kickstarted with a warm welcome address by Neeldhara Misra (faculty in CSE at IIT Gandhinagar and member of ACM-W), followed by the opening remarks from Heena Timani (chairperson of ACM-W India and co-founder and director of a data-analytics startup) and Nutan Limaye (faculty in CSE at IIT Bombay and vice-chairperson of ACM-W).
Shaping Futures
Providing a glimpse of ACM-W and its wing in India, Dr. Timani mentioned that this organization harnesses the talent of women who wish to spread the spirit of communities, supporting other women in computing. Aiming to increase the participation and impact of women in computer science education and research, it enables them to share experiences, acquire skills and establish collaborations, and provides excellent networking opportunities. With its very diverse activities such as the summer and winter schools, student and professional chapters, hackathons, scholarships, conferences, and regional celebrations, to name a few, the long term objective of this community is to positively shape the future of our societies by celebrating and advocating women in computing. Given the present pandemic situation, the majority of upcoming activities organized by ACM-W will be conducted virtually. Presently, there are 36 active chapters in the country, with three professional ones in Chennai, Delhi, and Pune. Dr. Timani also shared some interesting highlights from several events that happened last year.
Connect. Inspire. Succeed.
Discussing the primary goals of the Grad Cohorts, Dr. Limaye explained that the focus is to reach out to Indian women graduate students in the field of computational technology. In some sense, the current virtual event has given everyone a chance to connect with even more people this time. Over the years, ACM-W hopes to create a women community that brings together leaders, researchers, industrialists, and students in computing. It is a space to find inspiration and mentors as well as the right interaction with peers so that everyone can help each other in their journey. Giving some more insight into these gatherings, she further expressed that such events were initiated in the USA from 2004. Eventually, this approach was adopted to the Indian setting with its first edition conducted at IIT Bombay in 2018 and the second one happened at IIT Delhi in 2019. The present event is the third installment happening virtually at IIT Gandhinagar. This brief introduction was followed by a small guide-session by Meenakshi D’Souza (faculty at IIIT Bangalore) during which she explained the logistics of this event.
Machine Learning — Popular And Powerful!
The first keynote lecture of this workshop was delivered by Sunita Sarawagi who is a professor at IIT Bombay and the recipient of the Infosys Prize 2019 in Engineering and Computer Science for her pioneering work in developing information extraction techniques for unstructured data. Centered on the journey of machine learning models, it talked about their birth and went all the way to shed some light on how they are serving the real world in the present times.
According to Dr. Sarawagi, machine learning was still being incubated until about ten years ago and was tended to by a few thousand researchers across the world with a handful of applications. Since then, it has grown into one of the most popular fields of not just computer science but possibly the whole of science-related domains! ML has an amazing number of applications. Suppose, the input in the form of an image can be utilized in tasks like surveillance and medical diagnostics. Similarly, there are thousands of other real-world uses of ML models depending on the type of inputs they receive.
Even though the set of applications is very large, the core-algorithm that is used for creating such models is simple. The training objective is based on a principle from statistics known as the maximum likelihood estimation. To carry out the classification, one needs to design a conditional distribution. In between all this, one of the important factors is an estimation of how well any ML model performs when one applies it to new examples — a model is useless unless it generalizes to new samples (datasets)! This is a challenging problem given the diversity of data. We want models that can generalize to novel situations by getting trained with a finite amount of data.
Neural networks are one of the types of classifiers that have been around since the ’80s. Basically, they contain many linear classifiers side-by-side, and the output of one such function is fed to another function layer. Ultimately, a final score is obtained. This method showed some really powerful properties viz., high capacity, but until 2010 people thought that it was too hard to train, unstable and slow. This thinking changed around 2012 because of several reasons. Firstly, there was a sudden explosion in the amount of labeled data available for training. The second critical factor was the hardware advances with time. Another boost was seen in the optimization algorithm front coupled with a few engineering tricks. The final result was record-breaking accuracy (capable of surpassing the human capacity) that gave ML a lot of prominence in many tasks (such as object and speech recognition, translation, etc.)!
With this, the world entered into the new age of mega-models. These ML models can be trained easily as compared to earlier times on billions of parameters with the help of advanced GPUs and an impressively large number of other resources that we possess. They can be deployed in many different settings. In short, these models have high memorizing power and hence, great performance. However, a handful of challenges still need to be addressed. One of them is the domain adaptation problem, an example of which can be the issues faced during deploying a speech recognition model that is trained using data of North American accent into an Indian setting where the accent will, of course, be different. Essentially, the current methods of training often make the parameters biased towards the majority distribution.
A way to overcome this hurdle is the practice of fine-tuning. But, it is very tricky to implement — if gone wrong, it can cause parts of the useful knowledge learned by a model to be wiped off! An alternative to this approach is retraining the model on the union of user-provided data and revisited training data. Dr. Sarawagi described that it has shown to be much better as compared to fine-tuning. Another scenario is when one cannot revisit the source as sometimes, the data may not be available. In this situation, one can utilize another popular option called meta-learning. Yet, many of its methods have been found to be slow and contain black-boxes. The trick to solve this is having a simple recipe for efficient and reliable training. The main idea is how to choose domain-specific parameters so that they can be trained using a limited amount of domain data. Dr. Sarawagi concluded her talk with the message that although a lot of work has been done in this area, a lot more still needs to be done and this is the reason it continues to be an exciting direction for further research in computing.
The Different Career Paths
Titled ‘Higher Education: India v/s abroad + Masters vs. Ph.D.’, the next session was given by Manik Gupta, a current faculty member in computer science at BITS Pilani (Hyderabad) and an ex-software developer of the industry. She provided her perspectives on how young women can think about their careers in computing.
There are various options to go forward in one’s career. For starters, there are people who join the industry after completing their bachelor’s. There is also a set that goes directly from bachelor’s to pursue master’s, and yet another group does master’s but after working for a few years in the industry. Then, there is the option of going for a full-time or part-time master’s or research. Several integrated learning programs are also available. After a Ph.D., one can pursue a job in the industry or get involved in academia through a postdoc and/or professorship. Furthermore, one can transition between a career in industry and academia depending on phases of life. Apart from this, there are many other options available such as initiating a business or venturing into management, and so forth. These paths can be interrelated at various levels.
Each of these career paths has some pros and cons associated with them. Tangible outputs, monetary benefits, travel opportunities, and remote-working are some of the positives of working in the CS/IT industry. However, cutthroat competition and long working hours can be some of the hurdles in the way of one’s aim to achieve work-life balance. Developing expertise, a good track record, and planning one’s career breaks judiciously can be considered to be some of the tips to achieve a decent career in the industry. On the other hand, a lot of freedom is associated with the research domain in terms of academia. Some other plus points consist of flexibility, networking opportunities, and collaborations. Since a job in academia is often a collection of research and teaching activities, there is a need to have proper time and resource management skills. A few pro-tips include effective guidance from seniors and the importance of publication in this area, along with a good scholarly profile.
Research groups, research themes, and funding are some of the critical criteria for pursuing higher education in India as well as abroad. It is crucial to speak with the supervisor with whom one wants to work. Some pros of going for educational institutes in India for the students of this country are being closer to home and less expensive quality education. Also, dedicated coursework (that is a part of the Indian curriculum) helps in acquiring detailed knowledge of a particular domain and one can work on problems focused on India. But, there is a need to work on specific courses that hone the reading and writing skills of students (helpful in framing research articles and grant applications) along with trying out additional PG degrees in teaching practice — these are some of the USPs of the colleges and universities of the United Kingdom.
In the end, Dr. Manik Gupta encouraged the participants to embrace womanhood, be focused, work hard, carve their own career paths, and define their own success. Different plans work for different people based on their life-situations, whether it is industry or academia or industry-cum-academia.
Ikigai — The Thing That’s Worth Living For
Bhavana Kanukurthi, faculty in Computer and Automation at IISc Bangalore, spoke on the topic of choosing a research topic, advisor, and group. With research interests in cryptography, she has been a recipient of the 2010 Research Excellence Award from the CS department at Boston University. Opening her lecture with the Japanese term Ikigai, denoting the thing that makes one jump out of the bed every morning, she emphasized that we should pursue our research in something that we are good at. Our research interest should align with our skills and be something that we love to do. But, at the same time, it should also be practically relevant and we should be able to get paid for it.
There are multiple aspects to choosing a research topic — the importance of courses, the biases that one has, and career plans being a few of them. Courses are an excellent starting point to figure out what we like and can help us delve deeper into research. Moving forward, Dr. Kanukurthi talked about how certain public perceptions make us biased towards something over another and motivated participants to determine topics on which they would like to work, based on their own interests (and not of others). The key is to work with an open mind — be open to exploring new ideas! Career plans are based on some vital assessment questions — do we like to work alone or in collaboration? What are our financial ambitions/needs? Such questions can assist us to determine where we want to be careerwise.
The role of research advisors is to provide encouragement and critical feedback to their students at different points of time. A lot of contacts that students develop in the research community will be through their advisors. The bottom line is that there has to be a personality-match between a student and a supervisor. Although it is a personal choice depending on what works for whom, we can actually make an informed decision if we have well-planned meetings with our potential advisors and their research groups. Students should meet different faculty members in the same way as the latter meet different students while deciding the research groups that they want to be a part of. They should also talk to the current members of these groups to get an idea of parameters like productivity, connectivity, success, and group dynamics.
Aparna Taneja, a software engineer at Google research, talked about the practical aspect of this topic. She started by walking us through her own dissertation — building and keeping a map up to date. A lot of work is being done nowadays in building nice 3D models of urban environments and this was the motivation behind her research. Currently, she is a part of the team known as the Earth Observation Science in Google. The mission of her group is to use machine learning to solve geoscience problems through earth observation (satellite) imagery, encouraged by applications in climate adaptation and social impact. Satellites capture different types of images (multicentral as well as longitudinal) that can be utilized for tasks like averting disasters, monitoring fires, forecasting floods, crop-type mapping, measuring soil moisture and water distribution, estimating land cover, etc.
Digital Identity — A Part Of Personal Branding
Next followed the session on online presence and personal branding by Jaya Sreevalsan Nair (faculty at IIIT Bangalore). Giving examples of eminent minds in diverse fields, she stated that online presence is one of the components of personal branding. Individuality plays a critical role here and there are innumerable ways to build one’s individuality. One of the ways to portray it is one’s digital/cyber identity.
There is the concept of 5W2H called the Quintilian hexameter, given by Marcus Fabius Quintilian, who was a Roman educator and an expert in rhetorics. It is basically a questioning tool that can aid us to understand what online presence is all about! Who enables digital identity — this could be where one is working or an external organization. What is one’s online presence — this can refer to the causes that one stands for. Why does one need a digital presence — the most important point here is credibility (legitimizing our identity). How to be present online is the next question and there are various ways (different platforms, creative aspects, and so on) to do the same. How much of this online presence — we can manage most parts, but still, there are some that we cannot control depending on the type of work we do. When to take control of one’s online presence — career transitions are good options and it is built over time. Where to start — we can have our own websites and pages showcasing our research and projects.
Dr. Nair said that we can look at our online presence as a message that we send out to the world! Hence, it should have four layers like the four-sides model given by Friedemann Schulz von Thun — fact, self-revelation, relationship (with consumers), and appeal. There is also a time factor involved in building one’s online presence and regular updates are crucial. The final piece of the puzzle is the proper combination of one’s online and offline persona. All these aspects add up to form our personal branding.
Work-Life Balance And Working-From-Home
The first day of this workshop concluded with a thought-provoking panel discussion that touched upon some aspects of remote working and maintaining a proper balance between work and life. Moderated by Dr. Meenakshi D’Souza, the panelists of this session included Tulika Mitra (Provost’s Chair Professor of Computer Science in School of Computing at the National University of Singapore), Joycee Mekie (faculty at IITGN), Hemangee Kapoor (faculty at IIT Guwahati), Rekha Singhal (senior research scientist with TCS Research and Innovation) and Akanksha Agrawal (postdoctoral researcher in the Department of Computer Science at Ben-Gurion University of the Negev, Israel).
They talked about their personal and professional experiences and also shared some stories of their journeys in the field of computing. They motivated the participants and advised them on how to carve their own places in the domain of computer science.
Upcoming Teasers
Day 2 of this virtual event will begin in a few hours. Wanna know what it will house? Well, for starters, an interesting keynote lecture by Meena Mahajan, professor at the Institute of Mathematical Sciences (India), coupled with captivating sessions on the latest topics such as Confidence and Motivation, Quality versus Quantity Publishing and an intriguing panel discussion on Dealing with Uncertainty & Other Challenges During Graduate School! And, last but not least, screening of the documentary film based on the life and works of Maryam Mirzakhani, the first woman and the first Iranian to be honored by the highest prize in mathematics (the Fields Medal) in 2014. So, stay tuned for more updates!
