Friday, December 16, 2016

GHCI 2016 - a review


Conferences are always good to attend for many reasons like catching up with latest trends, what & where the industry is heading and of course goodies. Grace Hopper Conference 2016 was organized as a 3 day conference and is special for it had a theme of "Celebration" of women in computing which meant lot of technical tracks alongside motivational career track tailored for women. This is the first time I attended this 3-day conference. Being a family woman with 2 kids, I decided to attend the main 2 day event and skipped the first day event having some key note sessions. Nevertheless, I wanted to share my experiences and thoughts about this conference.




My inclination towards attending this conference was to understand the trend in Machine Learning and Data Science (ML & DS) and also in general how women are progressing in their technology career, the struggles & difficulties faced and how they circumvented them.

The 2nd day conference had following 4 tracks namely ML & DS, Open Source, Human Comuputer Interaction (HCI), Emerging Technology, of which I decided to focus on ML & DS.
The first presentation on "Data Science for Health Care" topic was about "Pneumonia detection using ML". It gave me a strikingly different perception about how ML is been used for pneumonia detection and early intervention. With the help of previous history of factors leading to pneumonia, doctors could now predict if a patient's symptoms could lead to pneumonia. The second presentation was also related to health care, where in the researcher deployed ML algorithms to understand bio-molecular interaction of a particular protein type. Though it was a bit difficult to follow, what made it interesting was, how she used SVM, Random Forest algorith,s on her dataset, undelining the fact that both could not extract much features and is not effective in prediction. I could learn, how to evaluate the applicability of a ML algorithm to any domain.

Followed this was the talk I was waiting for by our own Cisco Veteran Mr. Srikanth Narasimhan on how deep learning is applied to Business Applications. He shed light on how machine learning evolved over the decades starting with perceptrons and neural networks. However they had a limitation that those models cannot simalate new data from the learning and also could not scale.  Another main hindrance is the non availability of a large data set with labels to verify the output in real world.  This paved way to "Deep Belief Networks" which solved two significant problems seen in other methods, namely they could not infer information from the unobserved variables of dataset and could not generate new metadata from rich dataset. Unsupervised Machine Learning algorithms addressed these problems and Srikanth wonderfully linked this evolution history to the business application Diet Guru and explained how his team deployed RBM - an unsupervised machine learning algorithm. He stated how the next gen networks would infer, interpret and predict using such ML algorithms. Next came IBM's "Whatson" presentation - a ML based IoT platform with rich API's that can be customized to any domain like cancer research.



Image result for hololens autismWith more of DS & ML since morning, I thought of lending my ears towards Virtual Reality. Thanks for the decision for it turned out to be my best presentation in the conference. A team from Microsoft presented their Hololens product (similar to Google glass) and explained how its been used in treating person affected with autism spectrum. Besides the cause where its deployed, I was impressed by the way she presented the Hololens App platform, its simplicity and the need of apps to transform the lives of autism affected people. Though the underlying intention could be to sell their Hololens product and vouch on app developers, the manner in which it was put forth made the presentation stand out.

Following this was the workshop on "R & Python" in Predictive modelling. The handson session on R language given by Maheshwari from Cisco was very good, that I could grasp it instantly. Kudos to their efforts in sharing the preparatory slides beforehand which made me hit the ground running. Being already did Python and Scikit programing found the workshop to reinforce my learning. In crux, I understood R technically and also learnt that I need to open up, structure my thoughts and share the knowledge I have acquired in this domain.

Post Lunch sessions were from Flipkart and Amazon on how they are solving online retailer problems using ML. Some of them are identifying anomalous/rogue users, reviews & rating. It was interesting to know how Amazon benefited by using visual deep learning in delivering features like "Similar products/designs, recently bought ones" etc. The day however ended with an okayish session on "The art of secure coding". Perhaps after securing Cisco's Security Ninja belt, I might have found the session too naive.

The conference on third day had sessions on Career Mastery, Entrepreneurship, Management excellence, etc. I peeped into one of the sessions and found it to be all "known" stuff. These were the things which are perceived as "known" but hard to implement. I was looking for more practical ways to implement which I could not see them delivered on the sessions. So I decided to visit the company booths and demos, get to know their work and try my hand getting goodies. Below are the interesting updates from various companies.

1. GE Health care - It was exciting to know about Predix - The Industrial IoT platform which offers Data Science Analytics using cloud deployed machine learning algorithms. They claimed that this was already been deployed and used in big industries like Aerospace, Medical, Wind Power. Being inquisitive to know more, I had a 1-1 discussion and came to know how they developed Predix and their strategy for Go-To-Market. The platform is been deployed - Yes, however, it was deployed on the domains, where they own the assets. It is offered as Platform-as-a-Service (Paas) thereby incur recurring revenues. So they own both the risk and growth of the Predix, while slowly trying to acquire external customers. Very good strategy for innovation deployment indeed. (https://www.predix.io/)

2. Adobe - Adobe had a demo of their latest FREE products Photoshop Fix, Adobe Draw. Very easy to use and good to know to use it for creating my posters. (www.adobe.com/in/products/fix.html)

3. Alexa from Amazon - Alexa is a voice controlled IoT product from Amazon, something similar to Siri from Apple, Cortona from Microsoft. However, Alexa is a bit more intelligent that it could process complex commands from a person and act based on them. It can be integrated with other home appliances and also in industrial automation. Have to watch how this IoT product is evolving in the market. (https://developer.amazon.com/alexa)

4. Arm educational kit - Arm's educational wing demonstrated a ARM Cortex processor built by ST Micro which can be used as Robot. Their main aim is to reach to more university students by donating these products and give them their development environment to create products using ARM. I could infer that they want to get more students' mind share for ARM processor technical proficiency and also get their involvement in developing innovative products.

5. GE portable scanning machine - This is a path breaking device from GE health care which is a portable Ultrasound scanner device that can be taken to remote villages to perform abdominal scan, pregnancy scan tests etc. It would be a saviour to people living in remote areas. 

6. Microsoft bot - Got to know how Microsoft's bot framework operate and how easy it is to deploy bot for websites having social media plugins from various social media channels like Twitter, Skype etc. (https://dev.botframework.com/)

7. Cisco's Next Gen Network protection firewall - Last but not the least, paid a visit to Cisco's demo about next gen network protection. They demonstrated two wonderful demos about how our products safeguard the network threats. I was amazed to see Stealthwatch product (http://www.cisco.com/c/en/us/products/security/stealthwatch/index.html)  about the network anomaly detection using Machine Learning. It's exactly the similar idea which we presented in one of our Hackathon and later wanted to build on that. I wish I could spend more time on the Machine Learning technology on an innovative product that come under Cisco's market landscape.

Thanks to Cisco for giving me this opportunity to attend GHCI'2016
Fingers Crossed and Brains ticking - to present some cool ML product next time in GHCI?!