Research/Blog
Meeting Minutes from AI Lab session on Saturday 12th Oct in Bengaluru
- October 16, 2019
- Posted by: vsinghal
- Category: Computer Vision Deep Learning Healthcare Machine Learning Reinforcement Learning
#CellStratAILab #disrupt4.0 #WeCreateAISuperstars
Last Saturday saw some amazing sessions on advanced AI at the CellStrat AI Lab meetup.
![](http://www.cellstrat.com/wp-content/uploads/2019/10/Collage-1-1024x768.jpg)
Diabetes prediction with Machine Learning :-
First came a superb presentation by Dr Purnendu Das on diabetes prediction using ML. Dr Das started by discussing the data sources for healthcare and clinical data analysis. He covered the components of the Electronic Health Record or EHR. Later he provided overview of Diabetes Mellitus Type 2 (T2DM) and data science use cases around it. Then he discussed Machine learning approach on clinical data to predict patient level risk factor for Diabetes and other chronic disease conditions.
Dr Das mentioned that there are numerous applications of ML to diabetes situation, such as predicting progression of the disease, predicting future glucose fluctuations, impact of HbA1c measurement on predicting hospital readmission rates, predicting diabetic retinopathy etc.
Finally, Dr Das demonstrated a fantastic code demo on predicting readmission rates in hospital due to various demographic indications and progress of diabetes mellitus disease. It involves feature imputation, feature selection, model development and cohort scoring to decide on interventions. The dataset used was 100,000 T2DM patients from 30 hospitals which is available in CERNER HEALTH FACTS dataset. The model used various algos like DT, LogReg, SVC, RF, GBC etc.
Neural Style Transfer :-
Then came an excellent presentation on Neural Style Transfer by Gouthaman Asokan. Neural Style Transfer refers to a class of software algorithms that manipulate digital images, or videos, to adopt the appearance or visual style of another image. In this particular approach, it is seen that it is possible to separate the style representation and content representations in a CNN, learnt during a computer vision task. It is possible to blend content and style image together such that the input image is transformed to look like the content image, but “painted” in the style of the style image.
This is basically achieved by extracted intermediate layer features from an image processing network like VGG16 and style features from another network. The model trains via combined loss minimization of content image distance with respect to output image, along with style image distance from output image. The result of this cumulative loss minimization is that the final output image inculcates content features from the former image and style features from the latter image.
![](http://www.cellstrat.com/wp-content/uploads/2019/10/Comic-Style-Transfer-2-1024x563.png)
Deep Q Network (DQN) in RL :-
Then came a fabulous presentation by Niraj Kale on DQNs. Q Learning is an RL algorithm wherein one updates the Q values (indicating reward potential of various state-action pairs) table in an iterative process in an environment where multiple states, actions, reward potential concepts are involved (such as a game of Chess or a game of GO).
Deep Q network is extension of Q learning using the deep neural networks. In this example, Niraj showed how to train a model to play the Doom game using the DQN. The input frames are stacked to resolve temporal limitation problem. These stacked frames are fed to the 3 layers of Convolutional network followed by fully connected dense network which will be trained to take action in the game. Also, some improvements in DQN, ie. double DQN, deuling DQN, Experience replay buffer, were discussed.
![](http://www.cellstrat.com/wp-content/uploads/2019/10/Q-Value-update.png)
Image Credit : Dr Tom Mitchell
Wish to experience AI revolution unfolding right here in BLR ? If yes, then attend our AI Lab meetup this Saturday (19th Oct 2019) to participate in an intense Hands-On Workshop on QnA with NLP solution. Register for either of the two BLR locations below :-
Bellandur AI Lab (Saturday 19th Oct) :-
Register : https://www.meetup.com/Disrupt-4-0/events/vcqljryznbzb/
Topic : Hands-On Workshop on NLP – Ques-Ans System
Loc. : WeWork, Embassy Tech Village, ORR, BLR
Presenter : Abdul Azeez
Infantry Rd AI Lab (Saturday 19th Oct) :-
Register : https://www.meetup.com/Disrupt-4-0/events/264515296/
Topic : Hands-On Workshop on NLP – Ques-Ans System
Loc. : WeWork Prestige Central, Infantry Rd, BLR
Presenter : Indrajit Singh
See you this weekend for the AI Lab meetup ! Let’s disrupt with AI !
Questions ? Call me at +91-9742800566 !
Best Regards,
Vivek Singhal
Co-Founder & Chief Data Scientist, CellStrat
+91-9742800566