Meeting Minutes from AI Lab session on Saturday 31st Aug in Bengaluru
#CellStratAILab #disrupt4.0 #WeCreateAISuperstars
Last Saturday, we had AI Lab meetups at Bengaluru (Bellandur and Hebbal locations). Deep AI sessions were presented in both locations.
In Bellandur, we saw a great session on Singular Value Decomposition (SVD) by Sai Sundarakrishna. SVD is a matrix factorization technique to discover the constituents of a matrix. As opposed to Eigen Decomposition, each matrix can be factorized with SVD. SVD is used in denoising, compression or data reduction applications.
Then came a superb session by Niraj Kale on DeepSpeech. DeepSpeech is the speech recognition engine developed by Chinese Search Engine firm Baidu. It uses 7 BiRNN (bi-directional RNNs) with 2 layers of Conv2D for feature extraction. The loss calculation is done using CTC (connectionist temporal classification). Niraj discussed improvements and possible application of DeepSpeech for Indian languages.
Then Satyanarayana R. presented an intuitive session on RNNs. He started with explaining the architecture of RNN and its forward propagation and backpropagation. Satya also covered various types of RNNs. Later he presented character language model, along with the attendant exploding and vanishing gradient issues. Later he discussed the LSTM, GRUs, BiRNNs and Deep RNNs, all advanced variations of RNNs.
In Hebbal AI Lab, we had Pushparaj M. present an excellent session on 3D CNNs. 3D CNN maps the convolution kernel in a 3D space rather than the 2D space. It can also be used for video analytics with time (temporal) dimension as the 3rd dimension added to subsequent image frames of the video. In an application of 3D CNN for image segmentation models, Pushparaj showed 3D brain tumor segmentation with UNET model implemented in MATLAB. Here a 3D UNET uses 3D Conv layers to encode and decode the image in order to segment it.
Finally, Prameela Prakash presented a deep-dive session on time series and its various aspects such as trend, seasonality, stationarity etc. Prameela explained various techniques for time-series forecasting such as de-seasonalization and re-seasonalization, detecting stationarity in a time-series, white noise concept etc.
I invite you to be a part of the AI revolution. Attend our AI Lab this weekend in Bengaluru or Delhi to learn more :-
Bengaluru AI Lab (Saturday 7th Sep) :-
Register : https://www.meetup.com/Disrupt-4-0/events/263029888/
Topic : Speech2Face, Pose Estimation
Loc. : WeWork, Embassy Tech Village, ORR, BLR
Gurugram AI Lab (Saturday 7th Sep) :-
Register : https://www.meetup.com/Disrupt-4-0/events/263642316/
Topic : Gradient Descent, Image Classification and Segmentation, OpenCV
Loc. : WeWork Forum, DLF Cybercity, Phase III, Gurugram
Noida AI Lab (Sunday 8th Sep) :-
Register : https://www.meetup.com/Disrupt-4-0/events/263741000/
Topic : Decision Trees, RNNs, Computer Vision
Loc. : 91Springboard, Sector-2, Noida
See you this weekend for the AI Lab meetup! Lets disrupt the world with AI, together !
Questions ? Call me at +91-9742800566 !
Co-Founder & Chief Data Scientist, CellStrat