The AAAI makes efforts to advance research in the area of artificial intelligence and provide avenues for collaboration. Major AAAI activities include organizing and sponsoring conferences, symposia and workshops; publishing a quarterly magazine for all members; honoring individuals who have made distinguished contributions to the field, publishing books, proceedings, and technical reports; compiling a host of online resources and publications; and awarding grants and scholarships.
FACE GENERATION USING GAN
We used generative adversarial networks to generate new images of faces by using Celeba dataset. Implemented discriminator neural network that discriminates on images.
Implemented generator to generate an image using the input noise. We used Deconvolution at generator and applied batch normalization for generating images. We used Convolution at discriminator for Classifying generated image.
GENERATE TV SCRIPTS
In this project we have generated our own Simpsons TV scripts using RNNs. Neural Network we build will generate a new TV script for a scene at Moe's Tavern. We applied embedding with 300 dimensions to input_data using TensorFlow. We used two LSTM layers with 256 cells to the embedding output using TensorFlow. We used sequence to sequence model for generating scripts at the final layer.
LANGUAGE TRANSLATION
We are going to take a peek into the realm of neural network machine translation. We trained a sequence to sequence model on a dataset of English and French sentences that can translate new sentences from English to French. Implemented Encoder RNN layer and Decoder RNN layer. We used 2 LSTM layers with 256 cell on 300 dimensions embedded input.
IMAGE CLASSIFICATION OF CIFAR-10 DATASET USING CNN
Classified images from the CIFAR-10 dataset. Used 3 convolutional layers, 2 fully connected layers and one final output layer. Achieved 70% accuracy on test set which is 7 times better than random guess (10%).
2012-2016
Developed an Android project for Indian Defense Research and Development Organization. Won the best project award for classifying tumor in brain using SVM classifier.
What I’ve Learned
January 2016 - Present
Implemented a Resume parsing application which parses Name, Email address, Phone number, Skills present in the given resume using Stanford NER. Added indexing tool like Apache SOLR on top of MYSQL for faster fetch operation and also scaled it across multiple instances. Implemented a Speech to Text converter on the real time data.
My Career
Word2vec, PCA, Random Forest, Regression and Classification concepts. Neural Networks which include Recurrent, Convolutional and Adversial. Spring MVC, Spring JPA, Spring boot, REST Web services, Node JS. React, Angular and jQuery. Gradle, maven, gulp, Webpack, Browserify,