Browse Poster Wiki: 2022 Poster Award Winners, Awards Ceremony (Watch Recording from minute marker 1630), Posters by 2022 HPCC Systems Interns, Posters by Academic Partners, Poster Judges, About Virtual Judging, 2022 Poster Contest Home Page, Poster Contest Previous Years
Rahul Anbalagan is a 2nd Year Student at the RV College of Engineering, Bengaluru, India.
HPCC Systems, an open-source cluster computing platform for big data analytics consists of a Generalized Neural Network bundle with a wide variety of features which can be used for various neural network applications. To enhance the functionality of the bundle, this abstract proposes the design and development of a Transformer neural network on HPCC Systems platform using ECL, a declarative language on which HPCC Systems works.
A transformer is a state-of-the-art deep learning model that adopts the mechanism of self-attention, differentially weighting the significance of each part of the input data. It is used primarily in the fields of natural language processing and computer vision.
The Transformer architecture follows an encoder-decoder structure, but does not rely on recurrence and convolutions like RNN and LSTM. The task of the encoder is to map an input sequence of continuous representations, i.e., the encoder layer tries to capture the relations between the words in the sentence, which is then fed into a decoder. The decoder receives the output of the encoder together with the decoder output at the previous time step to generate an output sequence.
This project’s major goal is to develop and improve the existing Generalized Neural Network bundle provided by HPCC systems. This would enable to break the computation speed of already fast computation on HPCC with faster natural language processing as Transformers introduce parallelism whereas typical RNN and LSTM are sequential.
Advantages of Transformer model:
1. The attention mechanism provides context for any position in the input sequence.
2. Introduces parallel processing.
3. Can process entire input all at once.
All these advantages would self-explain why transformers models have taken over LSTM and RNN in the current world and would accelerate big data computation to much higher speed and standard.
In this Video Recording, Rahul provides a tour and explanation of his poster content.
Click on the poster for a larger image.