Faculty Publications
Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736
Publications by NITK Faculty
Browse
2 results
Search Results
Item Unsupervised Abstractive Text Summarization with Length Controlled Autoencoder(Institute of Electrical and Electronics Engineers Inc., 2022) Dugar, A.; Singh, G.; Balamuralidhar, B.; Anand Kumar, A.M.This work deals with taking an unsupervised approach to abstractive text summarization where a large set of sentences is converted into a concise summary highlighting the essential details. This is achieved with the use of an adversarial autoencoder model. The model encodes the input to a smaller latent vector and the decoder decodes this latent code to generate the higher dimensional output with some loss. Unlike variational autoencoders, AAE's use discriminators to learn using adversarial loss. K-Means clustering and language models are used to get the final summary. This model has been tested with different datasets like the Amazon, Rotten Tomatoes and Yelp reviews dataset to essentially do an opinion summarization task and this is finally evaluated using ROGUE-1, ROGUE-2,ROGUE-L and BLEU scores. The same task is also conducted on a dataset in Hindi. We obtain a ROGUE-1 score of around 24% for Amazon, Yelp and CNN/Daily Mail dataset and a score of 12% for Rotten Tomatoes while the score obtained for the Hindi news articles dataset is only 8%. © 2022 IEEE.Item Optimizing Machine Learning Operators and Models for Specific Hardware Using Apache-TVM(Institute of Electrical and Electronics Engineers Inc., 2023) Madathil, K.T.; Dugar, A.; Patil, N.; Unnikrishnan, U.Diligent utilization of hardware resources when dealing with computationally intensive jobs like machine learning (ML) that have a huge scope of compiler optimizations are often neglected due to the complexity of its implementation. The main reasons for its complexity is the wide range of architectures and the difference between the development and deployment environments. This leads to poor utilization of resources such as memory, hardware and increased execution time. These problems can be tackled using Apache-TVM - a compiler specifically designed to tune and optimize machine-learning models for specific hardware. We have implemented matrix multiplication on two types of hardware, x86 and Hexagon Digital Signal Processor (DSP), and have optimized it for specific hardware. Apache-TVM also supports tuning of whole ML models by applying various graph-level and operator-level optimizations. TVM can also automate the optimization of low-level programs to specific hardware characteristics using autoTVM which is a cost-based model for exploration of the search space for code optimization. We have obtained a significant reduction of upto 32.32% for Emotion FerPlus model and more than 150 times for matrix multiplication on hexagon DSP in execution time without reducing the accuracy or the performance. © 2023 IEEE.