In the Machine Learning Technologies group, we believe machine learning lies at the heart of creating cognitive systems. Being a part of IBM Research AI, our team works on cutting-edge research in the field of deep learning to tackle questions such as: What does it take to transform human natural language to forms that computers can process? How can computers perform cognitive tasks using their ability to move back and forth between natural language and its internal learned representation? What does it mean for a computer to learn human language? How can computers leverage external knowledge to help them learn faster and more accurately?
Deep learning is highly successful nowadays thanks to the vast amount of available data, increase in computational power, and new optimization algorithms. When tackling specific tasks or challenges, however, a network architecture that is suitable to the task plays a critical role in its performance. Our team creates novel network architectures for natural language and text analytics tasks.
We work on applying algorithms in various fields, using large-scale computing infrastructure. Our expertise ranges from traditional supervised and unsupervised machine learning to novel deep learning architectures, recommender systems, and representation learning. We are creating, developing, and training deep neural networks with an emphasis on sequence analysis using recurrent neural networks of all sorts, mainly for natural language tasks.
Our work involves close collaboration with IBM teams around the globe, academic institutions, and partners from industry. We believe the cognitive computing abilities that will be on the shelf tomorrow, are being researched and developed here today.