The most popular and comprehensive Open Source ECM platform
Liquid Machine Learning: Resilience When Encountering the Unexpected
Machine Learning (ML) algorithms ingest massive amounts of data and are able to identify and pick out patterns that recur in the data. Once trained, when the algorithm is able to identify similar patterns when presented with a new data set and make decisions about it.
Researchers at MIT are trying to make ML algorithms more dynamic and allow for continuous learning. The algorithm is, in effect, in a state of perpetual training. The technique that they’ve created is called Liquid Machine Learning. The algorithms constantly is tuning the parameters to respond to and adapt to new data inputs. This continual learning makes the algorithm significantly more flexible that standard machine learning techniques.
Ramin Hasani, MIT CSAIL researcher, said that “the real world is all about sequences, even our perception – you’re not perceiving images, you’re perceiving sequences of images. So time-series data actually create our reality… This is a way forward for the future of robot control, natural language processing, video processing — any form of time series data processing. The potential is really significant.”
The flexibility of the new method leads to greater resilience when presented with data that may be different or unexpected from what it has seen previously.
Hasani said that “the model itself is richer in terms of expressivity. We have a probably more expressive neural network that is inspired by nature. But this is just the beginning of the process. The obvious question is how do you extend this? We think this kind of network could be a key element of future intelligence systems”.