Collaboration in the realm of artificial intelligence leads to some strange bedfellows in the tech world. That's certainly the case today, when Microsoft and Amazon Web Services announced their collaboration on Gluon, a new programming library for machine learning. It works by providing a consistent interface for creating machine learning models using a variety of pre-built and highly optimized components. Gluon, which is available as an open source project, will provide a shared set of building blocks that people can use with both Amazon and Microsoft's preferred machine learning frameworks. Gluon's set of pre-built components are supposed to make it easier for developers to get started building models, and make it faster for machine learning experts to build prototypes of more complex systems that they might want to create by hand, according to a blog post from Matt Wood, AWS's general manager of AI.
Since the machine learning techniques are improving rapidly, it has been shown that the image recognition techniques in deep neural networks can be used to detect jet substructure. And it turns out that deep neural networks can match or outperform traditional approach of expert features. However, there are disadvantages such as sparseness of jet images. Based on the natural tree-like structure of jet sequential clustering, the recursive neural networks (RecNNs), which embed jet clustering history recursively as in natural language processing, have a better behavior when confronted with these problems. We thus try to explore the performance of RecNNs in quark/gluon discrimination. The results show that RecNNs work better than the baseline boosted decision tree (BDT) by a few percent in gluon rejection rate. However, extra implementation of particle flow identification only increases the performance slightly. We also experimented on some relevant aspects which might influence the performance of the networks. It shows that even taking only particle flow identification as input feature without any extra information on momentum or angular position is already giving a fairly good result, which indicates that the most of the information for quark/gluon discrimination is already included in the tree-structure itself. As a bonus, a rough up/down quark jets discrimination is also explored.
Deep learning systems have long been tough to work with, due to all the fine-tuning and knob-twiddling needed to get good results from them. Gluon is a joint effort by Microsoft and Amazon Web Services do reduce all that fiddling effort. Gluon works with the Apache MXNet and Microsoft's Cognitive Toolkit frameworks to optimize deep-learning network training on those systems. The problem with steps 1 and 2 is that they're tedious and inflexible. Hard-coding a network is slow, and altering that coding to improve the network's behavior is also slow.
GluonCV v0.6.0 added more video classification models, added pose estimation models that are suitable for mobile inference, added quantized models for video classification and pose estimation, and we also included multiple usability and code improvements. We now provide state-of-the-art video classification networks, such as I3D, I3D-Nonlocal and SlowFast. We have a complete model zoo over several widely adopted video datasets. We provide a general video dataloader (which can handle both frame format and raw video format). Users can do training, fine-tuning, prediction and feature extraction without writing complicate code.