Dear Apache MXNet (Incubating) community, Please find some meeting notes from the user group meeting on 28th August 2018.
# MXNet contribution Per joined the group to see and discuss how he can contribute to MXNet. He already had a PR opened [1] and was looking for the next steps. As a suggestion work on flaky tests was proposed. # Architecture review Hagen joined the group in pearson to review and discuss architecture that he and his company is using for categorizing uploaded by users documents. We had interesting discussion in this regards, some broader information on the topic came up. Here are some things worth checking out: * Best Practices for ML Engineering [2] * Powerful open source se2seq framework Sockeye [3] which is used for machine translation * Attention based methods, like Transformer [4], Attention Is All You Need [5]. Join us for the next sessions! Best Anton [1] https://github.com/apache/incubator-mxnet/pull/12331 [2] https://developers.google.com/machine-learning/guides/rules-of-ml/ [3] https://github.com/awslabs/sockeye [4] https://ai.googleblog.com/2017/08/transformer-novel-neural-network.html [5] https://arxiv.org/abs/1706.03762