Tomorrow Tue 4/3 talk on "Deep Learning & its application in Medicine" at 456 W. Olive Avenue, Sunnyvale. This is a free event & no pre-registration is required. Walk-ins are welcome. Come for talk, stay to network @machinelearnbot @MachineLearnDC @DeepLearningHub@deeplearning4j
Postdoc position in machine learning. Especially looking for candidates with experience in memory networks/neural turing machine/synaptic plasticity/neuroevolution/deep RL. Please RT/forward to interested candidates. Thanks! candidate.hr-manager.net/App…
A good example of a paper that pulls together quite a few useful techniques is: openreview.net/forum?id=S1NH… . CIFAR 10 SoTA, which is an interesting dataset IMO since it is quite small, so needs good augmentation, regularization, architecture, etc
[1801.01586] A practical tutorial on autoencoders for nonlinear feature fusion: Taxonomy, models, software and guidelines arxiv.org/abs/1801.01586
This looks like a very useful survey/tutorial paper.
The ML research community has long been driven by the need to publish, which results in a stark, sometimes ridiculous bias towards complexity. Remember to ask: "can we do this with k-means and logistic regression?"
One thing I realize people don't like: Being told to think critically about a problem.People: AI algorithms are not magic. When you're first learning them, take some time to consider the knobs you're turning. It's amazing what applying the scientific method (step by step ) does.
New: My research w/@andyguess/@JasonReifler providing 1st behavioral estimates of fake news exposure in 2016 dartmouth.edu/~nyhan/fake-ne…
Key findings:
-heavily concentrated among w/most conservative info diets
-Facebook key vector of exposure
-fact-checks did not reach those exposed
My publisher is doing a deal (today only) where you can get half off my book "Deep Learning with Python", as well as its R version, "Deep Learning with R" (co-authored with JJ Allaire) manning.com/dotd?a_aid=keras
A reinforcement learning agent that learns to program new neural network architectures.
Same/better results as LSTMs but with funky nonlinearities (sine, SeLus, etc) and new connections that result in different activation patterns😯
einstein.ai/research/domain-…arxiv.org/abs/1712.07316