NovaLabs_ML_Meetup_Notes_Jan5_2017

Notes from Machine Learning - Jan 5, 2017

- meeting is in Conference A
- learn the concepts, but then the libraries are going to make it just a function call


arxiv.org/pdf/1602.07261v2.pdf
updated Inception v4 results - 3.08% error rate


relu gives faster convergance than tanh or sigmoid
	autoML as a library that will automatically try to tweak the activation function

"Connectionism" neural net is the only one covered here

Yann LeCun
	- AI Lead at Facebook
	- doing summarization of images

Jeff Dean
	- at Google, distinguished engineer

Geoffrey Hinten
	- academic, lots of online videos

Andrew Ng
	- stanford university
	- CS229

Pete Shor
	- MIT professor of mathematics
	- in quantum mathematics sphere
	http://www-math.mit.edu/~shor/
	interesting quantum cryptography paper at http://www-math.mit.edu/~shor/NewDirections/
	

Demo of the TensorFlow playground
http://playground.tensorflow.org/#activation=relu&regularization=L1&batchSize=10&dataset=spiral&regDataset=reg-plane&learningRate=0.01&regularizationRate=0&noise=0&networkShape=4,5,4,4,4,4&seed=0.54787&showTestData=false&discretize=false&percTrainData=50&x=true&y=true&xTimesY=true&xSquared=true&ySquared=true&cosX=false&sinX=true&cosY=false&sinY=true&collectStats=false&problem=classification&initZero=false&hideText=false


http://yann.lecun.com/exdb/publis/pdf/lecun-98b.pdf

Discussion of L1, L2, and dropout
https://www.cs.toronto.edu/~hinton/absps/JMLRdropout.pdf


WordNet
http://wordnet.princeton.edu/wordnet/documentation/
http://wordnetweb.princeton.edu/perl/webwn?o2=&o0=1&o8=1&o1=1&o7=&o5=&o9=&o6=&o3=&o4=&s=citing&i=0&h=0000000#c
http://wordnet.princeton.edu/wordnet/download/current-version/



intro to neural
http://www.webpages.ttu.edu/dleverin/neural_network/neural_networks.html


discussion of Convolutional Neural Networks
	Andrej Karpathy - computer vision winter 2016
	https://www.quora.com/How-do-convolutional-neural-networks-work
	https://brohrer.github.io/how_convolutional_neural_networks_work.html


http://cs231n.github.io/convolutional-networks/

http://cs231n.github.io/assets/cnn/convnet.jpeg
http://cs231n.github.io/convolutional-networks/#conv
	- discussion of how the convolution aggregations can help 'tease' out the features


more on the recognition of images
http://yann.lecun.com/exdb/lenet/