|
Disclaimer:
These pages about different languages / apis / best practices were mostly jotted down quckily and rarely corrected afterwards. The languages / apis / best practices may have changed over time (e.g. the facebook api being a prime example), so what was documented as a good way to do something at the time might be outdated when you read it (some pages here are over 15 years old). Just as a reminder. Machine LearningK-means clusteringClustering introHow to find how many clusters (if you don know what you want) No easy way to get it. You might know in advance, e.g. cluster by digits 0-9 means you want 10 clusters. Could run algorithm for differnt values of K, K=2,3,... And compare variance (how many are in each cluster?), and pick the one with the smallest variance. Problem: is ideal when K=n (i.e. one in each cluster) In general, the more clusters you add, the lower the variance is going to be. Look at it visually in scree plot. Pick the K where the mountain ends and the rubble begins, i.e. less drastic changes: maximize 2nd derivate of V: point where rate of decline changes the most.Mean-shift Mean-shift, can be used to track objects between frames Neural networksGoogle Tech Talks - November, 29 2007How convolutional neural networks see the world Training and investigating Residual Nets Neural networks for computer vision The Back Propagation Algorithm A Beginner's Guide To Understanding Convolutional Neural Networks Trained image classification models for Keras Neural Network Architectures Toolsscikit-learn: Machine Learning in Pythonscikit tutorial PyHubs is a machine learning library developed in Python. It contains implementations of hubness-aware machine learning algorithms together with some useful tools for machine learning experiments. Curated list of machine learning frameworks, libraries and software TensorflowTensorFlowhttp://playground.tensorflow.org/ K-Means Clustering with TensorFlow Improvement on that K-means example Other k-means implmenetation Tensorflow sample code TensorFlow Implementation of Deep Convolutional Generative Adversarial Networks CS224D Lecture 7 - Introduction to TensorFlow (19th Apr 2016) https://github.com/aymericdamien/TensorFlow-Examples RNNS IN TENSORFLOW, A PRACTICAL GUIDE AND UNDOCUMENTED FEATURES Videos/LecturesCaltech machine learning 2012Microsoft Research: Deep Learning 7 December 2015 Deep Learning course by Google Neural Networks Demystified Lecture Lecture python numpy cs231n Andrej Karpathy 2016 UC Berkeley CS188 Intro to AI cs189 cs294-112 berkeley deeplearning t-SNEVisualizing data using t-SNEhttp://scikit-learn.org/stable/auto_examples/manifold/plot_lle_digits.html word2vec and similarImplementing Conceptual Search in Solr using LSA and Word2Vec: Presented by Simon Hughes, Dice.com (Oct 2015)http://multithreaded.stitchfix.com/blog/2015/03/11/word-is-worth-a-thousand-vectors/ http://eng.kifi.com/from-word2vec-to-doc2vec-an-approach-driven-by-chinese-restaurant-process/ https://github.com/dav/word2vec This is an implementation of the LexVec word embedding model (similar to word2vec and GloVe) that achieves state of the art results in multiple NLP tasks ClusteringFuzzy k-means in pythonFuzzifying clustering algorithms: The case study of MajorClust http://pythonhosted.org/scikit-fuzzy/auto_examples/plot_cmeans.html http://stackoverflow.com/questions/6736347/is-a-fuzzy-c-means-algorithm-available-for-python https://pypi.python.org/pypi/scikit-fuzzy Joint Unsupervised Learning of Deep Representations and Image Clusters (2016) https://github.com/jwyang/joint-unsupervised-learning Image classificationhttps://github.com/jcjohnson/densecapalternative to opencv, looks good, c++ Face detectionhttps://github.com/cmusatyalab/openfacehttps://bamos.github.io/2016/01/19/openface-0.2.0/ Where are they looking? NLPhttps://github.com/oxford-cs-deepnlp-2017/lecturescs224n NLP 2017 Python NLP package GeneralTensortalk, just links to AI stuff in a hackernews/reddit mannerUnderstanding Aesthetics with Deep Learning Netflix: extracting interesting regions/text, how they define simularity Starting points for deep learning and RNN https://github.com/kjw0612/awesome-rnn Approaching (Almost) Any Machine Learning Problem (2016) DeepWarp: Photorealistic Image Resynthesis for Gaze Manipulation (in ECCV'16) http://www.efimov-ml.com https://blog.insightdatascience.com/graph-based-machine-learning-6e2bd8926a0 http://news.efinancialcareers.com/uk-en/285249/machine-learning-and-big-data-j-p-morgan/ General AI booksSecond machine ageFooling networkshttp://www.evolvingai.org/foolingTheanoIntro to Deep Learning with Theano and OpenDeep by Markus Beissinger (2015)Pytorchhttp://blog.outcome.io/pytorch-quick-start-classifying-an-image/Nearest neigbhor packagesOverview of theory and pros/cons of some packageshttps://github.com/lyst/rpforest https://github.com/spotify/annoy http://www.cs.ubc.ca/research/flann/ https://github.com/primetang/pyflann https://github.com/mariusmuja/flann https://github.com/falconn-lib/falconn https://github.com/yahoo/lopq http://www.kgraph.org/index.php?n=Main.Home https://github.com/hdidx/hdidx http://mmp2.github.io/megaman/ https://arxiv.org/pdf/1603.02763.pdf https://github.com/kayzhu/LSHash http://ryanrhymes.github.io/panns/ Installing cunn: luarocks install cunn libcunnx.so malformed object http://stackoverflow.com/questions/26822010/install-name-tool-malformed-object-load-command-23-cmdsize-is-zero-mac-os-x th -e "require 'cutorch'; print(cutorch)" luajit -l libcutorch problem loading torch/install/lib/lua/5.1/libcutorch.so: https://github.com/torch/cutorch/issues/243 http://stackoverflow.com/questions/36312018/unable-to-import-require-cutorch-in-torch http://ubuntuforums.org/showthread.php?t=2264359 https://github.com/torch/cutorch/issues/244 https://github.com/torch/cutorch/issues/126 KerasCheck if using gpufrom tensorflow.python.client import device_lib print(device_lib.list_local_devices())https://tryolabs.com/blog/2013/03/25/why-accuracy-alone-bad-measure-classification-tasks-and-what-we-can-do-about-it/ accuracy = "how high percentage of data set did it get right" precision = "how relevant was the result returned, i.e. were there false positives" recall = "how high percentage of the relevant result did it return, ignore any false positives" Accuracy = (true positives + true negatives) / (total examples) Precision = (true positives) / (true positives + false positives) Recall = (true positives) / (true positives + false negatives) F1 score = (2 * precision * recall) / (precision + recall) Misclassification Rate = (FP + FN) / total (microsoft definition) TPR = TP / (TP + FP) FPR = FN / (TN + FN) For multi classification: Accuracy = (TP + TN) / (TP + TN + FP + FN) where TN is defined as "all items that should have been classified in all other classes MINUS False Positives") Precision (micro) = sum(all TP) / (sum(all TP) + sum(all TN)) Precision (macro) = sum(precision for each class) / numclasses F1 micro/macro in the same way ---------------------------------------------------------------------- OS X cd More programming related pages |
|