sf data mining_meetup
DESCRIPTION
These slides accompanied a demo of Deeplearning4j at the SF Data Mining Meetup hosted by Trulia. http://www.meetup.com/Data-Mining/events/212445872/ Deep-learning is useful in detecting identifying similarities to augment search and text analytics; predicting customer lifetime value and churn; and recognizing faces and voices. Deeplearning4j is an infinitely scalable deep-learning architecture suitable for Hadoop and other big-data structures. It includes a distributed deep-learning framework and a normal deep-learning framework; i.e. it runs on a single thread as well. Training takes place in the cluster, which means it can process massive amounts of data. Nets are trained in parallel via iterative reduce, and they are equally compatible with Java, Scala and Clojure. The distributed deep-learning framework is made for data input and neural net training at scale, and its output should be highly accurate predictive models. The framework's neural nets include restricted Boltzmann machines, deep-belief networks, deep autoencoders, convolutional nets and recursive neural tensor networks. Finally, Deeplearning4j integrates with GPUs. A stable version was released in October.TRANSCRIPT
![Page 1: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/1.jpg)
{Deep Learning
Machine Perception and Its Applications
Adam Gibson // deeplearning4j.org // skymind.io // zipfian
![Page 2: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/2.jpg)
Deep Learning = subset of Machine Learning
Machine Learning = subset of AI AI = Algorithms that repeatedly
optimize themselves. Deep learning = pattern recognition Machines classify data and improve
over time.
DL, a Subset of AI
![Page 3: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/3.jpg)
Why Is DL Hard?
We see this… Machines see this… (Where’s the cat?)
(Hat tip to Andrew Ng)
![Page 4: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/4.jpg)
Anything digitized Raw media: MP3’s, JPEG’s, text, video Sensor output: temperature, pressure,
motion and chemical composition Time-series data: Prices and their
movement; e.g. the stock market, real estate, weather and economic indicators
It’s setting new accuracy records everywhere
What Can It Handle?
![Page 5: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/5.jpg)
Recommendation engines: Anticipate what you will buy or click.
Anomaly detection: Bad outcomes signal themselves in advance: fraud in e-commerce; tumors in X-rays; loans likely to default.
Signal processing: Deep learning can estimate customer lifetime value, necessary inventory or an approaching market crash.
Facial and image recognition
What’s It Good For?
![Page 6: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/6.jpg)
Facial recognition & featurehierarchy
(Hat tip to Andrew Ng)
![Page 7: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/7.jpg)
DL4J Facial Reconstructions
![Page 8: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/8.jpg)
Nets need training data. You know what training sets contain. Nets learn training-set faces by
repeated reconstruction. Reconstruction = finding which facial
features are indicative of larger forms.
When a net can rebuild the training set, it is ready to work with unsupervised data.
How Did It Do That?
![Page 9: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/9.jpg)
Nets measure the difference between their results and a benchmark = loss function
They minimize differences with an optimization function.
They optimize by altering their parameters and testing how changes affect results.
Gradient descent, Conjugate gradient, L-BFGS
Technical Explanation
![Page 10: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/10.jpg)
Learning looks like this.
Note the local minima…
![Page 11: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/11.jpg)
Through pre-training, nets learn to locate signal in a world of noise
Generic priors initiate weights Reconstructions = representations Feature hierarchies intuition about
complex, abstract features
Representation Learning
![Page 12: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/12.jpg)
Facebook engages us more. (95-97% accuracy)
Government agencies identify persons of interest.
Video game makers build more realistic (and stickier) worlds.
Stores identify customers and track behavior, prevent churn and encourage spending.
Facial Recognition’s Uses
![Page 13: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/13.jpg)
Sentiment analysis ~ NLP Software classifies sentences by
emotional tone, bias and intensity Positive or negative - object-specific… Rank movies, books, consumer goods,
politicians, celebrities Predict social unrest, gauge
reputations, PR…
Sentiment Analysis & Text
![Page 14: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/14.jpg)
Restricted Boltzmann Machine (RBMs)
![Page 15: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/15.jpg)
A stack of RBMs. 1st RBM’s hidden layer -> 2nd RBM’s
input layer Feature hierarchy A DBN classifies data. Buckets images: e.g. sunset, elephant,
flower. Useful in search.
Deep-Belief Net (DBN)
![Page 16: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/16.jpg)
Two DBNs. The first DBN encodes data into
vector of 10-30 numbers. The second DBN decodes data back
to original state. Reduce any document/image to
highly compact vector. QA and information retrieval: Watson
Deep Autoencoder
![Page 17: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/17.jpg)
Some Results
![Page 18: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/18.jpg)
Image Search Results
![Page 19: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/19.jpg)
Good with images. ConvNets learn data like images in
patches. Each piece learned is then woven
together in the whole. Yann LeCun’s baby, now at
Facebook.
Convolutional Net
![Page 20: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/20.jpg)
Top-down, hierarchical nets rather than feed-forward like DBNs.
Sequence-based classification, windows of several events, entire scenes rather than images.
Features = vectors. A tensor = multi-dimensional matrix,
or multiple matrices of the same size.
Recursive Neural Tensor Net
![Page 21: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/21.jpg)
RNTNs & Scene Composition
![Page 22: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/22.jpg)
RNTNs & Sentence Parsing
![Page 23: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/23.jpg)
t-SNE for Data Visualizations
![Page 24: Sf data mining_meetup](https://reader038.vdocument.in/reader038/viewer/2022102815/557d60f2d8b42aba3d8b5048/html5/thumbnails/24.jpg)
DL4J + MNIST + t-SNE