Slide 1
Slide 2
Developer
Slide 3
Machine Learning is going viral...
Slide 4
Slide 5
Slide 6
❝Using #DeepLearning when all you needed was a few if statements. #MachineLearning #DataScience❞ —https://twitter.com/randal_olson/status/927157485240311808
Slide 7
Slide 8
Slide 9
Slide 10
Agenda Machine Learning Domain Dataset
Slide 11
Machine Learning
Slide 12
Artificial Intelligence Machine Learning Deep Learning !
Slide 13
https://blogs.nvidia.com/blog/2016/07/29/whats-difference-artificial-intelligence-machine-learning-deep-learning-ai/
Slide 14
General AI Human characteristics
Slide 15
AI Winter
Slide 16
Narrow AI Specific tasks
Slide 17
Slide 18
Facebook alt="Image may contain: ocean, sky, bridge, cloud, outdoor, water and nature"
Slide 19
https://www.facebook.com/ MaxIsDrawing/photos/a. 182284588581885/912506332 226370/?type=3&permPage=1
Slide 20
Slide 21
Slide 22
❝I made an AI watch 105 Tour de France races along with viewer numbers to make it simulate the most interesting Tour de France race possible. This is what it made. #ai #ml #machinelearning #artificialintelligence #wat❞ —https://twitter.com/thetafferboy/status/1039519923150630912
Slide 23
Slide 24
PS: A lot of Chatbots are not AI
Slide 25
Slide 26
Slide 27
❝Having been introduced to the wonders of Visual Chatbot by @JanelleCShane's AI experiments, I decided to see what it would think of two-headed snakes. As it turns out: mostly bananas. shutdowntheAI —https://twitter.com/AmeliaRMellor/status/1031006460598149121
Slide 28
Slide 29
Slide 30
Slide 31
Slide 32
❝Alice: I love stateless protocols! Bob: There has to be something bad about them. Alice: Bad about what?❞ —https://twitter.com/znjp/status/933405548678021120
Slide 33
Machine Learning Algorithms parse data → learn from it → make a determination or prediction "Trained" machine
Slide 34
❝Learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.❞
Slide 35
❝"Machine Learning is an emerging tech!" Logistic regression 1958 Hidden Markov Model 1960 Support Vector Machine 1963 k-nearest neighbors 1967 Artificial Neural Networks 1975 Expectation Maximization 1977 Decision tree 1986 Q-learning 1989 Random forest 1995❞ —https://twitter.com/farbodsaraf/status/977916871000412160
Slide 36
https://twitter.com/ algorithmia/status/ 1009486664933052416
Slide 37
❝But saying "powered by AI" is like saying you’re "powered by the internet" or "powered by computer code". By itself, it means nothing.❞ —https://twitter.com/jensenharris/status/999119292086960128
Slide 38
Learning Regression Ranking Clustering
Slide 39
https://twitter.com/ShawnWildermuth/status/932724124237123584
Slide 40
Slide 41
For children and machines Watch your language
Slide 42
Statistics 101: Linear Regression
Slide 43
❝We are leveraging machine learning.❞
Slide 44
https://twitter.com/LesGuessing/status/997146590442799105
Slide 45
Supervised Learning Input features and output labels are defined
Slide 46
Unsupervised Learning Unlabeled dataset Discover hidden relationships
Slide 47
https://xkcd.com/882/
Slide 48
Slide 49
Slide 50
Reinforcement Learning Feedback loop to optimize some parameter
Slide 51
Deep Learning Neural network producing a probability vector Lots of training and parallelization
Slide 52
https://www.youtube.com/watch?v=bxe2T-V8XRs
Slide 53
Access to a unique data set is inherently valuable
Slide 54
Slide 55
❝"What's the difference between AI and ML?" "It's AI when you're raising money, it's ML when you're trying to hire people."❞ —https://twitter.com/WAWilsonIV/status/925599712849174528
Slide 56
Domain
Slide 57
Patterns Trend (stationary) Cyclical Seasonal Irregular
Slide 58
Anomaly Point Anomalies Contextual Anomalies Collective Anomalies
Slide 59
Breakouts Mean Shift Ramp Up
Slide 60
Anomaly Detection with Machine Learning Supervised Learning Unsupervised Learning
Slide 61
Examples IT operations: Spiking 500s Security analytics: Unusual DNS activity Business analytics: Rare log message
Slide 62
Visual Inspection Complex, fast moving data Humans not made to stare at graphs Easy to miss
Slide 63
Where is the Anomaly?
Slide 64
Static Rules Definition False positives & negatives Tuning and adjustment
Slide 65
Which threshold?
Slide 66
Machine learning
Slide 67
! ❝OH: "Do you run any CPU intensive application on your laptop? Like, machine learning, or Slack?" ❞ —https://twitter.com/jpetazzo/status/932464823530430464
Slide 68
Frameworks TensorFlow Keras SciKit ...
Slide 69
How to build ML pipelines? ETL Data storage Optimization algorithms
Slide 70
❝I see you expected clean data. That's cute.❞
Slide 71
Slide 72
Model Baseline: What is normal?
Slide 73
Slide 74
Slide 75
Unsupervised
Slide 76
Evolves "Online" model learns continuously and ages out data
Slide 77
Single Time Series Example: Unusual traffic?
Slide 78
Multiple Time Series Multiple metrics or single metric split up Each series modeled independently Example: Unusual activity by country?
Slide 79
Slide 80
Dataset
Slide 81
nginx access log { "source": "/home/ec2-user/data/production-4/prod4elasticlog/_logs/access-logs541.log", "beat": { "hostname": "ip-172-31-5-206", "name": "ip-172-31-5-206", "version": "5.4.0" }, "@timestamp": "2017-03-08T11:44:51.562Z", "read_timestamp": "2017-06-20T08:49:58.538Z", "fileset": { "name": "access", "module": "nginx" },
Slide 82
"nginx": { "access": { "body_sent": { "bytes": "3262" }, "url": "/assets/blt1afcb054f02e257c/logo-activision.svg", "geoip": { "continent_name": "Asia", "country_iso_code": "IN", "location": { "lat": 20, "lon": 77 } },
Slide 83
"response_code": "200", "user_agent": { "device": "Other", "os_name": "Other", "os": "Other", "name": "Other" }, "http_version": "1.1", "method": "GET", "remote_ip": "192.19.197.26" } }, "prospector": { "type": "log" } }
Slide 84
Slide 85
Slide 86
Slide 87
Slide 88
Slide 89
Slide 90
Most of the internet went down
Slide 91
PS: When everything is on , nobody cares about your downloads
Slide 92
Slide 93
Counterfactual Reasoning Which host / IP / ... is involved in the anomaly
Slide 94
Slide 95
Slide 96
Slide 97
Combine Multiple Models
Slide 98
Slide 99
Correlation ≠ causation
Slide 100
https://xkcd.com/552/
Slide 101
https://xkcd.com/925/
Slide 102
Common problems Correlated features will mess up any model
Slide 103
Common problems Throw out most features if they are just noise
Slide 104
More features Future predictions
Slide 105
Slide 106
Conclusion
Slide 107
Agenda Machine Learning Domain Dataset
Slide 108
Rules of Machine Learning: Best Practices for ML Engineering http://martin.zinkevich.org/rules_of_ml/ rules_of_ml.pdf
Slide 109
43 rules Rule #1: Don’t be afraid to launch a product without machine learning Rule #14: Starting with an interpretable model makes debugging easier Rule #16: Plan to launch and iterate
Slide 110
Machine Learning ohne Hype Philipp Krenn @xeraa
