Machine Learning ohne Hype
A presentation at IT Tage in in Frankfurt, Germany by Philipp Krenn
Developer
Machine Learning is going viral...
❝Using #DeepLearning when all you needed was a few if statements. #MachineLearning #DataScience❞ —https://twitter.com/randal_olson/status/927157485240311808
Agenda Machine Learning Domain Dataset
Machine Learning
Artificial Intelligence Machine Learning Deep Learning !
https://blogs.nvidia.com/blog/2016/07/29/whats-difference-artificial-intelligence-machine-learning-deep-learning-ai/
General AI Human characteristics
AI Winter
Narrow AI Specific tasks
Facebook alt="Image may contain: ocean, sky, bridge, cloud, outdoor, water and nature"
https://www.facebook.com/ MaxIsDrawing/photos/a. 182284588581885/912506332 226370/?type=3&permPage=1
❝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
PS: A lot of Chatbots are not AI
❝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
❝Alice: I love stateless protocols! Bob: There has to be something bad about them. Alice: Bad about what?❞ —https://twitter.com/znjp/status/933405548678021120
Machine Learning Algorithms parse data → learn from it → make a determination or prediction "Trained" machine
❝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.❞
❝"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
https://twitter.com/ algorithmia/status/ 1009486664933052416
❝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
Learning Regression Ranking Clustering
https://twitter.com/ShawnWildermuth/status/932724124237123584
For children and machines Watch your language
Statistics 101: Linear Regression
❝We are leveraging machine learning.❞
https://twitter.com/LesGuessing/status/997146590442799105
Supervised Learning Input features and output labels are defined
Unsupervised Learning Unlabeled dataset Discover hidden relationships
https://xkcd.com/882/
Reinforcement Learning Feedback loop to optimize some parameter
Deep Learning Neural network producing a probability vector Lots of training and parallelization
https://www.youtube.com/watch?v=bxe2T-V8XRs
Access to a unique data set is inherently valuable
❝"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
Domain
Patterns Trend (stationary) Cyclical Seasonal Irregular
Anomaly Point Anomalies Contextual Anomalies Collective Anomalies
Breakouts Mean Shift Ramp Up
Anomaly Detection with Machine Learning Supervised Learning Unsupervised Learning
Examples IT operations: Spiking 500s Security analytics: Unusual DNS activity Business analytics: Rare log message
Visual Inspection Complex, fast moving data Humans not made to stare at graphs Easy to miss
Where is the Anomaly?
Static Rules Definition False positives & negatives Tuning and adjustment
Which threshold?
Machine learning
! ❝OH: "Do you run any CPU intensive application on your laptop? Like, machine learning, or Slack?" ❞ —https://twitter.com/jpetazzo/status/932464823530430464
Frameworks TensorFlow Keras SciKit ...
How to build ML pipelines? ETL Data storage Optimization algorithms
❝I see you expected clean data. That's cute.❞
Model Baseline: What is normal?
Unsupervised
Evolves "Online" model learns continuously and ages out data
Single Time Series Example: Unusual traffic?
Multiple Time Series Multiple metrics or single metric split up Each series modeled independently Example: Unusual activity by country?
Dataset
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" },
"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 } },
"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" } }
Most of the internet went down
PS: When everything is on , nobody cares about your downloads
Counterfactual Reasoning Which host / IP / ... is involved in the anomaly
Combine Multiple Models
Correlation ≠ causation
https://xkcd.com/552/
https://xkcd.com/925/
Common problems Correlated features will mess up any model
Common problems Throw out most features if they are just noise
More features Future predictions
Conclusion
Agenda Machine Learning Domain Dataset
Rules of Machine Learning: Best Practices for ML Engineering http://martin.zinkevich.org/rules_of_ml/ rules_of_ml.pdf
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
Machine Learning ohne Hype Philipp Krenn @xeraa
Machine Learning ist aktuell sowohl ein "heißer" wie auch mehrdeutiger Begriff.
Dieser Vortrag behandelt einen konkreten Aspekt davon – die Anomalieerkennung in Zeitserien. Das klingt zuerst sehr spezifisch, ist aber für den Betrieb und die Sicherheit von IT-Systemen konkret anwendbar: Was ist das erwartete Verhalten und wie schwerwiegend ist eine Abweichung davon. Dabei betrachten wir folgende Aspekte:
- Was steckt hinter Artificial Intelligence, Machine Learning und Deep Learning generell?
- Wann sind regelbasierte Ansätze die richtige Lösung und wann benötigt man Machine Learning?
- Was bedeutet maschinelles Lernen für Zeitserien?
- Was unterscheidet supervised und unsupervised Machine Learning in diesem Bereich?
- Wie könnte ein Beispiel mit einem Datensatz aus der Realität aussehen?
Buzz and feedback
Here’s what was said about this presentation on social media.
