Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Intro to AI and Big Data processing

20d0ddc61e80bce04a63680da0160756?s=47 Aletheia
November 11, 2020

Intro to AI and Big Data processing

Introductory talk about artificial intelligence and big data processing to leverage ML models in production

20d0ddc61e80bce04a63680da0160756?s=128

Aletheia

November 11, 2020
Tweet

Transcript

  1. www.neosperience.com | blog.neosperience.com | info@neosperience.com Neosperience Empathy in Technology AI,

    neural networks and applications Novembre, 11th 2020
  2. Luca Bianchi Who am I? github.com/aletheia https://it.linkedin.com/in/lucabianchipavia https://speakerdeck.com/aletheia Chief Technology

    Officer @ Neosperience Chief Technology Officer @ WizKey Serverless Meetup and ServerlessDays Italy co-organizer www.bianchiluca.com @bianchiluca
  3. Artificial Intelligence What is all this buzz about? Why now

    and not before (or after)?
  4. Different approaches How? Human intelligence relies on a powerful supercomputer

    - our brain - much more powerful than actual HPC servers and is capable of switching between many different “algorithms” to understand reality, each one of them is context-independent. • Filling gaps in Existing Knowledge • Understand and apply Knowledge • Semantically reduce uncertainty • Notice similarity between old/new The most powerful capability of our brain and the common denominator of all these features is the capability humans to learn from experience. Learning is the key.
  5. Artificial Intelligence “the theory and development of computer systems able

    to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.”
  6. Roadmap of Deep Learning

  7. Machine Learning What is exactly a machine that learns?

  8. Machine Learning The importance of Experience • Machine Learning (ML)

    algorithms have data as input, ‘cause data represents the Experience.
 This is a focal point of Machine Learning: large amount of data is needed to achieve good performances. • The Machine Learning equivalent of program in ML world is called ML model and improves over time as soon as more data is provided, with a process called training. • Data must be prepared (or filtered) to be suitable for training process. Generally input data must be collapsed into a n-dimensional array with every item representing a sample. • ML performances are measured in probabilistic terms, with metrics called accuracy or precision. An operational definition “A computer program is said to 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”
  9. Regression Regression analysis helps one understand how the typical value

    of the dependent variable (or 'criterion variable') changes when any one of the independent variables is varied, while the other independent variables are held fixed. Is a statistical method of data analysis. The most common algorithm least square method that provides an estimation of regression parameters. When dataset is not trivial estimation is achieved through is gradient descent.
  10. Regression — Use cases Statistical regression is used to make

    predictions about data, filling the gaps Regression, even in the most simple form of Linear Regression is a good tool to learn from data and make predictions based on data trend. Common Scenarios • Stock price value • Product Price Estimation • Age estimation • Customer satisfaction rate defining variables such as response-time, resolution-ration we can forecast satisfaction level or churn • Customer Conversion rate estimation (based on click data, origin, timestamp, ...)
  11. Classification Classification is the problem of identifying to which of

    a set of categories (sub-populations) a new observation belongs, on the basis of a training set of data containing observations (or instances) whose category membership is known. Most used algorithms for classification are: • Logit Regression • Decision Trees • Random Forest
  12. Common Scenarios • Credit scoring • Human Activity Recognition •

    Spam/Not Spam classification • Customer conversion prediction • Customer churn prediction • Customer personas classification Classification — Use cases Classification is used to detect the binary outcome of a variable Classification is often used to classify people into pre-defined clusters (good-payer/bad-payer, in/out target, etc.)
  13. Clustering is the task of grouping a set of objects

    in such a way that objects in the same group (called a cluster) are more similar (in some sense or another) to each other than to those in other groups (clusters). The difference between algorithms is due to the similarity function that is used: • Centroid based clusters • Density based cluster • Random Forest
  14. Common Scenarios • Similar interests recognition • Shape detection •

    Similarity analysis • Customer base segmentation Clustering — Use cases Clustering is used to segment data Clustering labels each sample with a name representing its belonging cluster. Labelling can be exclusive or multiple. Clusters are dynamic structures: they adapt to new sample coming into the model as soon as thy label them.
  15. a good ML model is just a component of an

    AI system
  16. An example Process data coming from the browser

  17. Neosperience Cloud Cloud Understand Engage Grow

  18. events from the browser

  19. Client JS library embedded into the webpage Browser generated events

    An user navigating the webpage produces events with a flexible structure that are sent to the backend Three types of events: • low-level: in response to mouse/touch events, agnostic • mid-level: related to webpage actions, domain-specific • high-level: structured customer-specific events Constraints • response time: beacon support is strict on time • volume: millions of events within a single month • throughput: events could peak to thousands within a few seconds
  20. a pipeline to ingest and process events

  21. convert analyze ingest Collect and process events through pipeline stages.

    From the browser to customer insights Data processing pipeline A service unable to ramp up as quickly as the events flow to the system would result in loss of data User Insight ingestion service collects data from many different customers leading to unpredictable load Events needs to be stored, then processed and consolidated into an user profile ingest events collect and send to storage store raw events extract, transform, load store baked events process events to build insights store customer profile
  22. A difficult task with a lot of uncertainty Storing and

    analyzing data The amount of data collected by database grows to millions of data points very quickly. i.e. for a single customer, ~130M events collected in just one month Data access pattern is not well defined (parameters within query) and could change whenever high level events are managed for a customer specific context Pulling data from DynamoDB with no clear access pattern means a full table scan for each query. It is not just slow, but also very expensive.
  23. A consolidated technology with unparalleled flexibility Introducing the Data Lake

    Amazon S3 - an object storage - 99.99% availability - designed from the ground up to handle traffic for any Internet application - multi AZ reliability - cost effective Amazon S3
  24. Lambda — invocations

  25. Extract, Transform raw events and load into a data catalog

    Data pipeline extract, transform, load - Processes events and load them into AWS Glue catalog, then saves to S3 - Aggregates events based on their visit time extracting user sessions - Transforms events encoding respective types into readable and compact format - Uses Apache Spark language to build processing jobs
  26. Data analysis stage Data pipeline - Data is loaded into

    AWS Glue catalog and into Amazon S3 from previous stage - Amazon Athena queries build customer insights, leveraging external ML services through Amazon SageMaker - Resulting insights are stored into Amazon Elasticsearch
  27. putting all together..

  28. User Insight data pipeline

  29. Wrap up

  30. Leveraging an ML model means having in place a data

    driven architecture Operations + Machine Learning —> MLOps • Think about streams of data flowing into an application • Collect data in a reliable way • Data preparation is often more relevant than ML models • ETL jobs to aggregate and transform data • Query data to filter relevant data • Deploy ML model and think about scalability • Consolidate data back into NoSQL data stores (i.e. a data lake)
  31. Empathy in Technology

  32. bit.ly/nsp-assi-ai-20201111

  33. www.neosperience.com | blog.neosperience.com | info@neosperience.com