El camino hacia el éxito con las bases de datos de grafos, la ciencia de datos de grafos y la IA generativa

© 2023 Neo4j, Inc. All rights reserved.
El camino hacia el éxito con
las bases de datos de
grafos, la ciencia de datos
de grafos y la IA
generativa
Luis Salvador
PreSales Engineer Iberia, Neo4j
1
Pierre Halftermeyer
PreSales Engineer France, Neo4j

DATA GROWTH IS ACCELERATING.
EVERYONE IS CONNECTED TO EVERYTHING
200+
Zettabytes
In cloud data storage
by 2025
65
Average number of
enterprise SaaS apps
41B
Connected IOT
devices by 2025
2
2

3
TODAY’S REALITY
Connections
in data are as
valuable as the
data itself
3

4
Legacy
systems
can’t keep up
“Relational” Databases
don’t handle relationships well
NoSQL Databases
don’t handle relationships at all

Graph creates a
more intuitive and
connected view of
data relationships,
unlocking deeper
insights and context
5
5

Hybrid Workload Duality
6
Intelligent
Applications
Transactions -
Security -
Performance & Scalability -
ACID Consistency -
Intelligent Modeling
- Extensive & Supported Algo Library
- Scalable
- Graph Visualization
- Graph Transformations
Graph
Transactions
Graph Analytics
& Data Science

THE PROPERTY GRAPH: SIMPLY POWERFUL
Employee City
Company
Nodes represent
objects (nouns)
Relationships are directional
Relationships connect nodes are
represent actions (verbs)
Relationships can have properties
(name/value pairs)
Nodes can have
properties (name/value
pairs)
name: Amy Peters
date_of_birth: 1984-03-01
employee_ID: 1
:HAS_CEO
start_date: 2008-01-20
:LOCATED_IN
7
7

8
NATIVE GRAPH ARCHITECTURE: FAST, FLEXIBLE, SCALABLE
Native Graph Database
At write time
Data is .connected.
as it is stored
At read time
.Lightning-fast. retrieval of data
and relationships via pointer
chasing
Relational Model
Graph Model
ACTED_IN
ACTED_IN
ACTED_IN
Tom Cruise
Mission
Impossible
Oblivion
Person Movie
Person-Movie
Tom
Cruise
Top Gun
Vs.

QUERY PERFORMANCE AS # OF JOINS INCREASE
Connectedness and Size of Data Set
Response
Time
Relational and
other NoSQL
databases
Native Graph Database
1000x Advantage
Minutes to milliseconds
5+ hops
3+ degrees
Thousands of connections
0 to 2 hops
0 to 3 degrees
Few connections
9
9

Analytics
Tooling
Graph Transactions
Data Integration
Dev.
& Admin
Drivers & APIs Discovery & Visualization
Graph Analytics &
Data Science
Native graph technology for apps & analytics
10
Developers
Admins
Applications Business Users
Data Analysts
Data
Scientists
Bloom
Kafka Connector BI Connector
Spark Connector
GRANDstack
(GraphQL, React,
Apollo, Neo4j)
Data Warehouse Conn.

Developer Productivity: Rich tooling and easy onramp
ops manager
11
data importer
Visualize and explore your data
Query editor and results visualizer
Code-free data loader and modeler
AuraWorkspace
Unified Workspace

PLUGS INTO YOUR EXISTING DATA ECOSYSTEM
12
Apache Spark
Connector
Data Warehouse
Connector
Apache Kafka
Connector
Neo4j BI
Connector

Enterprise-Grade: Security and Trust Built In
Single Sign-On Secure Development
Practices
Dedicated VPC Role- & Schema-Based
Access Control
Encryption
(At-Rest, In-Transit,
and Intra Cluster)
SOC 2 Type 1
13

14
19 Billion
1.2 Trillion!
128 Billion
Neo4j Fabric: Scaling Up & Scaling Out!

15
Neo4j
Graph Data Science

16
What’s in it for you:
● Improve model accuracy by 30%
● Simplify processes and remove
headaches
● More projects into production without
additional hiring
Neo4j Graph Data Science
Analytics
Feature
Engineering
Data
Exploration
Graph
Data
Science
Queries & Search
Machine Learning Visualization

When do you need Graph Algorithms?
Query (e.g. Cypher)
Real-time, local decisioning
and pattern matching
Graph Algorithms
Global analysis and iterations
You know what you’re looking
for and making a decision
You’re learning the overall structure
of a network, updating data, and
predicting
Local Patterns Global Computation

What’s important?
Prioritization
Who has the most connections?
Who has the highest page rank?
Who is an influencer?
What’s unusual?
Anomaly & Fraud Detection
Where is a community forming?
What are the group dynamics?
What’s unusual about this data?
What’s next?
Predictions
What’s the most common path?
Who is in the same community?
What relationship will form?
18
Pl
ay
s
Lives_in
In_sport
Likes
F
a
n
_
o
f
Plays_for
K
n
o
w
s
Knows
Knows
K
n
o
w
s
Graph Structure Improves Data Science Outcomes

19
With The Largest Catalog of Graph Algorithms
Pathﬁnding &
Search
Centrality &
Importance
Community
Detection
Supervised
Machine Learning
Heuristic Link
Prediction
Similarity Graph
Embeddings
…and more
Graph algorithms are a set of instructions that visit the nodes of a graph to
analyze the relationships in connected data.

20
65+ Graph Data Science Techniques in Neo4j
Pathfinding &
Search
• Shortest Path
• Single-Source Shortest Path
• All Pairs Shortest Path
• A* Shortest Path
• Yen’s K Shortest Path
• Minimum Weight Spanning Tree
• K-Spanning Tree (MST)
• Random Walk
• Breadth & Depth First Search
Centrality &
Importance
• Degree Centrality
• Closeness Centrality
• Harmonic Centrality
• Betweenness Centrality & Approx.
• PageRank
• Personalized PageRank
• ArticleRank
• Eigenvector Centrality
• Hyperlink Induced Topic Search (HITS)
• Influence Maximization (Greedy, CELF)
Community
Detection
• Triangle Count
• Local Clustering Coefficient
• Connected Components (Union Find)
• Strongly Connected Components
• Label Propagation
• Louvain Modularity
• K-1 Coloring
• Modularity Optimization
• Speaker Listener Label Propagation
Supervised
Machine Learning
• Node Classification
• Link Prediction
• Node Regression
… and more!
Heuristic Link
Prediction
• Adamic Adar
• Common Neighbors
• Preferential Attachment
• Resource Allocations
• Same Community
• Total Neighbors
Similarity
• Node Similarity
• K-Nearest Neighbors (KNN)
• Jaccard Similarity
• Cosine Similarity
• Pearson Similarity
• Euclidean Distance
• Approximate Nearest Neighbors (ANN)
Graph
Embeddings
• Node2Vec
• FastRP
• FastRPExtended
• GraphSAGE
• Synthetic Graph Generation
• Scale Properties
• Collapse Paths
• One Hot Encoding
• Split Relationships
• Graph Export
• Pregel API (write your own algos)

21
Pathfinding
Pathfinding and Graph Search
algorithms are used to identify
optimal routes, and they are often a
required first step for many other
types of analysis.
Applications: Shortest path,
Optimal paths, route availability,
What-if analysis, Alternate Routing,
Disaster Recovery
https://neo4j.com/docs/graph-data-science/c
urrent/algorithms/pathfinding/

22
Centrality
Finds important nodes based on
relationships to other nodes in the graph
Applications: Outlier detection,
preprocessing, Inﬂuencer detection,
Bridge points, points of failure,
vulnerabilities
Color and size represent inﬂuence
based on centrality scores
urrent/algorithms/centrality/

23
Community Detection
Evaluates how groups of nodes are
clustered or partitioned, as well as
their tendency to strengthen or
break apart
Applications: Recommendations,
homogeneity, disjoint communities,
outlier detection, preprocessing
urrent/algorithms/community/

24
Similarity
Evaluates how alike nodes are at
an individual level either based on
node attributes, neighboring
nodes, or relationship properties.
Applications: Recommendations,
What-if analysis, Disambiguation
urrent/algorithms/similarity/

25
Link Prediction
These methods compute a score for a
pair of nodes, where the score could
be considered a measure of proximity
or “similarity” between those nodes
based on the graph topology.
Applications: Context-enrichment,
contact tracing, spam detection,
association in social networks etc.
urrent/algorithms/linkprediction/

26
Embeddings
A graph embedding is a way of representing each node in your graph
as a ﬁxed-length vector.
• Preserves key features
• Reduces dimensionality
• Can be decoded
Different techniques may represent different aspects of a graph, and
may use different approaches to learn that representation

Node Embedding

Node Embedding
Encode nodes such that similarity in the
embedding space, i.e. cosine similarity,
approximates similarity in the graph

GraphML: Regression & autotuning
Node classiﬁcation:
“What label should this node have?”
Property regression:
“What’s the value for this missing property?”
Link Prediction:
“Should there be a relationship between these 2 nodes?”
f(x)
f(x)
f(x)
We discover the best model for you - you just supply the data!

GDS example: Fraud Detection
Graph algorithms for entity link analysis detect first party and synthetic
identity fraud across channels in financial sector and other industries
Detect patterns of
fraudulent activity across
channels
Isolate fraudsters by
running community
detection, centrality
and embedding
algorithms
Identify potential
fraudsters by computing
similarities between
known fraudsters and
customers
Prevent fraud by
flagging
transactions and
clients with higher
risk

And made it seamless for all ecosystems and pipelines
Graph Data Science
BI & VISUALIZATIONS
INGEST
STORE
PROCESS
Apache
Kafka
MACHINE LEARNING
Cloud
Functions
Neo4j
Bloom
PubSub
DataProc
Analytics
Feature
Engineering
Data
Exploration
Graph
Data
Science
Business
Applications &
Existing Systems
Files (unstructured,
structured)
TensorFlow
KNIME Python
Cloud Storage
AWS
Lambda
31
Graph Database

What is Neo4j Bloom?
Neo4j Bloom is our tool to Search, Explore and Discover graph data.
Bloom’s no-code UI makes it the ideal interface for analysts to extract
valuable information from knowledge graphs.
Neo4j Product
💪
Visualize
Large Graphs
🧠
Data Science
Capabilities
📖
Natural Language
Search

View the most well connected and influential nodes
Recommendations from shared user interactions and associations
Our Visualizations Make analysis easy to understand
33

Neo4j Visualisation tool - Bloom
34

35
Neo4j and
LLMs

Generative AI Is Predicted to Unlock
In Economic Value
The economic potential of generative AI: The next productivity frontier, McKinsey & Company, June 2023.
$6.6 Trillion
Up to 3.3% productivity improvement annually

37

38

LLM Hallucinations
Definition: Language models generate text
that is incorrect, nonsensical, or unreal.
• Appear to answer questions confidently
even if they don’t have facts
• May provide contradicting or inconsistent
responses to similar prompts

In Summary…

How to Help LLMs Do Better?
Few-Shot Learning
Fine-Tuning Grounding
Provide completed
examples “shots” to the AI
as context in prompts.
a.k.a In-Context Learning
Provide additional training
data to better tune GenAI
to your use case
Provide AI with the
information to use for
generating responses
All of these are useful, but grounding is where Neo4j adds value
Neo4j as the data source
for Grounding

Ground LLMs in Neo4j’s Knowledge Graph
42

43
https://neo4j.com/blog/vector-search-deeper-insights/

Vector Search - What is it?

Neo4j Should be the Database for Grounding
Vector DB Limitations
Knowledge Graph
Strengths
Neo4j Differentiators
Similarity ≠ Relevance or
Accuracy
Black-Box (Sub-Symbolic)
Duplicate & incomplete
results
Missing reference information
Challenging to answer
multi-hop questions
Difﬁcult for SME to correct
Relevancy beyond just
similarity
Transparent symbolic
representation
Condensed information
storage
References between
documents calculated before
query time
Enables human correction
LLMs understand Cypher
Vectors + Cypher
Index for many data types
(numeric, geopoints, dates)
Fine-grained security and
access control
ACID transactions,
high-availability, and scale
Ecosystem integration
Available on all clouds
Graph Data Science for
enhanced ML

API
Knowledge Graph
Neo4j AuraDS
Graph Data
Science
Graph DB
Intelligent Apps
Knowledge
Extraction and
Ingestion
Structured
Unstructured
Ontologies
GCP Vertex AI
Data Sources API Layer
Customer Service
Ticket Triaging
Recommendations
News Content &
Discovery
Enterprise
Knowledge Search
Patient
Prioritization
Clinical Decision
Support Systems
Pharmacovigilance
Health Assistants
FAQ Bots
Knowledge Graph and Generative AI Reference Architecture
GCP Document AI
Bloom
Generative AI Generative AI
Dataproc
Dataflow
Pubsub

G R A C I A S
49

El camino hacia el éxito con las bases de datos de grafos, la ciencia de datos de grafos y la IA generativa

El camino hacia el éxito con las bases de datos de grafos, la ciencia de datos de grafos y la IA generativa

Recommandé

Contenu connexe

Similaire à El camino hacia el éxito con las bases de datos de grafos, la ciencia de datos de grafos y la IA generativa

Similaire à El camino hacia el éxito con las bases de datos de grafos, la ciencia de datos de grafos y la IA generativa(20)

Plus de Neo4j

Plus de Neo4j(20)

Dernier

Dernier(20)

El camino hacia el éxito con las bases de datos de grafos, la ciencia de datos de grafos y la IA generativa