distributed deep rl on spark strata singapore

14

Upload: adam-gibson

Post on 16-Apr-2017

1.090 views

Category:

Data & Analytics


0 download

TRANSCRIPT

Page 1: Distributed deep rl on spark   strata singapore
Page 2: Distributed deep rl on spark   strata singapore

SKYMIND INTELLIGENCE LAYER (SKIL)REFERENCE ARCHITECTURE

Page 3: Distributed deep rl on spark   strata singapore

Overview

● Why am I up here?● Reinforcement Learning● Use cases● Demo!● Deep Reinforcement Learning● Rl4j● Dl4j● Spark/RL - why?

Page 4: Distributed deep rl on spark   strata singapore

Why am I up here?

Wrote this -->Book Giveaway!

Page 5: Distributed deep rl on spark   strata singapore

Reinforcement Learning

● Learn a “policy” with repeated trial and error

● An agent explores a search space● Learns from rewards and penalties

each time it takes a step● Think of win/lose scenarios● Rewards/punishment set by an

“environment”

Credit: http://ai.berkeley.edu/reinforcement.html

Page 6: Distributed deep rl on spark   strata singapore

Use cases (not games!)

● Risk analysis (loans)● Network Intrusion● Learning patterns from

simulations (MCMC)

Page 7: Distributed deep rl on spark   strata singapore

Demo!

Cartpole (Hello world of RL)

Page 8: Distributed deep rl on spark   strata singapore

Deep Reinforcement Learning

● Teach a neural net from environment● Policy determines gradient descent steps ● Most work has been based on raw frames

from games (pixel input)● Various techniques (A3C,Policy Gradients,Deep

Q,..)● Core idea: Neural net has a softmax

(probability distribution) mapped to actions to take in an environment

Page 9: Distributed deep rl on spark   strata singapore

RL4j

● Deep Reinforcement Learning library for Java

● Openai Gym Intregration● Deep Reinforcement Learning

with DL4j● Implementations of A3C,DeepQ,

Policy Gradients● Openai Gym Java Bindings

Page 10: Distributed deep rl on spark   strata singapore

Dl4j

Page 11: Distributed deep rl on spark   strata singapore

Dl4j

● Import keras models● Focus on running in production● Integrate with existing big data ecosystem● Transparent usage of cpus and gpus● End to end ecosystem for building data

products (not just algorithms!)

Page 12: Distributed deep rl on spark   strata singapore

Spark/RL Why?

● Spark is distributed compute● A lot of simulations and

environments to run● Distributed workers running

experiments in parallel● Data Parallelism with neural nets

Page 13: Distributed deep rl on spark   strata singapore

Summary

● Spark for orchestrating simulations● Spark for distributed training● Integrated storage with HDFS● Orchestrate GPU based spark jobs● Easy to hook in to production (java/scala)● Great streaming ecosystem for incremental

updates

Page 14: Distributed deep rl on spark   strata singapore