Slides PPT

Report
Incentivize Crowd Labeling under
Budget Constraint
Qi Zhang, Yutian Wen, Xiaohua Tian,
Xiaoying Gan, Xinbing Wang
Shanghai Jiao Tong University, China
Outline
 Introduction to Crowdsourcing Mechanism
 Problem Formulation and Mechanism Setting
 Mechanism Analysis
 Performance Evaluation
2
Background
 Crowdsourcing systems leverage human wisdom to
perform tasks, such as:
 Image classification
 Character recognition
 Data collection
3
Types of Tasks
 Tasks can be divided into two categories:
 Structured response format
 Binary choice
 Multiple choice
 Real Value
 Unstructured response format
 Logo design
4
Motivating Example
 Example: Image classification
Workers
Dog
Task
Dog
Cat
Allocation
Crowdsourcing
Platform
Inference
Algorithm
Dog
Cat
Dog
5
Framework: Reverse Auction
(1)Tasks
(2)Bids
(3)Winning bids determination
(4)Winning bids
(5)Answers
(6)Payments
6
Major Challenges(1)
To design a successful crowdsourcing system
 Task Allocation (winning bids)
• Tasks should be allocated evenly
 Payment Determination:
• Must provide proper incentives (monetary rewards)
 Inference Algorithm:
• Should improve overall accuracy
• Should address the diversity of the crowd
7
Major Challenges(2)
We need to model on
 Diverse task difficulty
• Dog or Cat
• Older than 30 or Not
 Diverse worker quality
Cat
8
Model on Tasks(1)
We focus on binary choice tasks
 Each task is a 0 – 1 question
(Assumption) Each worker is uniformly reliable
 Task
Soft Label
• Probability that the task is labeled as 1( by a reliable worker)
Crowd Label
0 or 1
9
Model on Tasks(2)
The soft label is viewed as a random variable
drawn from Beta distribution
Prior Parameters
Update parameters (a,b) by Bayes rule
Posterior
Inference
Prior
Likelihood
The task is inferred as 1
More than half agree
10
Framework: Reverse Auction
 The platform publicizes a set of binary tasks
 Workers reply with a set
of bids
• Each bid is a task-price pair
 (Allocation) The platform sequentially decide winning bids
 (Payment) Winning workers provide labels and get payment
11
Crowdsourcing Platform Utility
 After observing all crowd labels
updated as
, the distribution is
 Platform Utility: KL Divergence between
the initial and the final distribution
12
Problem Formulation
We want:
 Platform utility maximization under budget constraint
 Individual rationality
 Truthful about the cost
Truthful bid
Untruthful bid
 Computation Efficiency
13
Allocation Scheme (1)
The task allocation(winning bid determination) is sequential :
 Candidate selection
•
one candidate a round
Candidate
Remaining bids
 Proportional rule check
Discard
Winning bid
 Answer collection & Soft label update
The allocation scheme repeats the 3 steps until
All bids
Discard
Winning bids
14
Allocation Scheme (2)
 The candidate selection is greedy
• The largest platform utility gain per unit price
PU Gain
Candidate
Price
• Platform utility gain:
Current distribution
Updated distribution
15
Allocation Scheme (3)
 Proportional rule check
budget
price
fraction ratio
 Soft label update
•
Collect the answer from the winning bid
•
Update the soft label according to Bayes rule
16
Allocation Scheme (5)
Candidate selection
Proportional rule check
Soft label update
Computationally efficient !
17
Payment Scheme(1)
Winning bids
Discard
{A, B, C}
{D, E, F}
Kick out C
{ A,B,D,E,F }
Winning bids
{A,
B,
b2
b1
C
D,
b3
Discard
{F}
E}
b4
b1 is the minimum price
that bid C can replace bid A
 p(C) = max {b1,b2, b3, b4}
18
Payment Scheme(2)
(Proposition)The winning bid C is paid threshold payment.
 p(C)
C’s payment, b(C)
C’s bid
 if b(C) < p(C), C is a winning bid
if b(C) > p(C), C is discarded
Winning bids
{A,
B,
b2
b1
D,
b3
E}
b4
C
p(C)=max { b1, b2, b3, b4}
19
Payment Scheme(3)
(Proposition)The incentive mechanism is truthful

Each bid

Workers will truthfully reveal the cost as asked price
has a cost
Why?
Proof:
Threshold payment + Greedy candidate Selection
20
Individual Rationality
(Proposition)The incentive mechanism is individual rational

The utility of a winning bid is nonnegative
Proof: Let us consider the winning bid C
1.
C is the 3rd winning bid.
2.
The first 2 bids are the same
3.
b3 is the minimum price
that bid C can replace the new
New Winning bids
{A,
3rd
bid (D)
b1
B,
b2
D,
b3
E}
b4
It is true that b3 > b(c) !
p(C) = max {b1, b2, b3, b4}, p(C) > b3
{ A, B, C}
Original wining bids
p(C) > b(C)
21
Budget Feasibility
(Proposition, Payment Bound) Payment to each winning bid
is upper bounded by
•
Proportional rule:
•
Set
22
Performance Evaluation(1)
 Benchmark
1.
Untruthful Allocation: Workers’ cost is public information
2.
Random Allocation: Candidate selection is random
Truthful
Benchmark 1

Benchmark 2

My Mechanism

Running
Time
Platform
Utility
High
Low
Low
High
23
Performance Evaluation(2)
 Metric 1 : Platform Utility
•
Platform utility vs. Budget
Price of Truthfulness
Gain over random allocation
24
Performance Evaluation(3)
 Metric 2 : Budget Utilization
•
Payment / Budget
Budget utilization gain
Over random allocation
25
Thank you !
Presented by : Qi Zhang

similar documents