Incentivize Crowd Labeling under Budget Constraint Qi Zhang, Yutian Wen, Xiaohua Tian, Xiaoying Gan,...
-
Upload
jahiem-bess -
Category
Documents
-
view
212 -
download
0
Transcript of Incentivize Crowd Labeling under Budget Constraint Qi Zhang, Yutian Wen, Xiaohua Tian, Xiaoying Gan,...
Incentivize Crowd Labeling under
Budget Constraint
Qi Zhang, Yutian Wen, Xiaohua Tian,
Xiaoying Gan, Xinbing Wang
Shanghai Jiao Tong University, China
2
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
5
Motivating Example
Example: Image classification
Workers
Allocation
Crowdsourcing
Platform
Task Dog
Dog
Cat
Cat
Dog
Inference
AlgorithmDog
6
Framework: Reverse Auction
(1)Tasks
(2)Bids
(3)Winning bids determination
(4)Winning bids
(6)Payments
(5)Answers
7
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
8
Major Challenges(2)
We need to model on
Diverse task difficulty
• Dog or Cat
• Older than 30 or Not
Diverse worker quality
Cat
9
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
10
Model on Tasks(2)
The soft label is viewed as a random variable drawn from Beta distribution
Update parameters (a,b) by Bayes rule
Inference
The task is inferred as 1
Prior Parameters
PriorPosterior
Likelihood
More than half agree
11
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
12
Crowdsourcing Platform Utility
After observing all crowd labels , the distribution is updated as
Platform Utility: KL Divergence between
the initial and the final distribution
13
Problem Formulation
We want:
Platform utility maximization under budget constraint
Individual rationality
Truthful about the cost
Truthful bid Untruthful bid
Computation Efficiency
14
Allocation Scheme (1)
The task allocation(winning bid determination) is sequential :
Candidate selection
• one candidate a round
Proportional rule check
Answer collection & Soft label update
The allocation scheme repeats the 3 steps until
Remaining bids Candidate
Discard
Winning bid
All bids
Discard Winning bids
15
Allocation Scheme (2)
The candidate selection is greedy
• The largest platform utility gain per unit price
• Platform utility gain:
PU Gain
Price
Candidate
Updated distributionCurrent distribution
16
Allocation Scheme (3)
Proportional rule check
Soft label update
• Collect the answer from the winning bid
• Update the soft label according to Bayes rule
price
budget
fraction ratio
17
Allocation Scheme (5)
Candidate selection
Proportional rule check
Soft label update
Computationally efficient !
18
Payment Scheme(1)
p(C) = max {b1,b2, b3, b4}
Winning bids
{A, B, C}
Discard
{D, E, F}
Kick out C
{ A,B,D,E,F }
Winning bids
{A, B, D, E}
Discard
{F}
C
b1b2 b3 b4
b1 is the minimum price
that bid C can replace bid A
19
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
p(C)=max { b1, b2, b3, b4}
Winning bids
{A, B, D, E}
C
b1b2 b3 b4
20
Payment Scheme(3)
(Proposition)The incentive mechanism is truthful
Each bid has a cost
Workers will truthfully reveal the cost as asked price
Why?
Proof: Threshold payment + Greedy candidate Selection
21
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 3rd
bid (D)
It is true that b3 > b(c) !
p(C) = max {b1, b2, b3, b4}, p(C) > b3
p(C) > b(C)
New Winning bids
{A, B, D, E}
b1b2 b3 b4
{ A, B, C}
Original wining bids
22
Budget Feasibility
(Proposition, Payment Bound) Payment to each winning bid
is upper bounded by
• Proportional rule:
• Set
23
Performance Evaluation(1)
Benchmark
1. Untruthful Allocation: Workers’ cost is public information
2. Random Allocation: Candidate selection is random
Truthful Running Time
Platform Utility
Benchmark 1 High
Benchmark 2 Low Low
My Mechanism High
24
Performance Evaluation(2)
Metric 1 : Platform Utility
• Platform utility vs. Budget
Price of Truthfulness
Gain over random allocation
25
Performance Evaluation(3)
Metric 2 : Budget Utilization
• Payment / Budget
Budget utilization gain
Over random allocation
Thank you !
Presented by : Qi Zhang