$30 off During Our Annual Pro Sale. View Details »

Mitigating the Latency-Accuracy Trade-off in Mobile Data Analytics Systems

Anand Iyer
November 01, 2018
Research
0
200

Mitigating the Latency-Accuracy Trade-off in Mobile Data Analytics Systems

Anand Iyer

November 01, 2018
Tweet

More Decks by Anand Iyer

See All by Anand Iyer
DynaGraph: Dynamic Graph Neural Networks at Scale
anandpiyer
0
35
ASAP: Fast, Approximate Graph Pattern Mining at Scale
anandpiyer
0
290
Towards Fast and Scalable Graph Pattern Mining
anandpiyer
0
90
Monarch: Gaining Command on Geo-Distributed Graph Analytics
anandpiyer
1
100
Bridging the GAP: Towards Approximate Graph Analytics
anandpiyer
0
170
Automating Diagnosis of Cellular Radio Access Network Problems
anandpiyer
0
320
A Scalable Distributed Spatial Index for the Internet-of-Things
anandpiyer
0
300
Time-Evolving Graph Processing at Scale
anandpiyer
0
240
CellIQ: Real-Time Cellular Network Analytics at Scale
anandpiyer
0
240

Other Decks in Research

See All in Research
最先端NLP論文紹介：A Watermark for Large Language Models
kuribayashi4
0
190
Coco-Nut: 自由記述文による声質制御に向けた多話者音声・声質自由記述ペアデータセット
aya_watanabe
0
190
LLMとプランニングの未来
smatsumori
1
280
第59回名古屋CV・PRMU勉強会：ICCV2023論文紹介（自己教師あり学習）
naok615
0
190
What Are You Token About? Dense Retrieval as Distributions Over the Vocabulary
ryou0634
0
290
Mean-Variance RL (JAFEE 2023)
masakat0
1
300
論文紹介：What Learning Algorithm is In-Context Learning? Investigation with Linear Models
kosuken
0
360
CoRL2023速報
rpc
1
1.2k
大学生の学習における生成系AI活用方法と今後の課題点 (2023.09.05)
kuanyi_n_n
1
320
大規模言語モデル
chokkan
PRO
40
13k
論文紹介: Backpack Language Models
tatsuropianooo
2
340
未満児保育・保育士確保
yowata
0
100

Featured

See All Featured
Building Adaptive Systems
keathley
29
1.6k
Designing Dashboards & Data Visualisations in Web Apps
destraynor
225
51k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
113
17k
Infographics Made Easy
chrislema
236
17k
5 minutes of I Can Smell Your CMS
philhawksworth
199
19k
Refactoring Trust on Your Teams (GOTO; Chicago 2020)
rmw
24
2k
Bootstrapping a Software Product
garrettdimon
PRO
299
110k
Learning to Love Humans: Emotional Interface Design
aarron
265
39k
Fireside Chat
paigeccino
18
2.3k
How GitHub Uses GitHub to Build GitHub
holman
467
280k
Facilitating Awesome Meetings
lara
37
5.2k
実際に使うSQLの書き方 徹底解説 / pgcon21j-tutorial
soudai
52
19k

Transcript

  1. Mitigating the Latency-Accuracy Trade-
    off in Mobile Data Analytics Systems
    Anand Iyer ⋆, Li Erran Li⬩, Mosharaf Chowdhury✢, Ion Stoica ⋆
    ⋆UC Berkeley ⬩Fudan University/Pony.ai ✢University of Michigan
    MobiCom, November 1, 2018

    View Slide

  2. § Many emerging domains
    Mobile Data Analytics Very Popular

    View Slide

  3. § Many emerging domains
    Mobile Data Analytics Very Popular

    View Slide

  4. § Many emerging domains
    Mobile Data Analytics Very Popular

    View Slide

  5. § Many emerging domains
    Common Goal: Understand user/entity behavior
    Mobile Data Analytics Very Popular

    View Slide

  6. Mobile Data Analytics

    View Slide

  7. Mobile Data Analytics

    View Slide

  8. Mobile Data Analytics

    View Slide

  9. Mobile Data Analytics

    View Slide

  10. Uplink SINR >
    -11.75
    RSRQ > -16.5
    RSRQ
    Available?
    Success Drop
    Uplink SINR >
    -5.86
    CQI > 5.875
    Drop
    Drop
    Yes No
    Yes No
    No
    Yes
    Success
    No
    No
    Yes
    Yes
    Success
    Mobile Data Analytics

    View Slide

  11. Uplink SINR >
    -11.75
    RSRQ > -16.5
    RSRQ
    Available?
    Success Drop
    Uplink SINR >
    -5.86
    CQI > 5.875
    Drop
    Drop
    Yes No
    Yes No
    No
    Yes
    Success
    No
    No
    Yes
    Yes
    Success
    Mobile Data Analytics
    Tasks operate on data ingested in (near)
    real-time for low-latency decisions

    View Slide

  12. Uplink SINR >
    -11.75
    RSRQ > -16.5
    RSRQ
    Available?
    Success Drop
    Uplink SINR >
    -5.86
    CQI > 5.875
    Drop
    Drop
    Yes No
    Yes No
    No
    Yes
    Success
    No
    No
    Yes
    Yes
    Success
    Mobile Data Analytics
    Tasks operate on data ingested in (near)
    real-time for low-latency decisions
    Model/predict per-user/per-entity behavior

    View Slide

  13. Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy

    View Slide

  14. Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy

    View Slide

  15. Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Statistically
    insignificant

    View Slide

  16. Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Statistically
    insignificant

    View Slide

  17. Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Statistically
    insignificant
    High latency

    View Slide

  18. Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Statistically
    insignificant
    High latency
    1 hour latency for
    94% accuracy

    View Slide

  19. Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Statistically
    insignificant
    High latency
    1 hour latency for
    94% accuracy

    View Slide

  20. Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Statistically
    insignificant
    High latency
    1 hour latency for
    94% accuracy
    Staleness enforces
    short interval analyses

    View Slide

  21. Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Statistically
    insignificant
    High latency
    1 hour latency for
    94% accuracy
    Staleness enforces
    short interval analyses
    Highest achieved
    accuracy ~66%

    View Slide

  22. Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Statistically
    insignificant
    High latency
    1 hour latency for
    94% accuracy
    Staleness enforces
    short interval analyses
    Highest achieved
    accuracy ~66%
    Need to update models frequently

    View Slide

  23. Mitigating Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy

    View Slide

  24. Mitigating Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy

    View Slide

  25. Mitigating Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Efficient Task
    Formulations

    View Slide

  26. Mitigating Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Efficient Task
    Formulations
    Intelligent Data
    Grouping

    View Slide

  27. Hybrid Multi-Task Learning
    Mitigating Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Efficient Task
    Formulations
    Intelligent Data
    Grouping

    View Slide

  28. PCA based partitioning
    Hybrid Multi-Task Learning
    Mitigating Latency-Accuracy Trade-off
    Data collection latency
    Model Accuracy
    Efficient Task
    Formulations
    Intelligent Data
    Grouping

    View Slide

  29. Cellular RAN Performance Diagnostics

    View Slide

  30. Cellular RAN Performance Diagnostics

    View Slide

  31. Goal: Diagnose problems using data collected at base stations
    Cellular RAN Performance Diagnostics

    View Slide

  32. Base stations vary widely in data
    ��
    ����
    �����
    �����
    �����
    �����
    �����
    �����
    �����
    � � � � � �� ��
    �����������������
    ���������������������������������
    ���������

    View Slide

  33. Base stations vary widely in data
    ��
    ����
    �����
    �����
    �����
    �����
    �����
    �����
    �����
    � � � � � �� ��
    �����������������
    ���������������������������������
    ���������
    Many base
    stations do not
    collect enough
    data in small
    intervals

    View Slide

  34. Latency-Accuracy Trade-off in RANs
    0
    20
    40
    60
    80
    100
    0 1 2 3 4 5 6 7 8
    Accuracy (%)
    Data Collection Latency (minutes)
    Random Forest Lasso Regression
    10 60
    (Call drops) (Throughput)

    View Slide

  35. Latency-Accuracy Trade-off in RANs
    0
    20
    40
    60
    80
    100
    0 1 2 3 4 5 6 7 8
    Accuracy (%)
    Data Collection Latency (minutes)
    Random Forest Lasso Regression
    10 60
    High latency
    incurred for
    good accuracy
    (Call drops) (Throughput)

    View Slide

  36. Latency-Accuracy Trade-off in RANs
    0
    20
    40
    60
    80
    100
    0 1 2 3 4 5 6 7 8
    Accuracy (%)
    Data Collection Latency (minutes)
    Random Forest Lasso Regression
    10 60
    High latency
    incurred for
    good accuracy
    Staleness
    causes huge
    variance and
    errors
    (Call drops) (Throughput)

    View Slide

  37. Cellscope Architecture
    CellScope
    Domain-Specific MTL
    Gradient Boosted Trees
    RAN Performance Analyzer
    ML Lib
    Bearer Level
    Trace
    Dashboards
    Self-Organizing
    Networks (SON)
    Throughput Drop
    Feature
    Engineering
    PCA-Based Similarity Grouping
    Streaming
    Hybrid MTL

    View Slide

  38. Cellscope Architecture
    CellScope
    Domain-Specific MTL
    Gradient Boosted Trees
    RAN Performance Analyzer
    ML Lib
    Bearer Level
    Trace
    Dashboards
    Self-Organizing
    Networks (SON)
    Throughput Drop
    Feature
    Engineering
    PCA-Based Similarity Grouping
    Streaming
    Hybrid MTL

    View Slide

  39. Cellscope Architecture
    CellScope
    Domain-Specific MTL
    Gradient Boosted Trees
    RAN Performance Analyzer
    ML Lib
    Bearer Level
    Trace
    Dashboards
    Self-Organizing
    Networks (SON)
    Throughput Drop
    Feature
    Engineering
    PCA-Based Similarity Grouping
    Streaming
    Hybrid MTL

    View Slide

  40. Cellscope Architecture
    CellScope
    Domain-Specific MTL
    Gradient Boosted Trees
    RAN Performance Analyzer
    ML Lib
    Bearer Level
    Trace
    Dashboards
    Self-Organizing
    Networks (SON)
    Throughput Drop
    Feature
    Engineering
    PCA-Based Similarity Grouping
    Streaming
    Hybrid MTL

    View Slide

  41. Multi Task Learning (MTL)

    View Slide

  42. Multi Task Learning (MTL)
    Jointly learn many tasks by exploiting commonalities and differences

    View Slide

  43. Multi Task Learning (MTL)
    Jointly learn many tasks by exploiting commonalities and differences
    ℎ " = $(&' (
    , &*
    " , … , &,
    ("))

    View Slide

  44. Multi Task Learning (MTL)
    Jointly learn many tasks by exploiting commonalities and differences
    Data Train Model
    Task 1
    Data Train Model
    Task 2
    Data Train Model
    Task N
    …
    ℎ " = $(&' (
    , &*
    " , … , &,
    ("))

    View Slide

  45. Multi Task Learning (MTL)
    Jointly learn many tasks by exploiting commonalities and differences
    Data Train Model
    Task 1
    Data Train Model
    Task 2
    Data Train Model
    Task N
    …
    ℎ " = $(&' (
    , &*
    " , … , &,
    ("))
    Data
    Task 1
    Data
    Task 2
    Data
    Task N
    …
    Model
    Model
    Model
    …
    Train

    View Slide

  46. Multi Task Learning (MTL)
    Jointly learn many tasks by exploiting commonalities and differences
    Data Train Model
    Task 1
    Data Train Model
    Task 2
    Data Train Model
    Task N
    …
    ℎ " = $(&' (
    , &*
    " , … , &,
    ("))
    Data
    Task 1
    Data
    Task 2
    Data
    Task N
    …
    Model
    Model
    Model
    …
    Train
    ℎ " = $./
    (&' (
    , &*
    " , … , &,
    ("))

    View Slide

  47. Multi Task Learning (MTL)
    Jointly learn many tasks by exploiting commonalities and differences
    Data Train Model
    Task 1
    Data Train Model
    Task 2
    Data Train Model
    Task N
    …
    ℎ " = $(&' (
    , &*
    " , … , &,
    ("))
    Data
    Task 1
    Data
    Task 2
    Data
    Task N
    …
    Model
    Model
    Model
    …
    Train
    ℎ " = $./
    (&' (
    , &*
    " , … , &,
    ("))
    Assumes that all tasks are related

    View Slide

  48. MTL in Cellscope

    View Slide

  49. MTL in Cellscope
    Train
    Data
    Task 1
    Data
    Task 2
    Data
    Task N
    …
    Model
    Model
    Model
    …
    ℎ " = $%&
    (() *
    , (,
    " , … , (.
    ("))

    View Slide

  50. MTL in Cellscope
    Train
    Data
    Task 1
    Data
    Task 2
    Data
    Task N
    …
    Model
    Model
    Model
    …
    ℎ " = $%&
    (() *
    , (,
    " , … , (.
    ("))
    … …
    Train
    Data
    Task 1
    Task 2
    Task K
    …
    Model
    …
    Group 1
    Data
    Data
    Model
    Model
    Train
    Data
    Task 1
    Task 2
    Task K
    …
    Model
    …
    Group N
    Data
    Data
    Model
    Model
    ℎ " = $0(%&)
    (() *
    , (,
    " , … , (.
    ("))

    View Slide

  51. MTL in Cellscope
    Problem: Scalable maintenance of large number of models
    Train
    Data
    Task 1
    Data
    Task 2
    Data
    Task N
    …
    Model
    Model
    Model
    …
    ℎ " = $%&
    (() *
    , (,
    " , … , (.
    ("))
    … …
    Train
    Data
    Task 1
    Task 2
    Task K
    …
    Model
    …
    Group 1
    Data
    Data
    Model
    Model
    Train
    Data
    Task 1
    Task 2
    Task K
    …
    Model
    …
    Group N
    Data
    Data
    Model
    Model
    ℎ " = $0(%&)
    (() *
    , (,
    " , … , (.
    ("))

    View Slide

  52. min $ % ℎ ': )*+
    , - + /|| 1(': )*+
    )||
    Hybrid MTL
    Model estimation by L1 regularized loss minimization

    View Slide

  53. min $ % ℎ ': )*+
    , - + /|| 1(': )*+
    )||
    Hybrid MTL
    Model estimation by L1 regularized loss minimization
    Prediction error

    View Slide

  54. min $ % ℎ ': )*+
    , - + /|| 1(': )*+
    )||
    Hybrid MTL
    Model estimation by L1 regularized loss minimization
    Per base-station
    parameters
    Prediction error

    View Slide

  55. min $ % ℎ ': )*+
    , - + /|| 1(': )*+
    )||
    Hybrid MTL
    Model estimation by L1 regularized loss minimization
    Per base-station
    parameters
    Regularization
    parameter
    Prediction error

    View Slide

  56. min $ % ℎ ': )*+
    , - + /|| 1(': )*+
    )||
    Hybrid MTL
    Model estimation by L1 regularized loss minimization
    Per base-station
    parameters
    Regularization
    parameter
    Prediction error
    Decompose parameters into shared common set fc
    and base station specific set fs

    View Slide

  57. min $ % ℎ ': )*+
    , - + /|| 1(': )*+
    )||
    Hybrid MTL
    Model estimation by L1 regularized loss minimization
    Per base-station
    parameters
    Regularization
    parameter
    Prediction error
    Decompose parameters into shared common set fc
    and base station specific set fs
    $( ∑% ℎ ': )+
    , )5
    , - + /||1(': )+
    )||) + /||1(': )5
    )||

    View Slide

  58. min $ % ℎ ': )*+
    , - + /|| 1(': )*+
    )||
    Hybrid MTL
    Model estimation by L1 regularized loss minimization
    Per base-station
    parameters
    Regularization
    parameter
    Prediction error
    Decompose parameters into shared common set fc
    and base station specific set fs
    Base-station specific
    $( ∑% ℎ ': )+
    , )5
    , - + /||1(': )+
    )||) + /||1(': )5
    )||

    View Slide

  59. Hybrid MTL
    Structure of determines efficient implementation
    ℎ : #$
    , #&

    View Slide

  60. Hybrid MTL
    Structure of determines efficient implementation
    Restrict models to be of form w . x
    ℎ : #$
    , #&

    View Slide

  61. Hybrid MTL
    Structure of determines efficient implementation
    Restrict models to be of form w . x
    Leverage
    ensemble methods
    ℎ : #$
    , #&

    View Slide

  62. Hybrid MTL
    Structure of determines efficient implementation
    Dataset
    Restrict models to be of form w . x
    Leverage
    ensemble methods
    ℎ : #$
    , #&

    View Slide

  63. Hybrid MTL
    Structure of determines efficient implementation
    Dataset
    Model 2 Model 3 Model 4 Model N
    Model 1 …
    Restrict models to be of form w . x
    Leverage
    ensemble methods
    ℎ : #$
    , #&

    View Slide

  64. Hybrid MTL
    Structure of determines efficient implementation
    Dataset
    Ensemble Model
    Model 2 Model 3 Model 4 Model N
    Model 1 …
    Restrict models to be of form w . x
    Leverage
    ensemble methods
    ℎ : #$
    , #&

    View Slide

  65. Hybrid MTL
    Structure of determines efficient implementation
    Dataset
    Ensemble Model
    Model 2 Model 3 Model 4 Model N
    Model 1 …
    Restrict models to be of form w . x
    Leverage
    ensemble methods
    ℎ : #$
    , #&
    f1
    f2
    f3
    f4
    fN

    View Slide

  66. Hybrid MTL
    Structure of determines efficient implementation
    Dataset
    Ensemble Model
    Model 2 Model 3 Model 4 Model N
    Model 1 …
    Restrict models to be of form w . x
    Leverage
    ensemble methods
    Gradient Boosted
    Trees
    ℎ : #$
    , #&
    f1
    f2
    f3
    f4
    fN

    View Slide

  67. Hybrid MTL More details in the paper
    Structure of determines efficient implementation
    Dataset
    Ensemble Model
    Model 2 Model 3 Model 4 Model N
    Model 1 …
    Restrict models to be of form w . x
    Leverage
    ensemble methods
    Gradient Boosted
    Trees
    ℎ : #$
    , #&
    f1
    f2
    f3
    f4
    fN

    View Slide

  68. Data Grouping for MTL
    Key Idea: Use Principal Component Analysis (PCA) to find normal behavior

    View Slide

  69. Data Grouping for MTL
    Project large number of correlated dimensions to a
    small set of orthogonal dimensions.
    Key Idea: Use Principal Component Analysis (PCA) to find normal behavior

    View Slide

  70. Data Grouping for MTL
    Project large number of correlated dimensions to a
    small set of orthogonal dimensions.
    Key Idea: Use Principal Component Analysis (PCA) to find normal behavior
    …

    View Slide

  71. Data Grouping for MTL
    Project large number of correlated dimensions to a
    small set of orthogonal dimensions.
    Key Idea: Use Principal Component Analysis (PCA) to find normal behavior
    …
    n

    View Slide

  72. Data Grouping for MTL
    Project large number of correlated dimensions to a
    small set of orthogonal dimensions.
    Key Idea: Use Principal Component Analysis (PCA) to find normal behavior
    …
    …
    …
    …
    …
    …
    …
    …
    n
    m
    Measurement matrix

    View Slide

  73. Data Grouping for MTL
    Project large number of correlated dimensions to a
    small set of orthogonal dimensions.
    Key Idea: Use Principal Component Analysis (PCA) to find normal behavior
    …
    …
    …
    …
    …
    …
    …
    …
    n
    m
    Measurement matrix

    View Slide

  74. Data Grouping for MTL
    Project large number of correlated dimensions to a
    small set of orthogonal dimensions.
    Key Idea: Use Principal Component Analysis (PCA) to find normal behavior
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    m
    n
    k
    Measurement matrix

    View Slide

  75. Data Grouping for MTL
    Project large number of correlated dimensions to a
    small set of orthogonal dimensions.
    Key Idea: Use Principal Component Analysis (PCA) to find normal behavior
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    m
    n
    k
    Measurement matrix

    View Slide

  76. Data Grouping for MTL
    Project large number of correlated dimensions to a
    small set of orthogonal dimensions.
    Key Idea: Use Principal Component Analysis (PCA) to find normal behavior
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    m
    n
    k
    loadings
    Measurement matrix

    View Slide

  77. PCA Similarity
    Find the similarity between the principal components

    View Slide

  78. PCA Similarity
    Find the similarity between the principal components
    …
    …
    …
    …
    …
    …
    …
    …
    mA
    n
    …
    …
    …
    …
    …
    …
    …
    …
    mB
    n

    View Slide

  79. PCA Similarity
    Find the similarity between the principal components
    …
    …
    …
    …
    …
    …
    …
    …
    mA
    n
    …
    …
    …
    …
    …
    …
    …
    …
    mB
    n
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    k
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    k

    View Slide

  80. PCA Similarity
    Find the similarity between the principal components
    …
    …
    …
    …
    …
    …
    …
    …
    mA
    n
    …
    …
    …
    …
    …
    …
    …
    …
    mB
    n
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    k
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    k

    View Slide

  81. PCA Similarity
    Find the similarity between the principal components
    …
    …
    …
    …
    …
    …
    …
    …
    mA
    n
    …
    …
    …
    …
    …
    …
    …
    …
    mB
    n
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    k
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    k
    S"#$%%&'()$
    = +
    ,-.
    /
    +
    0-.
    1
    |30,
    − 50,
    |

    View Slide

  82. PCA Similarity
    Find the similarity between the principal components
    …
    …
    …
    …
    …
    …
    …
    …
    mA
    n
    …
    …
    …
    …
    …
    …
    …
    …
    mB
    n
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    k
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    …
    n
    k
    S"#$%%&'()$
    = +
    ,-.
    /
    +
    0-.
    1
    |30,
    − 50,
    | × 7809:;1'$(=,?)

    View Slide

  83. Implementation & Evaluation
    § Implemented on Apache Spark
    § Extends Mllib and provides a simple API for grouping
    § Evaluated using data from a live RAN
    § Data over several months
    § Models for two metrics: drops and throughput
    prediction
    § Also analyzed several issues in the wild

    View Slide

  84. Cellscope Reduces Latency
    0
    20
    40
    60
    80
    100
    0 1 2 3 4 5 6 7 8
    Accuracy (%)
    Data Collection Latency (min)
    Per Base Station Cellscope
    10 60
    > 90% accuracy with 3
    minutes data (compared to
    60 minutes)
    3 x

    View Slide

  85. Cellscope Improves Accuracy
    0
    20
    40
    60
    80
    100
    0 1 2 3 4 5 6 7 8
    Accuracy (%)
    Data Collection Latency (min)
    Per Base Station Cellscope
    10 60
    Achieves high
    accuracy in
    small timespans
    1.4 x
    4 x

    View Slide

  86. Real-world Analysis with Cellscope
    § Cellscope can significantly reduce operator efforts
    § Reduces the need for field trials, can build accurate models quickly
    § Up to 2 order of magnitudes (10s of hours → minutes)
    § Cellscope found new issues previously unknown
    § E.g., Grouping revealed high interference base station clusters
    § Cellscope can aid domain expert
    § Can reduce the troubleshooting search space significantly

    View Slide

  87. Much more in the paper…
    § Extending the techniques to other domains
    § Straightforward & effective
    § Comparison with strawman grouping techniques
    § Why they’re not sufficient
    § Implementation details of our hybrid MTL & API
    § Extensive evaluation
    § Real-world analysis & findings

    View Slide

  88. Summary
    § Mobile data analytics popular
    § Need low-latency decisions on live data
    § Latency-Accuracy Trade-off
    § Not enough data in small timespans, staleness determines bounds on
    data collection latencies
    § Intelligent grouping and efficient task formulations
    § Hybrid MTL and PCA based partitioning
    http://www.cs.berkeley.edu/~api
    [email protected]

    View Slide