Never miss a bus again with this one crazy trick Cooper Sloan - - PowerPoint PPT Presentation

never miss a bus again with this one crazy trick
SMART_READER_LITE
LIVE PREVIEW

Never miss a bus again with this one crazy trick Cooper Sloan - - PowerPoint PPT Presentation

Never miss a bus again with this one crazy trick Cooper Sloan Background - Undergrad: MIT, BS in Computer Science - Masters: MIT, MEng in Artificial Intelligence Overview Problem Risks/Challenges Implementation Final Design


slide-1
SLIDE 1

Never miss a bus again with this

  • ne crazy trick

Cooper Sloan

slide-2
SLIDE 2

Background

  • Undergrad: MIT, BS in Computer Science
  • Masters: MIT, MEng in Artificial Intelligence
slide-3
SLIDE 3

Overview

◻Problem ◻Risks/Challenges ◻Implementation ◻Final Design ◻Takeaways

slide-4
SLIDE 4

Problem

  • Bus schedules are

unreliable

  • Complicated interactions
  • Deep learning may have

the answer

slide-5
SLIDE 5

Problem cont.

slide-6
SLIDE 6

Challenge 1: Feature Selection

  • Clumping
  • Dwell Time
  • Travel Time
  • Schedule Adherence
  • Temporal Features
slide-7
SLIDE 7

Challenge 2: Noisy/Incomplete Data

  • Urban Valleys
  • 1: Naive Approach
  • 2: Interpolation
slide-8
SLIDE 8
slide-9
SLIDE 9

Challenge 3: Model Design

  • Train/test split
  • Data partitioning
  • Routes
  • Architecture
  • Evaluation
slide-10
SLIDE 10

Iteration

  • Overfitting
  • Signal to noise
  • Temporal data
slide-11
SLIDE 11

Architecture

slide-12
SLIDE 12

Results

slide-13
SLIDE 13

Comparison

  • Knet
  • Pytorch
  • Imperative
  • Easy Debugging
  • More expressiveness
  • Mocha
  • Tensorflow/Keras
  • Declarative
  • Easy to write (usually...)
  • Hard Debugging
  • Good for simple models
  • Less expressive
slide-14
SLIDE 14

Summary

  • Predict bus arrival times using

neural networks

  • GPS data from MBTA
  • Use travel time between stops as

features

  • 3 hidden layers + RNN