A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

14
A Jack of All Trades Pamela J. Williams LMI March 26, 2010

Transcript of A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

Page 1: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

A Jack of All Trades

Pamela J. WilliamsLMI

March 26, 2010

Page 2: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

2

Application 1:

Controlling Mass Transit Systems

Page 3: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

3

Numerous optimization problems arise in the area of advanced automatic train control

• On a daily basis, the control system

experiences approximately 20 delays

of 5 or more minutes

• Decrease in travel time will reduce

number of needed trains, thereby

saving $2M per car

• Short Term - improve passenger

comfort for the Bay Area Rapid

Transit (BART) District

• Long Term - minimize energy

consumption

Page 4: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

4

Our enhanced control objective is to smooth interference during acceleration

• while– adhering to the

schedule,

– maintaining the worst case stopping distance,

– making required station stops,

– travelling at safe speeds, and

– braking into a station at a controlled rate.

0 g(x)

0, (x) t.s.

f(x) min

h

x

Page 5: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

5

What tools are needed to solve this problem?

• Disciplines– Physics– Numerical Analysis– Operations Research

• Programming– C++ (for UNIX systems)– Java

• Software Tools– Matlab– Excel

• Soft skills learned– Teamwork– Client Interactions

Items in orange font indicate my knowledgebase at the start of the project.

Page 6: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

6

Application 2:

Detecting protein phosphorylation sites

Page 7: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

7

Why is phosphorylation important?

• Controls DNA repair• Regulates metabolism• Can prevent diseases

Page 8: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

8

What was the research question?

• Problem Statement– Given a set of known phosphorylated sites, can we use

machine learning techniques to accurately detect the sites?

• Approach– Apply neural networks, Hidden Markov Models, etc. – Use majority vote to identify a site– Compare accuracy of individual methods and majority vote

Page 9: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

9

Integration of existing tools was the heart of this project

• Disciplines– Data mining

• Programming Languages– CSH– Perl– Python

Page 11: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

11

How to stock shelves?

• Key Questions– How much to buy?– When to buy?

• Driving metrics– Customer wait time– Inventory investment– Workload

• Items with sporadic demand are hard to manage. – The past may not be a perfect barometer for the future

Page 12: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

12

What is Sporadic Demand?

Frequent Demand

0

1

2

3

4

5

6

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

Months

Dem

and

s

Sporadic Demand

02468

101214

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

MonthsD

eman

ds

Infrequent demands, quantities vary greatly (not always low quantities!)

Frequent demands, quantities vary moderately

Page 13: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

13

What tools are needed to solve this problem?

• Disciplines– Statistics– Inventory control

• Programming– Visual Studio C++

• Database Tools– FoxPro– Microsoft Access

• Soft skills learned– Project management

Items in orange font indicate my knowledgebase at the start of the project.

Page 14: A Jack of All Trades Pamela J. Williams LMI March 26, 2010.

14

Who is LMI?

• Formerly known as Logistics Management Institute• A not-for-profit government consulting company• Headquarters in Northern, VA• We are hiring!

– Go to http://www.lmi.org/careers/careers.aspx– Computer scientist/mathematician posting