Podcast Special Episode 2 – The Future of AI with Dr. Ed Felten


In this audio-only Becoming a Data Scientist Podcast Special Episode, I interview Dr. Ed Felten, Deputy U.S. Chief Technology Officer, about the Future of Artificial Intelligence (from The White House!).

You can stream or download the audio at this link (download by right-clicking on the player and choosing “Save As”), or listen to it in podcast players like iTunes and Stitcher. Enjoy!

Show Notes:

The White House Names Dr. Ed Felten as Deputy U.S. Chief Technology Officer

Edward W. Felten at Princeton University

Dr. Edward Felten on Wikipedia

White House Office of Science and Technology Policy (OSTP)

The Administration’s Report on the Future of Artificial Intelligence (White House Report from October 2016)

Artificial Intelligence, Automation, and the Economy (White House Report from December 2016)

Ed Felten on Twitter: Official / Personal

Freedom to Tinker blog



Source: becomingadatascientist.com – Podcast Special Episode 2 – The Future of AI with Dr. Ed Felten

Some CNN visualization tools and techniques

Deep Visualization Toolbox

Github: https://github.com/yosinski/deep-visualization-toolbox

Understanding Image Representations by Inverting Them

Paper: https://arxiv.org/pdf/1412.0035v1.pdf

Learning FRAME Models Using CNN filters

Project page:  http://www.stat.ucla.edu/~yang.lu/project/deepFrame/main.html

Convergent Learning: Do different neural networks learn the same representations?

Github: https://github.com/yixuanli/convergent_learning

Torch-visbox

https://github.com/Aysegul/torch-visbox

Plot caffe models online

http://ethereon.github.io/netscope/#/editor

Grad-CAM: Gradient-weighted Class Activation Mapping

https://github.com/ramprs/grad-cam/

Quiver: Interactive Feature Visualization for Keras

https://github.com/jakebian/quiver

CS231 Stanford notes on Visualization

http://cs231n.github.io/understanding-cnn/

 

The post Some CNN visualization tools and techniques appeared first on A Blog From Human-engineer-being.

Source: Erogol – Some CNN visualization tools and techniques

DataDiving with Marks & Spencer


Running DataDives is part of DataKind’s DNA. However, over the years, we have experimented with different formats and formulas. At DataKind UK, we’ve been partnering with Marks & Spencer, a UK food and clothing retailer, to run internal DataDives with them for three years running.

Similar to other DataDives, the event involves bringing together groups of volunteer data scientists and selected charities to work on data-for-good projects. Unlike other DataDives though, the events are not open to the public. Rather, they are attended by M&S’s internal data analyst community and invitations are extended to some of their suppliers.

On Thursday, October 27th and Friday, October 28th this year, 40 Marks & Spencer data analysts came together to help three fantastic charities: Oasis Community Learning, Shelter, and the Welcome Centre. After much coffee and number crunching, the assembled brain power produced some spectacular results. 

Check out highlights of findings from each project below!

Oasis Community Learning

“…It is the best level of human resources analytics that Oasis Community Learning has ever had and it is great to see the educational impact being clearly mapped to the turnover of our staff.”    
John Barnaby, Chief Operating Officer, Oasis Community Learning

Oasis Community Learning is one of the top three Academy providers in the UK with 47 schools across primary, secondary and 6th form serving 22,000 students with 4,300 teachers. Oasis wanted to look at their human resources data to understand staff turnover and absence, and what this means for pupil performance.

The volunteer analysts found that staff turnover was higher in schools with students that have special educational needs and English as an additional language. They found that primary schools spend twice the amount per day to cover staff absence compared to secondary schools. The analysts also found that primary schools tend to underestimate these absence costs. While these are all provisional findings that require further analysis, the DataDive enabled Oasis to see these patterns in their HR data for the first time and Oasis has accelerated their plans to become more data-driven.

Shelter

“…The DataDive has equipped us with a set of ideas and insights that has helped to clarify which direction to head in to develop a deeper understanding of our clients...”
Dean Robinson, Business Systems & Analysis Manager, Shelter   

Shelter helps millions of people every year struggling with bad housing or homelessness through advice, support and legal services. They wanted to dig into their outcomes data to understand what happened to their clients. What is the result of their help? What are the changes for the client? How do these changes compare for different people accessing different services around the country?

The M&S analysts dived in and, in no time at all, whipped up an interactive dashboard to enable Shelter staff to explore these very questions. The volunteers looked at the number of hours Shelter staff spend delivering services in different parts of the country. They also explored the rate at which cases were dealt with. For example, those over age 65 are more likely to have their cases resolved than those aged between 25 to 34. Shelter’s business systems team was thrilled, and they have even started learning R, a data analytics programming language!

The Welcome Centre

“A very well organised and structured event, the outcomes of which will make a genuine difference to our organisation’s business processes…”
Andrew Tomlinson, Trustee, The Welcome Centre

The Welcome Centre is a food bank in Huddersfield and South Kirklees that supports people experiencing crisis through practical help. They wanted to understand their clients better and identify those who would benefit from additional support, advice and referral to other services. In particular, the Welcome Centre wanted to know who is most likely to become a repeat user of the food bank, as those individuals tend to need extra support.

The all-star team of pro bono analysts got to work and were able to find factors that predicted how likely it was that someone would need extra support. Based on a person’s age, their number of dependents, and the reason for their referral to the foodbank, we can begin to predict the kind of support they will need. The Welcome Centre is looking to develop the model further so they can identify who needs support earlier, what future demand for the service might be, and to test hypotheses for which interventions work best with which clients. 

A huge thank you to Marks & Spencer’s Plan A team and to Pete Williams, Head of Enterprise Analytics at M&S, for driving another successful DataDive. We look forward to next year’s!



Source: DataKind – DataDiving with Marks & Spencer

How to get a data science job

You’ve done it. You just spent months learning how to analyze data and make predictions. You’re now able to go from raw data to well structured insights in a matter of hours. After all that effort, you feel like it’s time to take the next step, and get your first data science job.

Unfortunately for you, this is where the process starts to get much harder. There’s no clear path to go from having data science skills to getting a data science job. You’ll need to put in a lot of hard work to forge your own.

But don’t give up hope! Getting a data science job after learning on your own is very possible. In this post, we’ll discuss the things you should be doing to put yourself in position to start getting data science interviews. In a subsequent post, we’ll cover the interview process itself, and how to prepare.

Afterwards, you might get a snazzy new company laptop!

If you feel like your data science skills aren’t yet well developed enough to start looking for a job, you might want to check out…

Source: Dataquest – How to get a data science job