HNHacker News
TopNewBestAskShowJobs

madenine

596 karma · joined September 2, 2016

Data Scientist© (whatever that means in $CURRENT_YEAR)

Current: CHI Previous: PHL, DC, NY

submissionscomments
madenine··on Azure Jupyter Notebooks
Depending on what you're doing - deploy a webservice with Azure ML?
madenine··on We turned $140k on Kickstarter into $40k debt and broke even
Getting started from 0 with ads on facebook is hard.

If you don't have good email lists, have the money to throw at ads to figure out your ad groups / demographics (just because purple fish people are your #1 customers doesn't mean they buy things off of facebook ads), and don't have a whole ecosystem (landing pages, cross platform re targeting, etc), its an uphill battle that even the pros can't always get working right.

Good on you for trying. Back when I ran ads, a common story from small clients was "I spent $50 - $200 and didn't get any sales, so I figured it wasn't for us" - At least you figured out it wasn't for you, or it wasn't for your budget/time scale.

madenine··on Intel Declares War on GPUs at Disputed HPC, AI Border
> 100X reduction in training time compared to GPUs

Sounds great, when will it..

>Intel loosely laid down plans to eventually integrate the unique Nervana architecture with Xeons this week, but was hesitant to put a year stamp on when that might happen.

So their theoretical eventual integration is hypothetically 100x faster than current-gen GPUs?

If they can deliver, great news for deep learning - but this seems like a lot of speculation.

madenine··on Show HN: Stop the Bullshit
Fake news site are just the latest symptom of the two fold problem of anyone being able to put anything on the internet and people tending to like things that validate their existing views.

If its not news sites, its images with some text over them, videos, etc.

madenine··on Code I’m Still Ashamed Of
> 'I wasnt even manager level but I lost it and yelled at the CTO.'

Bold strategy.

madenine··on How an accused drug dealer revealed Jacksonville's facial recognition network
The term comes from the games community - its what they call it.

For example, in DotA: http://wiki.teamliquid.net/dota2/Pseudo_Random_Distribution

madenine··on Web Scraping with Beautiful Soup (2014)
My web scraping toolkit (Python):

-Beautiful Soup

-Requests Lib

-JSON Lib

-Selenium

-Urllib2

-Cookielib

Handles 99% of things I encounter. With dynamic sites, you're often better off simulating the request than controlling the browser.

Then you get to parse your way through someone's annoyingly formatted JSON.

madenine··on Facebook admits it must do more to stop the spread of misinformation
Their monetezation strategy is advertorial content posts.

People are so used to clickbait as a way to get impressions and clicks from display ads that many don't realize there isn't a single 'traditional' display ad on BuzzFeed's entire site.

They make their money by charging for references to 'advertisers' products and services throughout BuzzFeed content - something they do really well.

Doesn't come cheap either - big cash outlays iirc.

madenine··on The Steve Jobs email that outlined Apple’s strategy a year before his death
I think some formatting got lost. The context:

– we invented Digital Hub concept – PC as hub for all your digital assets – contacts, calendars, bookmarks, photos, music, videos – digital hub (center of our universe) is moving from PC to cloud – PC now just another client alongside iPhone, iPad, iPod touch, ...

Train of thought seems to be, "Appled invented the digital hub, with the PC at the center for all your contacts, calendars...etc - That's moving to the cloud"

Hence why its under the topic "2011: Year of the Cloud"

madenine··on Announcing RStudio v1.0
Check out Rodeo[0] from Yhat. Its not RStudio, but its an IDE with data science in mind. Been getting better and better each release.

[0] http://rodeo.yhat.com/

madenine··on Asus to surpass Apple in notebook sales: TrendForce
I wouldn't discount the ThinkPads so quickly.

I use a T460s for work - its a great machine. I prefer it to my 2013 MBP.

madenine··on Google AI invents its own cryptographic algorithm
On a second look, seems like a combination of Alice not being good at encrypting (in a sense that it had to encrypt in a way that makes decrpytion possible), and Bob needed to get up to speed on decrypting (even with the key).

As soon as that relationship clicks, Bob get better very quickly, while Eve improves at a slower rate for a while, before Alice adjusts (causing bob to get worse for a bit, until it catches up)

Would be interesting to see a set up where Alice / Bob get trained first, then Eve is activated and see how they adjust.

madenine··on Google AI invents its own cryptographic algorithm
Really interesting that for the first ~6500 steps Eve was better at decrpyting the messages than Bob, which had access to the key.

Would be cool to see a setup like this in which Eve is a cohort of decryption tools, otherwise your Alice network will just overfit against its mechanics. Ie, if Eve uses lockpicks to open a door, an Alice solution would eventually replace the key-lock with a number pad - not necessarily any more secure, but will foil Eve every time.

madenine··on Cognitive bias cheat sheet
Great article, but what on earth is the point of the huge wheel chart? Its pretty, but I'm not sure what conclusions it helps me draw, other than sorting sources of bias by group/subgroup in a difficult to read manner.
madenine··on The Strange and Sudden Disappearance of a Coding Bootcamp Founder
My bootcamp was for data science, so not exactly comparable to software/web dev camps - but it cost me about $9k + living expenses.
madenine··on SpaCy v1.0: Deep Learning with custom pipelines and Keras
As someone else in the thread said, spaCy is my 95% solution; its what I use first to get me going.

The biggest benefit you'll see right off the bat is speed. SpaCy is much faster than most other NLP tools out there.

madenine··on The Strange and Sudden Disappearance of a Coding Bootcamp Founder
I did a bootcamp program. In-person, 13 weeks, 9-5, 5 days a week, plus HW/Weekly projects/Capstone project.

If I had locked myself in my apartment could I have learned the same things in the same amount of time? Maybe.

Going through an organization took the legwork out and let me focus on the learning. I didn't have to seek out my own materials (I did anyways, but that was for added depth into things that interested me along the way), I didn't have to trust my ability to be self critical in order to evaluate my progress, I didn't have to prepare exercises and projects, etc.

Plus, I was going through it with 20 other students who were learning the same things at the same time as me - people to learn with, bounce ideas off of, ask different questions in class, etc.

Plus I got a ton of career support. They brought in panels of industry leaders to talk to us, took us to different types of companies to meet their teams and get a good sense of the different types of organizations we could work for. They helped us with our resumes, online presence, interview technique - and hosted 'career day' events where they brought in a staggering number of companies who were looking to hire someone just like us, and helped us with applications, networking etc.

Again, could I have taught myself the same amount of material in 13 weeks solo? Maybe. Would I have walked away with a new job as quickly as I did (less than 10 days after finishing the course), and with a decently sized professional network? Probably not.

On top of that, the physical learning environment was great (coffee, food, etc), plus going to school every day meant it was easy to stay in learning/work mode, instead of getting distracted or procrastinating at home.

madenine··on Bill Belichick Throws in the Tablet
Most people who use/consume mass-market, popular technology have grown accustomed to user-friendly environments and devices that just work right out of the box - so they have an expectation that things will just work.

In my experience, that isn't limited by age - and if anything I see the opposite, with the young being so used to great UX that their concept of technology stops at the GUI.

madenine··on Researchers reach human parity in conversational speech recognition
Its frustrating when people assume any reference to AI means AGI or strong AI.
madenine··on OpenCAPI Unveiled: AMD, IBM, Google, Xilinx, Micron and Mellanox Join Forces
> "NVIDIA is a member of the OpenCAPI consortium, at the "contributor level", which is the same level Xilinx has. The same is true for HPE (HP Enterprise)"

Awesome.

madenine··on A Mexican architect has a vision for a city straddling the U.S.-Mexico border
DC resident here - diagonal boulevards sure looks pretty, but 5+ way intersections are always a mess.

All I saw from the road-map of this city was waaaay to many 6+ way intersections of major (perimeter or axis of triangle/hexagon areas) roads.

Look at the central north/south road, you've got a 6-way intersection every block. No thanks.

madenine··on New data challenges the cultural consensus on public housing
> "However, for black boys growing up in the very poorest projects, the benefits of housing assistance when they entered the labor force were also indistinguishable from zero."

So public housing/vouchers slightly raises income and reduces the likelihood of going prison, except when it doesn't.

madenine··on Microsoft SQL Server 2016 Express
Loving Power BI.

Dashboards, reports, etc are a pain - but its what the clients want. They don't really care what we're doing in the background, they just want their metrics and analytics reports. Power BI has made that part of my job 100x better.

madenine··on Atom 1.11
Only thing keeping me with Atom is Hydrogen[0], which lets me have all the things I like about jupyter notebooks without some of the things I don't like about jupyter notebooks.

I've tried VS Code and liked it - if anyone knows of an equivalent/replacement for Hydrogen in VS Code I'll make the switch.

[0] https://atom.io/packages/hydrogen

madenine··on OpenAI is Using Reddit to Teach An Artificial Intelligence How to Speak
Pretty much.

It uses their P100 HPC cards instead of consumer grade cards (8x P100s), plus two Xeon E5v4 chips, half a TB of RAM and 7.5 TBs of SSD storage - all wrapped up nicely configured for you with their CPU-GPU speed up stack.

I believe the only way to get P100s right now is in the DGX-1, so there's that.

madenine··on Not OK, Google
Hate to be the bearer of bad news, but your "crappy dumb phone" is already telling someone about your every move with a degree of accuracy ranging from an area around the closest cell tower to within a few feet.

The courts are still deciding when/whether that information requires a warrant.

madenine··on Cathy O’Neil on Weapons of Math Destruction
Her PR team seems to have been doing a great job. Can't go more than a day or two without seeing something about this book.
madenine··on Machine Learning 101: An Intro to Utilizing Decision Trees
If you want to do Andrew Ng's ML course, but want to do it in python: https://github.com/icrtiou/coursera-ML
madenine··on Using R to detect fraud at 1M transactions per second [video]
Microsoft has gone all in on R, excited to see where things go. Some cool stuff out of Ignite this week.
madenine··on Using R to detect fraud at 1M transactions per second [video]
doesn't Bloomberg have a custom, in-house R IDE?
← PreviousPage 5 of 6Next →