Sure. I don't have any holistic survey to prove my point, but an example of recent progress in terms of sample efficiency is this paper[0]. Derivatives of this paper have been used to solve Sudoku[1], Starcraft II[2] and more [3]. This paper enabled more efficient use of data by creating a probabilistic graphical model between logical sets.
Who cares about this garbage if the tool isn't even open source? There are lots of ML deployment tools that are open source. I know haters will downvote my post, but it's the truth. If I can't actually fork and evaluate a tool, it is hyped up garbage to me.
Meanwhile, here is a list of open source ML deployment packages:
i agree, and I recommend doing it for those wishing to get into the field. i believe one of the things that helped me get where i am now is my performance in kaggle competitions.
i do have one concern if it ever became an industry prereq, it becomes a filter for those who have more free time to work on side projects.
That is incorrect. It's more like because equal opportunity is so difficult, people target equal outcome as a second best. I (decidedly liberal) only value equal opportunity. But since opportunity appears to be hereditary, an equal outcome can, over time, equalize opportunity.
It sure would be nice to skip that first step, and I think it's probably not going to turn out effective for a number of reasons. I don't know what would be more effective.
The following is from http://www.salon.com/2015/09/26/how_to_explain_the_kgbs_amaz... and describes the way the Russians implemented SELECT * WHERE CIA FROM EMBASSY_EMPLOYEES: "differences in the way agency officers undercover as diplomats were treated from genuine foreign service officers (FSOs). The pay scale at entry was much higher for a CIA officer; after three to four years abroad a genuine FSO could return home, whereas an agency employee could not; real FSOs had to be recruited between the ages of 21 and 31, whereas this did not apply to an agency officer; only real FSOs had to attend the Institute of Foreign Service for three months before entering the service; naturalized Americans could not become FSOs for at least nine years but they could become agency employees; when agency officers returned home, they did not normally appear in State Department listings; should they appear they were classified as research and planning, research and intelligence, consular or chancery for security affairs; unlike FSOs, agency officers could change their place of work for no apparent reason; their published biographies contained obvious gaps; agency officers could be relocated within the country to which they were posted, FSOs were not; agency officers usually had more than one working foreign language; their cover was usually as a “political” or “consular” official (often vice-consul); internal embassy reorganizations usually left agency personnel untouched, whether their rank, their office space or their telephones; their offices were located in restricted zones within the embassy; they would appear on the streets during the working day using public telephone boxes; they would arrange meetings for the evening, out of town, usually around 7.30 p.m. or 8.00 p.m.; and whereas FSOs had to observe strict rules about attending dinner, agency officers could come and go as they pleased." I read the book. When a CIA agent's cover was blown, the CIA had a spare care and apartment and the agent's replacement needed just that, so they tended to reuse the car and apartment. And wondered why the replacement was then identified so quickly.
So. After that long digression, here comes a hypothesis: Organisations that can keep their mistakes secret, can make themselves seem much more capable than other, similarly large organisations.
We uh treat models as code, but also have NFS shares setup for the storage and GitLab runner talking to a Slurm cluster to run the models. Results and cross validation upload to GitLab. Main thing we haven’t built out yet are performance dashboards for showing improvement across commits, but with the GitLab APIs that’s a script away (currently we do it by hand)
Tell them to become a data cleaning monster - This is a great way to get in at a lower level and get an interview. Bigger teams would appreciate help in that way.
Graphical models[0] & probabilistic programming[1], with the latter making it easier for developers to dive into this growing AI trend. Research in the field for the past decade has been steadily booming with more companies like Microsoft leading the way. I recommend checking out some MOOCs[2] in coursera.
Uber | Software Engineer, UI Platform Open Source | San Francisco | Onsite
As a frontend engineer on the UI Platform team at Uber, you will build the foundation for all web applications at Uber. This team focuses on providing a performant, secure and reliable web ecosystem for all of our users (riders, drivers and our internal operations & logistics teams) through the creation and support of developer tools, systems and frameworks. The team’s main goal is to make Uber’s web engineers productive and its web applications high quality.
Our current tech stack utilizes React.js & Redux, ES2016+, and Node.js. Our design team uses Figma to create user interface designs.
Finally, 90% of finance books you can read at Borders or Barnes & Noble and not need to buy and keep as a reference. Do that, rather than spend hundreds on them, and buy only the ones you think you'll want to reread or keep on hand for reference.
We have a components+properties model, driven in Python, which gets translated to HTML+JS+CSS. We also have a visual editor, and abstractions over a bunch of the client-server stuff that's typically icky on the Web.
The challenge of such a system is that the web platform is so huge and sprawling, and constantly growing - and all of it's written in JS/HTML. So we've made the pragmatic choice to prioritise usability by non-web developers, and open an "escape hatch" for doing layout etc in HTML if you really need to. But most of our users don't need it, and we're constantly working to extend the boundary of what you can do with simple properties and components.
1) If you already have a good grasp of Python, I always advise to start with the AI for Robotics MOOC at Udacity, which is my favorite class of all time. Once that's done, I'd take a look at their Deep Learning classes and the Self-Driving Car Nanodegree.
2) I think it's crucial to get to grips with how the whole stack works, so I always advise to get to grips with a middleware like ROS. Also, don't be afraid to dabble in algorithms (think problems in motion planning, computer vision, etc.)
3) The traditional programs (think PhD programs) create a lot of specialists focused on a single domain, but the industry is in dire need of more generalists. An engineer who is able to dive into any part of the stack is a huge value-add!
[0]https://arxiv.org/abs/1706.01427 [1]https://arxiv.org/abs/1711.08028 [2]https://arxiv.org/abs/1806.01830 [3]https://arxiv.org/abs/1806.01822