An update on Signal Data Science (an intensive data science training program)

In De­cem­ber 2015, Robert Cord­well and I cofounded Sig­nal Data Science (web­site), which we an­nounced on Less Wrong.

Our first co­hort has just con­cluded, and over­all went very well. We’re plan­ning an­other one in Berkeley from May 2nd – June 24th. The pro­gram is a good fit for peo­ple who are both ex­cited to learn how to ex­tract in­sights from data sets and look­ing to pre­pare for in­dus­try data sci­ence jobs. If you’re in­ter­ested at­tend­ing the next co­hort, we would love to hear from you. You can ap­ply here, or con­tact us at sig­nal­data­

We offer in­quiry-based learn­ing and an un­usu­ally in­tel­lec­tu­ally cu­ri­ous peer group. Un­like typ­i­cal col­lege classes, Sig­nal Data Science fo­cuses on learn­ing by do­ing. You’ll learn from a com­bi­na­tion of lec­tures, short knowl­edge-re­in­force­ment prob­lems, and longer, more open-ended as­sign­ments fo­cus­ing on an­a­lyz­ing real datasets. (That’s your chance to dis­cover some­thing new!) Don’t worry if that sounds daunt­ing: our in­struc­tors will be there to sup­port you ev­ery step of the way.

You’ll learn both the the­ory and the ap­pli­ca­tion of a wide ar­ray of data sci­ence tech­niques. We offer a pair pro­gram­ming-fo­cused cur­ricu­lum, al­low­ing stu­dents to learn from each other’s strengths. We cover ev­ery­thing from ba­sic lin­ear re­gres­sion to ad­vanced, in­dus­try-rele­vant meth­ods like sup­port vec­tor ma­chines and di­men­sion­al­ity re­duc­tion. You’ll do an ad­vanced, self-di­rected pro­ject at the end of the course. Cu­ri­ous? Check out our show­case of past stu­dents’ fi­nal pro­jects. What­ever your in­ter­ests are—from do­ing some­thing with real-world, in­dus­try-rele­vant ap­pli­ca­bil­ity to ap­ply­ing cut­ting-edge neu­ral nets—we’ll work with you to find a pro­ject to match your in­ter­ests and help you show­case it to prospec­tive em­ploy­ers.

Less Wrong read­ers might be es­pe­cially in­ter­ested by Olivia Schaefer’s pro­ject, which de­scribes re­sults of do­ing some nat­u­ral lan­guage pro­cess­ing on the Less Wrong com­ment cor­pus, ex­plain­ing how the words pic­tured in differ­ent col­ors be­low are at op­po­site ends of an axis.