BEGIN:VCALENDAR VERSION:2.0 PRODID:icalendar-ruby CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/Los_Angeles BEGIN:DAYLIGHT DTSTART:20180311T030000 TZOFFSETFROM:-0800 TZOFFSETTO:-0700 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:PDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20181104T010000 TZOFFSETFROM:-0700 TZOFFSETTO:-0800 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:PST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20261021T133648Z UID:mskogQ DTSTART;VALUE=DATE:20181010 DTEND;VALUE=DATE:20181011 CLASS:PUBLIC CREATED:20180919T182445 DESCRIPTION: On October 10\, join ~180 devs at SF Python's presentation nig ht and learn more about how you can use Python to run gradient descent and lots more! \n\n If you'd like to present a 5 mins Lightning talk or a 10- 15 mins Short Talk at this or future meetup\, please submit your talk idea s here. \n\n Our generous sponsor Yelp will also provide pizza and drinks for this evening. \n\n PROGRAM \n\n Lightning talks \n\n \n Mehrdad Yazda ni - Winning a guessing game with Python and Scikit-learn \n Tim Ansell - FuPy - FPGA MicroPython - Making FPGA SoC creation and usage easy! \n \n\n Short talk(~10 mins + Q&\;A) \n\n Aaron Wiegel - The tyranny of the no ise in custom manufacturing \n\n In an automated custom manufacturing envi ronment\, the variance of the process can often make it challenging to des ign experiments to improve the performance of the factory. Namely\, the ri ght risk balance needs to be found between the relatively high cost of exp eriments and potential false negatives. I explain how the data science tea m at Synthego uses statistical simulations in Python to aid in design of e xperiments with an appropriate frequency of dog memes. \n\n Main talk (~25 mins + Q&\;A) \n\n Michael Stewart - Gradient Descent\, Demystified \n \n Speaker Bio \n\n Stu (Michael Stewart) is a machine learning engineer a t Opendoor in San Francisco. Previously\, he was a data scientist and engi neer at Uber\, and an economic researcher at the Federal Reserve Bank of N ew York. He is a National Science Foundation Graduate Research Fellowship Awardee in Economic Sciences. \n\n Abstract \n\n The audience and I will w alk through a live coding practicum (in a RISE Jupyter Notebook slideshow) in which I implement an initial gradient descent algorithm for logistic a nd linear regression\, demonstrating the flexibility of the optimization t echnique and the decidedly un-scary code required to get our prototype up- and-running. \n\n Details \n\n Gradient descent (GD) is a fundamental opti mization algorithm that sounds much scarier than it is. Many users of Scik it-learn et al. can apply GD through these tools\, but do not grok what GD is really doing. Other more engineering-oriented practitioners are put of f entirely by the seeming complexity. I walk through a live coding practic um (in a RISE Jupyter Notebook slideshow) in which I implement an initial gradient descent algorithm for logistic and linear regression\, demonstrat ing the flexibility of the optimization technique and the decidedly un-sca ry code required to get our prototype up-and-running. I compare the result s of our hand coded algorithm to those generated by Scikit-learn (and the closed-form normal equation\, for linear regression) and show equality. \ n\n The focus of this talk is on the practicum of implementation one’s o wn GD algorithm\, though I review the most important mathematical and theo retical components of GD to ground the practicum for attendees. Mathematic al review touches on the nature of gradients\, what they are\, how they re late to derivates\, and how they enable iterative optimization over a para meter space. This talk does not include a formal derivation of\, eg\, the loss functions used in linear and logistic regression\, nor does it requir e mastery of calculus. Attendees will leave the talk with a better underst anding of iterative optimization and a template of their own for implement ing GD in Python\, should they feel this would enrich their understanding. \n\n The live-coding setup has two major advantages over static slides + disquisition. First\, it is more engaging to participants. Second\, it rei nforces the central tenant of the talk: that GD is much much less intimida ting than it seems\, and indeed is implementable in quite few lines of cod e. \n\n Many tutorials on gradient descent (1) make only a cursory nod tow ards the underlying math\, or (2) start with something tangentially relate d to — and far removed from — the algorithm\, eg\, an MLE derivation o f the logistic regression loss function. But the fundamental iterative nat ure of the algorithm is intuitive and unfussy: by starting with this itera tive intuition before fleshing out how one arrives at this mechanism parti cipants should be able to understand the mathematical components of the al gorithm in broader context\, rather than in a vacuum. \n\n Practicum parti cipants are aware that Iterative solutions are often computationally expen sive. I review\, also\, closed-form solutions (the normal equation) to lin ear regression and explain the benefits of GD over the seemingly more stra ight-forward closed-form model. Perhaps the main reason that gradient desc ent remains relevant in industry is its flexibility: the second largest po int of emphasis (after the compactness and elegance of the algorithm) is t he clear parallel between the GD implementation for linear and logistic re gression. \n\n Seeing is believing. A final benefit of live-coding GD is that we can compare our hand coded algorithm with the implementation users know and love from Scikit-learn. Scitkit-learn source code can be inscrut able to those note well-versed in object oriented python: by showing (roug h) algorithmic equivalence participants may better convince themselves tha t directly inspecting Scikit source is within reach. \n\n AGENDA \n\n 6:00 p - Check-in and mingle\, with food provided by our generous sponsor! \n\n 7:05p - Welcome \n\n 7:30p - Door close \n\n 7:10p - Announcements\, ligh tning talks and main talk \n\n 8:15p - More mingling \n\n 9:30p - Hard sto p \n\n SF Python is run by volunteers aiming to foster the Python communit y in the Bay Area. Please consider making a donation to SF Python and sayi ng a big thank you to Yelp for providing pizza\, beer\, and the venue for this Wednesday's meetup. \n\n Yelp sees 89 million mobile users and 79 mil lion desktop users every month. Keeping everything running smoothly requir es the best and brightest in the industry. Their engineers come from diver se technical backgrounds and value digital craftsmanship\, open-source\, a nd creative problem-solving. They write tests\, review code\, and push mu ltiple times a day. Come out and talk to them. \n LAST-MODIFIED:20240728T211358 LOCATION:Yelp\, 140 New Montgomery Street\, San Francisco\, CA ORGANIZER:mailto:grace@pybay.com SUMMARY:Live coding gradient descent in Python and more! URL;VALUE=URI:https://ti.to/sf-python/oct-2018-presentation-night URL;VALUE=URI:https://ti.to/sf-python/oct-2018-presentation-night END:VEVENT BEGIN:VEVENT DTSTAMP:20261021T133648Z UID:mskogQ1022811 DTSTART;TZID=America/Los_Angeles:20181010T180000 DTEND;TZID=America/Los_Angeles:20181010T213000 CLASS:PUBLIC CREATED:20180919T182454 LAST-MODIFIED:20181009T123441 LOCATION:Yelp \, 140 New Montgomery\, San Francisco\, CA ORGANIZER:mailto:grace@pybay.com SUMMARY:[Live coding gradient descent in Python and more!] SF Python Meetup URL;VALUE=URI:https://ti.to/sf-python/oct-2018-presentation-night URL;VALUE=URI:https://ti.to/sf-python/oct-2018-presentation-night END:VEVENT END:VCALENDAR