[Proposal]: Backprop from scratch

[prakharcode]

• Abstract (2-3 lines) This talk will cover the necessary mathematics and coding to implement backprop from scratch

• Brief Description and Contents to be covered Backpropagation is the most beautiful work of derivatives and mathematics this talk aims to cover up backprop from ground up

• Pre-requisites for the talk Basic mathematics and coding

• Time required for the talk 25-30mins

• Link to slides

• Will you be doing hands-on demo as well? Probably yes

• Link to ipython notebook (if any)

• About yourself

• Are you comfortable if the talk is recorded and uploaded to PyData Delhi's YouTube channel ? Yes

• Any query ?

[manojpandey]

Adding people to review here: @shagunsodhani @Dawny33 @rishikksh20

[shagunsodhani]

Waiting for the slides.

[Dawny33]

Hi. Thanks for proposing. Instead of simply helping people code up backprop, how about explaining backprop in general? [Just an opinion though]

Most people I know tend to not know how backprop works, thanks to the ready-made api's of the ML libs.

[prakharcode]

Acknowledged. I may not be clear enough but I'll explain the math behind it and then, will show the code and how maths works in sync with code. Ill

[prakharcode]

I'll submit the slides in a couple of days!

[prabhant]

Suggestions: Try to make it a general talk(not a theory lecture with some code) covering back prop as well as variations and advances in backprop in recent year, so that it won't be just another talk on backprop and CNN (which is easily available online)

[manojpandey]

@prakharcode ping with slides ? :)