brohrer/brohrer.github.io

Public web page for brohrer

brohrer.github.io

This is the code for my landing page, a one stop shop just in case anyone is curious about stuff that I write or build.

brohrer.github.io's Issues

Mixed content

You currently have mixed content issues with this repo because of kdnuggets.com, currently their SSL cert is 637 days out of date and needs renewing. Once that's fixed you can change all the http://kdnuggets.com referances to https://kdnuggets.com. For now I'd suggest maybe uploading the images to this repo or using a free service like imgur to host the images.

For example LN64 in blog.html

<img src="http://www.kdnuggets.com/images/top-kdnuggets-blogger-2016-sep-gold.jpg" width=90 height=90 alt="Top KDnuggets Blogger - Gold, September 2016">

Ref: https://twitter.com/OmgImAlexis/status/780592227454156800

Output of NN not normalized

Hi sir!

In your excellent document "how_convolutional_neural_networks_work.html", in the section "Fully connected layers", we show the picture of a fully connected layer with .92 and .51

I guess they are probabilities of labels "X" and "O" to be chosen (92% and 51%) but the sum does not make 100%. So it would have been better to have used normalized values (such as .92 and .08 or .49 and .51). This would have been less beginners confusing (like me when I learnt). In tutorials, the output of NN is always poorly explained so beginners believe outputs are "yes" or "no" but in fact it is 92% of chance to be "X" and 8% of chance to be "O"