Tips Tricks 15 - Understanding Binary Cross-Entropy loss

[@rohitdhakad-k5s]

I have watched only 4 minutes till now but within that time I understood that the video is going to be interesting. greak work 🔥

[@blackswann9555]

Brillant explainer! I really appreciate your teaching style.

[@ankitvaghasiya3789]

great explanation with using "entropy" term, so easy to understand

[@ishanaphale6373]

Thank you! Going through your videos has been an immense help for me in securing an internship!

[@DigitalSreeni]

Glad to know and happy to help. Good luck with your internship.

[@pbberlin]

I have been watching episodes 73ff and 204ff - $2 to express my appreciation

[@DigitalSreeni]

Thanks for your appreciation Peter. I hope you'll find other videos also to be useful / educational.

[@adriandip8448]

Thanks Sreeni! Crystal-clear explanation that doesn't dumb things down -- or make them hard to understand for the sake of showing it's a hard problem.

[@ksgfl]

감사합니다.

[@DigitalSreeni]

Thank you very much for being generous.

[@sandips700]

Thanks

[@DigitalSreeni]

Thank you very much.

[@viveksthanam]

Thanks!

[@DigitalSreeni]

Thank you very much Vivek. Please keep watching...

[@wessauder7708]

Thank you for such a well presented example of the calculation! Really made the concept easier to understand.

[@pablouribe1522]

Great video. I needed a refresh in the concept and it was perfect!

[@DigitalSreeni]

Glad it was helpful!

[@jeehunkang5648]

Amazing. video. Your explanations are so thorough and on to the point!

[@DigitalSreeni]

Glad you think so!

[@manuelpopp1687]

Danke!

[@DigitalSreeni]

Many thanks Manuel for being very generous. Please keep watching.

[@manuelpopp1687]

Thank you for producing these educational videos!

[@handsome_seagull]

Awesome explanation, you were my hero for today :D

[@noorabentaher1295]

You are doing a great job. Thank you very much for your efforts.

[@DigitalSreeni]

Thanks and welcome

[@harisjoseph117]

Great explanation. Thank you for making it simple.

[@rameshwarsingh5859]

sreeni sir ,thanks ✨✨✨✨

[@AnasHawasli]

Thank you for this video really helpful

[@kevian182]

Excellent tutorial! Thank you!

[@DigitalSreeni]

You're very welcome!

[@bruh334]

Thank you soooo very much.

[@sann6688]

That was really useful! Thanks a lot

[@inteligenciamilgrau]

Really good!!

[@DigitalSreeni]

Glad you think so!

[@ExV6120]

Thank you! please make videos about GAN for medical image segmentation sir

[@stoilacindy9730]

what is ti in the equation

[@shaelanderchauhan1963]

Do we use it in Yolo object detection?

[@herhappykitchen]

your videos are really

[@MohamedAbomokh]

Thank you, much needed video :)

[@DigitalSreeni]

You're welcome 😊

[@3DComputing]

As always, why look at another channel to understand these things. Thanks for the plain English

[@DigitalSreeni]

My pleasure!

[@georgemiller6010]

Hey, great stuff! Im working with some of your earlier code from U-Net from video 207. Im following along but using personnel imagery that is RGB instead of the Mitochondria grayscale. How do I set the channels to 3? And how when I select a random image, how do I display it as RGB, as opposed to cmap='gray'

[@proatemp211]

Thanks Dr from amazing stuff. By the way, I am interested in cell segmentation in NAFLD case, using mask-rcnn, please direct me to any of your already existing video on your channel. Respect from JP.

[@gustavojuantorena]

Great video! Thank you

[@DigitalSreeni]

Glad you liked it!

[@theoverseer1289]

Thanks for the video! I've been following your channel for a while and it has been really helpful.
I have a question about a binary cross-entropy problem i'm into: what means that the loss function decreases to 0.07 more less but accuracy doesn't increase over 0.4??
It have something to do with the data??? Sorry, i'm just new in deep learning. Thanks in advance!

[@nathalieroos1999]

Hi, it could be possible that accuracy is not the best metric for your problem :)

[@theoverseer1289]

@@nathalieroos1999 Thanks! And how is that? What can be those metrics?

[@nathalieroos1999]

@@theoverseer1289 like for example in the problem I'm dealing with I have a big class imbalance, I only have a few pixels labelled 1 in a batch, so the accuracy might be high but that might be because it is very good in predicting the 0 label (from which they are a lot more), so I might use the mean IoU or true positives as a metric. If I were you I would just play around with some metrics and see if they improve when your loss function decreases 😊

[@akashprabhakar6353]

👌

[@bahaya7123]

bless u

[@siddharthct4799]

Great video

[@DigitalSreeni]

Thanks!

[@李白-i8m]

thank you!

[@suryatej839]

u r totally amazing!!

[@DigitalSreeni]

Thank you so much!!

[@EB3103]

awesome!

[@ncroc]

If y's are known and only p's are unknown then this is maximum likelihood.

[@amiroacid]

Thanks!

[@DigitalSreeni]

Thank you very much.

[@jurischaber6935]

Danke!

[@DigitalSreeni]

Thank you very much for your kind contribution. Danke Schön