Fine-tuning a Neural Network explained

[@deeplizard]

Machine Learning / Deep Learning Tutorials for Programmers playlist: kzbin.info/aero/PLZbbT5o_s2xq7LwI2y8_QtvuXZedL6tQU Keras Machine Learning / Deep Learning Tutorial playlist: kzbin.info/aero/PLZbbT5o_s2xrwRnXk_yCPtnqqo4_u2YGL

[@mohammadaminhosseinniya7849]

Brief, Clear, Understandable and Complete, without usual youtubers playing round. Thank you!

[@marsrocket]

Your videos are by far the best on KZbin for this kind of thing. Thank you for producing them.

[@deeplizard]

Check out the corresponding blog and other resources for this video at: deeplizard.com/learn/video/5T-iXNNiwIs

[@tymothylim6550]

Thank you very much for this video! The use of an example and the specific explanation on freezing layers really helps me understand how fine-tuning can conceptually be done for similar tasks!

[@sathvikchowdaryveerapaneni1290]

Found the perfect explanation of fine tune.

[@CosmiaNebula]

0:10 intro 2:00 car recognition example

[@THEGASDRIP]

Thanks 4 the video and good info.

[@nawmeerahman8574]

All of your tutorials are amazing.Simply explained.Helps me a lot.

[@RemiDav]

Exactly what I needed to know before starting that finetuning tutorial ! :)

[@CaptainBravo87]

What a great channel ... am surprised it still has less than 100k subs!

[@mohammadmassri2394]

Thank you for this video!

[@codedlAnguage]

Brilliant 🙄🖐️🖐️🖐️🖐️👉💫👉🎶👉📲👉🔌👉❤️👉🙄👉👆👆👆👉🙏👉😃👉😐

[@saqlaingujjar6764]

Thanks for the explanation in a very simple way.

[@AnshumanKumar007]

Thanks for simplifying the topic.

[@richarda1630]

this is so interesting...I can already see the applications for foreign language adaptation, disease classification, drug research...

[@richarda1630]

also are the weights and biases updated per batch? or per epoch?

[@deeplizard]

per batch

[@snehabhapkar7620]

Awesome explanation as always :)

[@fritz-c]

I spotted a slight typo in the article for this video. recofnizing ↓ recognizing I really enjoy your courses so far, by the way. I've stopped and started a few times with studying ML in the past, but this has been a pleasure to go through.

[@deeplizard]

Fixed, thanks Chris! :D

[@aram1rasul751]

I really enjoyed your lectures. it will be grateful if you could go through an example step by step for hand-digit recognition or car or anything else. It will be much better using code from scratch rather than using prepared libraries.

[@JayTheMachine]

best video ever

[@fahnub]

great explanation

[@fahnub]

Came back to revise the concept. Still helpful.

[@gourab_sarker]

Great Content !! Please if you can make videos on transfer learning... Thank You.

[@bindukhadka]

Thank you ✨

[@shaneebkottakkal3491]

very help full video

[@benabdallahrahma8245]

thank u, it's a very useful video

[@mdyeasinarafath4450]

God bless you, Teacher

[@deeplearningpartnership]

Good stuff.

[@ajeetverma3679]

{ "question": "Why do we freeze some of the initial layers of the network in transfer learning", "choices": [ "to use these layers without updating the weights", "to exclude these layers while training", "to update the weights of these layers", "to initialize the weights with zero for these layers " ], "answer": "to use these layers without updating the weights", "creator": "Ajeet Verma", "creationDate": "2021-05-08T09:07:14.299Z" }

[@Waleed-qv8eg]

Great job! Thank you so much!

[@TheAmhdar]

Thanks a lot !

[@behrampatel4872]

Wow I think you just unlocked something . Would this be similar to clip-skip in comfyUI ? Cheers 💯

[@meravchkroun4197]

Thank you for the video! Can you send a link to an example how to do fine tuning?

[@deeplizard]

deeplizard.com/learn/video/oDHpqu52soI

[@hyunwookim4601]

thanks for video

[@benny4013]

you are amazing guys

[@knowhowww]

amazing explanation. thanks a lot.

[@deeplizard]

Thanks, Knowhow! And you're welcome :)

[@coop4476]

For more general neural networks like LLM, would it be possible to fine tune it to basically anything text-based?

[@hamzawi2752]

You are extraordinary :)

[@MRBOBBY8706]

Is this just a substantial amount of the If else then statement application? (Not a coder or anything, just remember some stuff from Computing at U of H) I am interested in machine learning stuff as it pertains to my theory/book I am writing. I am using some stuff from this video, that I will reach out to you about later, as I am doing the same with all others whose content I am referencing. Thank you so much for this simplistic explanation.

[@imsteven3044]

Thank you for your video, i have a question i have a model of speech to text in spanish (i didn't train it) i want to improve the model, can i use fine tuning right?

[@justchill99902]

Thank you :)

[@lingjiefeng3196]

thank you

thanks for this video

[@deeplizard]

You're welcome, Lan!

[@DM-py7pj]

how does one freeze some weights please?

[@gerelbatbatgerel1187]

ty

[@noCOtwo]

after each batch are the weights updated? does each epoch see the same input data? thanks!

[@deeplizard]

Yes, updates occur after each batch. More on this in the blog below. This kind of update is called "mini-batch gradient descent." Each epoch sees the same input data, but with batches of data shuffled in a different order. deeplizard.com/learn/video/U4WB9p6ODjM

[@N-e0N]

I have a question, if I have a model that can detect license plates of cars, would it be feasible or possible even, to fine tune the model to make it able to detect license plates of motorbikes?

[@deeplizard]

Yes

[@tammanakarthik1288]

Could anyone say whether Fine Tuning and Transfer learning are same? If not, then how they are different? Thanks in advance!

[@ozgunsungar9370]

just wow

[@dr.hafizurrahman9374]

Thanks,

[@tammanakarthik1288]

Can anyone say whether Fine tuning and Transfer learning are same?? Thanks in advance!

[@NVworld2507]

its just like inheritance?? where we can aquires the features of parent class

[@deeplizard]

Hey Nilesh - Kind of. With fine-tuning, we take an existing model that has already been trained on a data set, and we "tune" it by modifying layers, parameters, etc. and re-train it on new data. So in a sense, the fine-tuned model is inheriting its parameters from the original model and then being changed incrementally to accommodate its new task/new data.

[@angelleal3005]

How is Fine-tuning related to Transfer Learning ?

[@asifahmed2147]

What is steps_per_ephocs?

[@deeplizard]

The number of batches that it will take to complete one epoch. It is introduced in our blog here: deeplizard.com/learn/video/daovGOlMbT4

[@dennisasamoah2213]

wow

[@LayneSadler]

How is this different from transfer learning? I figured you would be tweaking individual weights based on the title