Deep Learning -Part-2 | How Neural Network Functions? | Deep Learning Applications | Malayalam

[@Mystiq8.i]

Sir after plus 2 AI, deep learning, nlp okke padikkanum, career optionsum collegeum okke onnu paranju tharumo

[@mpowerfuture]

theerchayayum. ithinte oru video cheyyam soon.

[@Mystiq8.i]

@@mpowerfuture sir sirnodu oru karyam chodikkanundayirunnu, virodamillel whatsapp number tharamo plzz🥺, instayil msg ayachitt marupadi illa athanu plzzz

[@mpowerfuture]

@@Mystiq8.i njan insta il theerchayayum reply cheyyum. korachu delay undakumennu mathrame ullu. apologies.

[@okhanio7941]

Ordinary ML has only decision layers, features (color, size etc) should be inputted to it. DL has feature extraction layer also. So actual input (images). DL needs huge data to accurately extract features. Ordinary ML has advantage over DL, that it works for less data also.

[@mpowerfuture]

many thanks for the valuable inputs. this is helpful for many. appreciate.

[@ArunkumarM_AKM]

Wonderful.. Simple and straight explanation. Easy to get the concept. Thank you.

[@mpowerfuture]

many thanks for the kind feedback. appreciate😍

[@shironjaison4484]

Sir very informative video,thank you for that, would like to know more about deep reinforcement learning with double Q- learning and dueling double DQN

[@mpowerfuture]

for this, I need get into more specifics like a course which may not become a generic concept video which i do in this channel. however thanks for the the suggestion and lets see how to do this later. thank you.

[@D...015]

Neuromorphic computing ne kurich detailed aayitt asap oru video cheyyavo sir..plz?

[@mpowerfuture]

athu kurachu technical aavam. let me review. thanks for the msg.

[@anujasuraj4037]

Explained in simple words...👍

[@mpowerfuture]

Thanks a lot 😊

[@nikstatusworld2896]

How is deep learning differ from machine learning? Where I go to study deep learning course?

[@mpowerfuture]

thanks for the Q. the basic differences were mentioned in this video. which is : a. DL uses the algorithms inspired by the structure and functions of brain(learning from the hidden patterns).
b. DL uses neural networks and suitable for handling large amount of data.
c. in ML feature engineering happens manually but in DL feature extraction is automatic.
before choosing the courses, it is important to know the mode of content delivery which is online or onsite, accordingly can suggest few. thanks.

[@REDROSE-be3br]

CNN ne patti oru video cheyyumo?

[@mpowerfuture]

future il for sure.

[@manjulachandrasekhar5361]

Sir please give training on Regression analysis multiple variables for machine learning

[@mpowerfuture]

thanks for the suggestion. unfortunately i don't do course specific videos in this channel. apologies.

[@sreelakshmi7932]

Sir pls make a video about LSTM

[@mpowerfuture]

thanks for the suggestion. will plan for the future releases.

[@jibinpreji]

Relu and sigmoid difference entha? Why relu is used everytime

[@mpowerfuture]

relu is easy to use compared to sigmoid as an activation function. it is faster to compute. Relu just needs to pick max(0, x) and not perform expensive exponential operations as in Sigmoids. hope this helps.

[@mystackquest]

Sigmoid ഒരു hyperbolic tangent function ആണ്. അതായത് ഒരു curve ആണ്. അതിനാൽ തന്നെ hidden layer-ൽ നടക്കുന്ന differentiation-ന്റെ (chain rule apply ചെയ്ത) ഫലമായി ഇടയ്ക്കുള്ള neuron-ന്റെ ഗ്രേഡിയന്റ് പൂജ്യമാകും. അതിനാൽ തന്നെ ഇടയിലെ ഒരു ന്യൂറോണിന്റെ weight സീറോ ആകും. അതു കൊണ്ട് subsequent ആയി വരുന്ന എല്ലാ ന്യൂറോണും weight ഇല്ലാത്ത dead neuron ആകും. കാരണം നമ്മൾ ഈ സീറോ weight വെച്ചാണ് ഫോർവേഡ് and ബാക്ക് propagation സമയത്ത് multiply ചെയ്യുന്നത്. ഈ ഒരു പ്രശ്നത്തെ ആണ് vanishing gradient problem എന്ന് പറയുന്നത്.
ReLU അഥവാ rectified linear unit ഉപയോഗിക്കുമ്പോൾ ഈ പ്രശ്നം പരിഹരിക്കപ്പെടും. എങ്ങനെ എന്നുള്ള മാത്തമാറ്റിക്കൽ explanation ഒരു കമന്റിൽ എഴുതാൻ പരിമിതി ഉണ്ട്. എന്നാൽ ReLU is also not perfect due to its exploding gradient problem.

[@deekshithmenon7195]

Well explained sir

[@mpowerfuture]

Thanks and welcome. keep watching.

[@greeshmaasharaf3617]

Really simplified and informative as well.. Sir please make a video on digital twin technology...

[@mpowerfuture]

Thank you for the feedback. sure will do as soon as possible.

[@fayismahmood8612]

Deep learning il inputs kodukkille??

[@mpowerfuture]

Yes, definitely

[@shahmafidhu5559]

❤❤

[@mpowerfuture]

thank you so much :)

[@lathabala6873]

Sir...

[@mpowerfuture]

yes tell me :)

[@ansiftc936]

Part3 venam

[@mpowerfuture]

sure 😊 will try.