In the last part of the video you said we can fit a known distribution to a continuous set of data. However, you continued to then write that the probabilities can be calculated by taking the product of the pdf evaluated at different values of the feature and label. The pdf does not provide probabilities however, as it needs to be integrated to inform one of the probabilities of an event. This part of the video seems imprecise. However, the video in general was great. Thanks.
9:37 you made conclusion based on P(X=[0,2] | Y), I think the correct way is to calculate P(Y|X=[0,2]). In case P(Y=1) is very small, the answer can be Y=0.
If I search for any ML Algorithm I just first check your channel If you have created the video on the same... You are my first preference for ML/DL Algo Explanation. Just a request please make a video on Deep Learning Algorithm too like CNN, RNN & LSTM "from scratch". It will really help people who want to become practitioners in AI like me.
The explanation is so cool! But it would be even cooler if you added some examples with continious features and fitting a distribution, this part wasn't so clear...
Maybe i'm wrong but I think the hypothesis is not that X1 and X2 are independant but that X1 and X2 are conditionnaly independant. It was very clear otherwise thank you !
I think the hypothesis is that you assume each feature to be (w.r.t other features) 1) globally independent (in the global sample space) 2)conditionally independent w.r.t the occurence of each class label (under the subset sample space where the particular class event has occured) If these assumptions are not met, then it does not seem possible to build the mathematics, because as far as I see, if events A and B are independent, that does not naturally imply conditional independence between events (A|C) and (B|C)
Ok, I've given up on the video after 45 secs. You said "stated clearly", if you hadn't I'd have kept watching. You point to an array of features called X. What are they? Are they features of the array itself (its size / rank / dimension?), are they features of the thing the array of describing (measurements in a house?), or a list of possible attributes (the ingredients on a pizza?) Then you introduce a label. So what, is this like a python dictionary? Plus, I've no idea what sort of issue we're supposed to be tackling? Is it probability? Is it rationality with limited knowledge? I only guess that because I've heard of Bayes before. Instead you launch into calculations when I have not the first idea what you're calculating. Why would I listen to that? Tell you what, I'll give it another 30 secs. If there's no illustrative example / clear explanation of what the hell you're covering I'm gone.