Status Updates From Machine Learning

The author blabbers way too much before getting to the subject matter at hand, forced me to give up.

Complicated

Kivaa lämmittelyä viimeiselle vuodelle yliopistolla tältä erää

This book keeps on getting more and more complicated, literally feel like every concept that’s “trying” to be summarized needs to have a course to explain what the hell it is.

Three important considerations.
-Lots of data does not mean we have appropriate data. This data should contain information about the variable we want to predict.
- The learning algo. must be efficient. It actually about the max performance with min computational power. Sometimes it's important to have light-weighted models.
- The data we used to train a model must be up to date, in some cases the data changing.

Pattern Rec.
I know that before, the book goes with the real basics of machine learning. Since I know these topics before it is kinda boring, and does not meet with my expectations.

He discussing some of the topic in ai

The explanation of implementing XOR by a perception reveals what makes this book unnecessarily difficult to read: “if you have two cases [...] and if your want either to be enough, you need to give each of them a weight larger than the threshold so that the neuron fires when any one of them is true. But when both are true, the overall activation will be twice as high and cannot be less than the threshold.”

I love this definition, very succinct: “Going from disease to symptom is the causal direction - that is what the disease does; going from symptom to disease is diagnostics - that is what the doctor does. In the general case, diagnostics is the inference of hidden factors from observed variables.”

Definitely a good overview for folks wanting general principles. I may not finish this --- I know most of the material. I was mostly reading it to see if it was appropriate to give to my father and a few other folks. Will write a longer review of my thoughts later.