eitlt @methodsoftruth - Tumblr Blog

Day 9.

Still haven’t gotten the original gradient descent to work properly yet. Kind of feel like my values for theta are being updated much too soon; normalization did squat.

Oh well, got a quiz done so at least it wasn't a complete loss.

#why are none of you good at machine learning and can give advice #self inflicted grad school

Day 8. Finally getting to some of the more meaty portions... as if there wasn’t meat enough. But I’m still excited since there will be the inevitable slowdown happening later on. Yay regularization methods~ (~u~)

Day 7.

The review questions and quizzes are easy enough but somehow that entire section on advanced optimization makes me feel temporarily inadequate.

Like, maybe I should do a course on numerical computing or something to compensate.

Day 6~ Sigmoid/Logistic Regression. Different from linear regression since it’s a method with a discrete-valued output.

Forgot to upload yesterday. But hey, it’s up now.

Any blog with any post that begins with “Since you have” immediately gets unfollowed. Too bad I can’t mark as spam.

This is 2019. No time for the bullshit.

Gradient Descent Mk I

No idea why this is wrong. Any insight would be greatly appreciated.

Effectively, the answer for the first value of theta is too big and the value for the second value of theta is too small. This has happened consistently, even with the test cases.

Code:

Answer:

#machine learning #gradient descent #linear regression

Also, I'm not going to do any major progression tomorrow so there won't be any significant updates.

Day 5. Decided to take it a bit easier today and finish up some conceptual materials.

Day 4. Forgot to post this yesterday but focusing a bit more on conceptual understanding.

Day 3.

A linear review and some other stuff dealing with linear regression. Hopefully, I’ll get to post some code soon. However, this too is necessary it seems.

Day 2. Considerably interesting. I knew I’d encounter partial derivatives but didn’t know they’d turn out to be so integral (see what I did there?)

Interestingly enough, this is still the easy part. Haha

Day 1 was actually yesterday but was too tired to scan anything. I suppose I should do my convex notes as well but sometimes I find them much less coherent.

Old undergrad research on chemical thin-films.

#math #mathematics

A small drill involving the powers of 2.

Note: k < 0 is trivially true so it is left as an exercise to the viewer.

#math #mathematics #proofs #proof

There is No “I” in “Team”.

The saying “You don’t have to act like everything is okay if it isn’t” is NOT universally applicable.

Explanation:

In the corporate world, you do, since the upper management will never approach you directly with any concerns or otherwise "perceived conflicts of interest". They would prefer to hang on every word of your direct manager such that it doesn't matter what is actually happening; you're just a number and highly replaceable at the end of the day.

Even if they claim otherwise; it's all just public relations to make themselves feel better.

And yes, we fully understand the argument of "you know how many [numbers] they would have to address if they did that for every [digit] in the [machine]", as well as "it's just not feasible for the sheer number of [cogs] in [the wheel]", amongst many of the other textbook excuses that would be thrown.

We are also fully aware of the "just call human resources" but since that approaches being another report on another pile never to see the light of day, as time t approaches infinity, then re-read the first six words in the first line of the explanation.

As long as the lies will be propagated, then do not call the workers by any deceitful words such "family", "asset", or any other buzzword from that dictionary. The servitude here implying that the alternative is homelessness or otherwise falling into a state of utter disrepair.

On the other hand, now, let's assume that the open set of phrases is not a lie. Choosing, "we're here for you", "you can talk to me", and "if I can't [assist] then I'll find someone who will be able to". Then, if nothing changes over the course of 13 months, in a supposedly fast-paced environment where the unit of measure is in "weeks" *, then

[person using phrases] has been exhausted as a resource and is able to do no more,

[person using phrases] has not done all that they are able, but is unable/unwilling to do any more

[person using phrases] has not done anything at all (clearly contained within 2.)**

[person using phrases] is necessarily capable but insufficiently capable. ***

In any of these instances, it would then imply that the subject's best interest would be to cease interaction and explore other avenues. However, without loss of generality, heavens forbid that the subject chooses that option for then the lies and slander begin once more; since they are clearly unable to be anything but a sore loser even though the thing that they lose was valued as “inconsequential”, or “garbage”, by them in the first place.

*For conversion sake, let's assume that 13 months is 52 weeks. Clearly, this would mean that colloquially 52 weeks = 1 year = 13 months, which in reality is false, but let's be generous; no one wants a 48 week year.

**This case could be extended to [person using phrases] making the active effort to work against any form of actual assistance provided it might be in the subject's best interest as opposed to their own.

***Along the lines of the classic sitcom person fixing the sink but having the lights flicker when the tap is turned, once the task is complete. (or some other equivalently catastrophic result; even approaching the given room being flooded from inadequate capabilities)

I temporarily forgot why I was writing it like a journal so Days 9 and 10 got lumped together.

Also, BOLZANO-WEIERSTRASS!!!

This shit is so handy like you wouldn’t believe.

#math #mathematics #real analysis #calculus

Trending Blogs

Recently Viewed Blogs

eitlt