www.theparticle.com
Main
Services
Products
Resources
Forum
About Us
Contact
Particle Revelation
Hardware Destruction
Quotes
Humor [alpha]
Murphy's Laws
Programming
Java
Java Data Structures
C# Data Structures
Database Design
Graphics Tutorial
Artificial Intelligence
Downloads
SQLRunner
Graphics Tutorials
Hacking Tutorials
Java Applets
MIDI Music
Gov & Misc Docs
Games
Chess Game
Asteroids
TicTacToe
Tetris
Applets
DRAW!
FlightBox
pWobble 3D
pRunner
NYU HWs
Swarms
Geometry
Chaos
Machine Learning
Academic
CISC 7700X
CISC 7510X
Net (old)
OS (old)
AI (old)
App Dev (old)
C++ (old)
OOP (old)
Web (old)
Perl (old)
DBMS (old)
Perl (old)
ProgLangs (old)
PHP (old)
MltMedia (old)
Oracle (old)
Misc
Privacy Policy
Publications
profphreak.com




Welcome to www.theparticle.com.
It's the newest preIPO dot bomb that's taking the world by storm.
Now is a perfect time to buy lots of worthless and overpriced shares! 

Internet is becoming more and more polluted with
junkmail, people selling crap, and businesses which don't know their place on the net.
They're all trying to make this wonderful place (i.e.: the net) in to hell (i.e.: real
world). Internet should be viewed as a place of imagination, creativity, and most of all:
fun. Internet is not some really advanced tool for searching for people to ripoff. It's
about searching, and finding, things which are useful, helpful, and promote the sharing of
ideas. This is what this site is striving to become.  
News, Updates, & Rants...

Happy Halloween!!!
The future is not set. It doesn't exist until it happens. We remember the past, we don't remember the future. Yet past somehow seems to exist, we can refer back to it, we can look at and study historical events, etc. But is there really that much difference between the past and the future?
Whatever we know about the past is actually a recording. Unless we have something memorized in the present, then we don't know about it from the past. If there was an event in the past that nobody recorded, then it's just as mysterious to us as the event happening in the future. We can guess at the probabilities, etc., but in studying the past, only the present matters.
Whenever a recording device is involved, there's some bitratespeed and accuracy with which it can record information. Similarly, there's a retrieval rate. The univesal recording device cannot possibly be recoding everything: it is very observer centric: unless an observation was made, then the event could've happened any possible way.
For example, an electron going through the doubleslit experiment: if we record which way it goes, we have made an observation and a recording. We can refer back to that path later by saying ``yes, the electron went through the left slit''. But if we don't make a recording of which slit the electron went through, then the electron actually goes through all possible paths.
At a later stage, we cannot refer back to history and say ``ah, but if we look at the history (refer to the past), then the electron went through the left slit'', because there's no history, and we don't have a recording of it in the present... So as far as ``history'' is concerned, the electron went through all possible paths.
The same applies to the future. If we send an electron through a double slit experiment and explicitly don't look/make observations/recordings, then it appears to actually go through both slits. If we setup the experiment to check which way it goes, then it goes through the slit we observe, randomly of course. If it wasn't random, then there would be no point in recording the information (it would not be `information').
What would the universe without observations look like? Imagine if nothing was ``recorded'' in the present. Time would appear to stand still. In fact, it wouldn't ``appear'' anything, it just wouldn't be anything. So for time to move, events need to be observed and recorded.
Lets look at the dead cat in the box. A cat is locked in a box with some poison. A random event (truly quantumrandom) may release the poison. Now, in our macro world, the cat is either dead OR alive. We don't know which one until we open the box. In the quantum world, the cat is "both" dead AND alive (it has yettobeobserved histories of it dead AND alive). There's no information yet. If the cat behaved like the electron flying through the left or right slit, then technically it would be flying through "both", which is hard to imagine for a macroscopic object like a cat in a box.
But what's the process by which the past gets recorded and gets turned into the present? How many bits of information can the present hold? (we've already seen that we cannot record ``everything''). Is that number of bits constant, or does it expand as we make observations?
If it's constant, then making leftorright slit observation must ``forget'' some recorded bit somewhere. Perhaps the universe leaks bits on the edges (or black holes?).
 Alex; 20181031
October 31st at wikipedia...  

Below are my views and do not reflect the opinions of anyone else:
Valuations: there are several ways of evaluating investments. For example, bonds give us a very direct way of evaluation: a bond pays us a well defined interestso there are only three things we have to worry about: whether the bond will eventually be repaid, interest rates, and inflation.
In case of government (and many municipal) bonds, we don't need to worry about default, and in many cases inflation is correlated with interest rates, so there's only one thing we really need to worry about. High future interest rates means that our bonds lose value and vice versa.
If we expect interest rates to go higher, locking in a rate via a bond is a bad idea (on the flip side, borrowing at a low rate is generaly a good idea).
Bonds of corporations are best avoided altogether. Ask yourself, if you had a company and you needed to raise capital, which one would you issue, bonds or stock? The mathematics are pretty straight forward: if you expect to earn a bigger return than the interest you pay, you issue bonds, otherwise you issue stock.
In other words, if the company is sound, and they're issuing bonds, you'd make a better return from stocks. If the company isn't sound and is issuing stock, then you'd better avoid the stocks of such a comapny. Never buy bonds (or stock) if you believe the company is on its last leg. (unless you think they're a buyout target, or if liquidation of the company is expected to bring in more than the price).
For stocks, the evaluation is a bit tricky.
We can take the bond approach: treat equity as a contract that pays us some ``interest'' (dividend) every so often until the end of time. How much are those infinite dividends worth?
Money today is worth more than money tomorrow. Two in the bush vs one in the hand.
So we pick some reasonable discount rate (often approximately the current interest rate), and discount all the future dividends by that rate until we hit zero. The sum of all such payments is what the contract is ``worth'' (this is essentialy how much gain we expect to get out of the contract over its entire lifetime). Note that there's no principle in this scenariothere's just dividends, and nothing else. So that's the only thing we use to evaluate the worth of a stock certificate.
For example, if a company pays us $1 in dividends this year, and we expect that to continue, and we also expect interet rates to be about 3% or so (which is historically very low), we can say that each share is worth about: $30.
Obviously the price of each share could be $45 in the market. Or it could be $25. The $30 number just says if you look at dividend alone, you really shouldn't pay more than that. In this scheme, you ignore the fact that you could potentially sell the share for $45 next year: the basic idea is that if you owned the share for a lifetime, what financial benefit you would get from the company because of said ownership.
Many companies don't pay dividends. We could say such comapnies are worth exactly $0 (as we don't benefit from being the shareholders).
We could also use a proxy measure: earnings. Assuming earnings are reinvested and build up value of the company, each share should gain by the amount of the earnings. Discounted into the future of couse.
For example, if our company made $2 per share, how much would we expect such earnings to be worth? Lets assume 5% discount rate, or about $40 price.
Why did we use 5% for earnings and 3% for dividends? Guesses. But generally we should discount earnings at a higher rate than dividends, as money in our pocket is worth more than money in someone else's pocket. (money that we control is worth exactly what it is, but if some greedy/incompetent manager is in charge of our money, it's worth much less than the actual dollar amount).
In the above, valuation we ended up with P/E of 20. (price/earnings, 40/2).
Let's pretend that we had almost a decade (from 2008 til 2017) of extraordinary low rates (01%), and the discount that many analysts used was 1%. That company that earned 2$ a share would ``demand'' a $200 share price, or P/E in the 50100 range.
Then Fed steps in and tightens the rates to 2%, now our valuation should point towards a $100/share price. But the intersia in the market keps markets afloat at $200.
The Fed tightens things a bit more: rate is 3%. Now that $2 earnings demands only $66 or so share price. The market is still up: the stock is still priced at $200.
Let's put this in perspective: if we buy a government bond at 3% (the 10 year treasury bond), we'd expect a higher return on our money than buying that $2 earnings for $200.
Yes, earnings could come up. But they could also come down. If a comapny has been earning $2 a year for a while, bumping that up to $4 is hard. Many companies are already optimize and squeezed quite a bit... bot pulling out more earnings out of the same process is difficult.
What happens when rates go to 4%?
That's why the market panics every time the Fed bumps up the rate number. Nobody wants to be the last one holding those shares at $200 when they fall back to $40 or so.
Revenue is another thing folks have used to evaluate stocks. But we got to be careful: revenue isn't the same as profit. High revenue could mean high losses.
Large enterprises with no earnings (and no dividends) is another thing to look out for. Yes, the company could be putting everything its got back into research and development (which often increases company value)or they could be lavishly rewarding the management or funding gonowhere projects or badly concieved aquisitions.
 Alex; 20181029
 

Decided to rebuild the desktop. Tentatively thinking of AMD Threadripper as the CPU.
 Alex; 20181028
 

My primary desktop quit. It shutdown, and now no indication of life when I press the power button. The board is getting power (there's a small LED on the board, but that's about it. No fan. No disk spin, nothing. Seems completely dead :/
 Alex; 20181007
 

Happy BDay to yours truly :D
 Alex; 20181005
 

Finished reading The Book of Why: The New Science of Cause and Effect by Judea Pearl and Dana Mackenzie. This is definitely an interesting book to read. Not sure I agree with everything in it, but it's worth putting this on your toread list (if you're into such things).
Basically the language of statistics doesn't include causation. If two things appear to happen together, or one after the other, no matter how closely, there's no way to precisely state that one causes the other using the language of statistics. Correlation does not imply causation. From the observational data perspective, we just get points in N dimensionsand just from that data alone it's impossible to say which dimension causes the other (all we can say is how values are correlated).
Anyway, the book makes an attempt at defining causation. Essentially there's a concept of conditional probability: P(YX), which is probability of Y given that X has happened. This does not in any way imply that X causes Y, in fact this could be flipped using the Bayes rule. What the book defines is the "do" operator. In other words: P(Ydo(X)), which is the probability of Y given that we "do" X (we force it, we don't just let nature run its course).
The situations where P(YX) != P(Ydo(X)) are exactly the instances where doing X causes Y. Counterfactually, not doing X wouldn't "cause" Y.
The book doesn't clearly define the "do" operator, and I'm not sure if that's intentional or not (it's not a techy book). The do operator is explained in words as: we fix the value of X, while leaving everything else run their coursethat way any influence on Y from fixing X would be only due to fixing X and not some confounded variable.
The book is full of very interesting examples and stories regarding misuse of causation vs correlation. It's also full of author's opinions, which sometimes do become a drag on the text (and mostly explains why some reviewers gave this book low marks). The author often comes off as I'm so clever and great, and my approach solves problems that all these other researchers completely screwed up. Or something like that. The tone of the book starts out OK, but gets like that someplace in the middle.
All that said, we're assuming that there really is such a thing as causation, which isn't supported by all the physical laws, quantum mechanics, nor anything in mathematical language. Relativity just needs a fourdimensional spacetime, but as for causeeffect things can just happen in any ordereverything is just a curve in spacetime.
 Alex; 20181002
 

Happy Birthday... you know who you are.
 Alex; 20181001
 

Visiting the liberty bell :)
 Alex; 20180930
 

Driving out to Philadelphiagonna see the liberty bell :)
 Alex; 20180929
 

Cut my finger while shaving :/. The blade slipped out of my hand, and while trying to catch it (yeah, wasn't a smart idea), sliced a chunk of my finger off.
 Alex; 20180923
 

Ahoy, maties! International Talk Like a Pirate Day! Arr...
In other news, did an MRI of my knee. So for the MRI, they asked, ``what music would you like to listen to'', and I just said anything, so they gave me headphones with classic music... but then that machine went clunkclunk, chungchung, dangdang, bangbang, clunkclunk, etc., for about 40 minutes. Can barely hear anything. If I ever get one of these things again, I'll ask for heavy metal.
 Alex; 20180919
 

Resubscribed to World of Warcraft for a bit (preordered the expansion, but haven't logged in yet).
 Alex; 20180915
 

Finished reading Measurement by Paul Lockhart. This is one amazing book. Wish I read it during my high school years.
 Alex; 20180912
 

9/11: It's that time of the year again.
 Alex; 20180911
 

Went to doc for the knee. Did Xray, and ``nothing serious''. Still hurts though. Scheduling MRI :/
 Alex; 20180831
 

Think last week's hike hurt my knee. It's been bothering me since the hike, today more than before :/
 Alex; 20180824
 

Crazy thoughts: been thinking about connection between ntuple method and quantum mechanics. If we make n large (in ntuple) then we get more precise results, but less generalization. If we make n=1, then we get average behavior. The link to quantum mechanics is that if we're only working with one quantum system, we get average behavior, but if we're working with many particles, we get some sensible predictable power. (e.g. we can predict how masses will fall down to earth because they're made up of many many particles... but at individual atom level that's hard... same for ntuple method, we can do prediction based on a large ntuple, but if n=1 then prediction only goes as far as overall average).
 Alex; 20180823
 

Hicking up Mnt.Washington. 3hours up via BootSpur trail, 1hour on top (loop around Lake, and upto summit building, and 2hours down via Lion's Head trail.
Much of the Tuckerman trail is actually closed, and there's a detour via a VERY steep road that just seems to go up up up (or downdowndown). It's shorter than the trail, but it's much harder to walk down on.
 Alex; 20180819
 

Driving out to Mnt.Washington.
The mountain is all foggy. Drove upto summit, and there's no view. Very cold and wet :/
At the fancy hotel, Suneli found a bullet under one of the shelves.
 Alex; 20180818
 

Finished reading The ThreeBody Problem by Cixin Liu, Ken Liu (Translator). This book is... well... different. It's very very good.
Sometimes a book has quite unique concepts and a story. Kind of like Einstein's Bridge by John Cramer. The Three Body problem is kind of similar... very deep and unique SciFi.
Definitely recommend.
 Alex; 20180817
 

Kiddo doc appointment day. Leaving early for the 8am appointment. As merging into LIE, right past exit 19, was changing lanes, checked the side mirror and blind spot, and at that instant the car ahead of me suddenly stopped.
Noticed it a bit too late. Slammed on break, and still rolled into the back. My front bumper got cracked, and the other car's metal bumper protector got damaged (no damage to the car itself). My bumper got fixed after a kick (it got a scratch on it now, but besides that, it's all good).
Pretty scary experience. Blocked two lanes on the LIE that morning (until it became apparent that this wasn't anything major). The other driver didn't speak english, so didn't understand me when I was asking whether I should call emergency. The cops never showed up. A passingby state trooper stopped traffic and let both of us drive onto the shoulder.
Missed kid's doc pointment :/
 Alex; 20180815
 

...and back in New York :)
 Alex; 20180806
 

Back to Ambala, and then drive to Delhi airport.
 Alex; 20180805
 

Feeding cows at a cow sanctuary :)
Getting to Noor Mahal Hotel, and enjoying swimming pool.
 Alex; 20180804
 

Taking morning commuter train from Ambala Cantt to Chandigarh.
Visiting Chandigarh rose garden, and having cheesecake at Elante mall.
 Alex; 20180803
 

Morning minihike to Jakhoo temple.
Driving back to Ambala.
 Alex; 20180801
 

Driving to Shimla.
 Alex; 20180731
 

Landed in Delhi. Driving to Ambala.
 Alex; 20180728
 

Flying out to India.
 Alex; 20180727
 

Does Microsoft Surface Go tablet come with a prorprietary charger? (e.g. does it use USBC for power like every other modern tablet these days, or does it require its own proprietary power adapter?).
In other news, a friend flying on Hawaiian airlines discovered that I'm mentioned by name in an inflight magazine, in a story about Mauna Loa. ``Hona Hou!'' magazine, June/July 2018, Volume 21, No. 3. page 90 :)
 Alex; 20180724
 

Given these numbers:
[50, 16, 5, 54, 1, 92, 42, 20, 63, 19, 29, 17, 75, 98, 99, 83, 84, 69, 8, 27, 71, 65, 30, 73, 69, 54, 37, 63, 46, 85, 54, 45, 36, 35, 93, 14, 26, 91, 22, 23, 0, 12, 58, 98, 34, 75, 51, 44, 61, 56]
The mean is: 49.44.
The mean of the first 20 is: 47.550.
The mean of the last 20 is: 48.238.
Mean of every even index is: 45.76, and odd indexes 53.12.
If we randomly pick out 20 numbers, what do you think the mean will be? Yep, about 50ish.
With that in mind, we can estimate any integral (sum) by randoly sampling a few numbers.
 Alex; 20180720
 

Future Value = Present Value * (1 + rate)
If rate is 0.045 a year (4.5%), then the future value of $100 in a year is $104.50.
It works in reverse too. Present Value = Future Value / (1+rate).
A $100 a year from now is worth: 100/(1+0.045) = $95.694
If we raise (1+rate) to a power, we can calculate it over many years. e.g. A promise to pay $1000 in 18 years at 4.5% a year is worth: 1000/(1+0.045)^18 = $452.80 in today's dollars.
What rate would we need to turn $100 into $1000 over 18 years? We can solve for rate:
rate = (Future Value / Present Value)^(1/years)  1
or (1000/100)^(1/18)  1 = 0.13646, or 13.65%
We can check that, 100*(1+ 0.13646)^18 = $999.94
How many years would we need to get $100000 if we start with $1000 and expect a rate of 8% a year?
years = (log(future value)  log(present value)) / log(1+rate)
or (log(100000)  log(1000)) / log(1+0.08) = 59.838 years.
So far, we looked at interest that's compounded annually (the interest is added at the end of the year). What if it's compounded 4 times a year? Then the future value of $100 a year from now (at 4.5% interest) is:
100 * (1+0.045/4)^4 = 104.58, or
Present Value = Future Value / (1+rate/n)^n
where n is the number of compounding intervals.
Notice that 104.58 is greater than 104.50 we got before. So more frequent compounding makes more money. What if we compound continuously... like every nanosecond... (let n go to infinity). We get:
Future value = Present value * exp(rate * time)
where time is in years.
The future value of $100 a year from now (at 4.5% interest) is then
100 * exp(0.045*1) = 104.60
This works in reverse too... What's the present value of $5000 five years form now?
5000 / exp(0.045 * 5) = $3992.6
 Alex; 20180719
 

Berkshire has lifted restrictions on what they'd consider a fair price to buy back shares. Stock poped by 5%. Before, the limit was 1.2 book value: if share price fell below 1.2 times book value, the company would buy back their own shares. Now the limit is anyone's guess; perhaps 1.5 book value?
In either case, a company buying back shares is pretty ugly business. On one end, if the executives are doing it because they think they're cheap, then they're taking advantage of the shareholders (by using their own money to buy them out).
On the flip side, if executives are doing it irrelevant of the price (e.g. 1.5 book value maybe), then they're disadvantaging the existing shareholders buy overpaying the infated share prices.
So either case, the company is acting bad towards its shareholders.
With that in mind, it's more likely the company is buying them back because they're underpriced...
 Alex; 20180718
 

If we wanted to multiply two 2 digits numbers, AB * CD, we could start by changing the representation of the numbers... e.g. AB is actually A*10+B, and CD is actually C*10+D.
So to multiply AB*BC, the result is actually
(A*10+B) * (C*10+D) =
A*C*100 + A*D*10 + B*C*10 + B*D =
A*C*100 + 10*(A*D + B*C) + B*D =
x*100 + y*10 + z
where x=A*C, y=A*D+B*C, and z=B*D
There are several ways of computing A*D+B*C. First observe that
x+y+z = A*C + A*D+ B*C + B*D = (A+B)*(C+D)
So to get A*D+B*C we can just take (A*C + A*D+ B*C + B*D)  A*C  B*D, leaving only A*D+ B*C. We already have A*C and B*D calculated, as x, and z respectively. In other words:
y = A*D+B*C = (A+B)*(C+D)xz
The A*D+B*C has two multiplications, while (A+B)*(C+D)xz has one multiplication. So the total AB*CD multiplication can be done with 3 multiplications, and not 4 as in grade school algorithm.
And this can be used recursively for numbers that are bigger than 2 digits.
The above is known as the Karatsuba algoritm, and apparently it was invented in 1960.
 Alex; 20180717
 

If computer programming can be absracted away into implementing predicates, then every program is essentially defining a set of numbers.
In other words, { x  is_even(x) } defines a set of even numbers. There's a predicate is_even that defines this set. In previous rant (7/13) we've seen that any program (particularly those that return a numeric result) can be implemented as one that outputs a 0 or 1 value.
So every computer computer program defines a set.
Now, lets take the set of integers, and take a power set of it. What we get is a set of all subsets of all integers.
Looking back at the predicate view, each of these subsets is esesntially juset a binary number, indicating presense or absense of a value.
For example, is_even(x) may be represnted as:
010101010101010101010101010101010101010101...
Is divisible by 3 may be:
001001001001001001001001001001001001001001...
And divisible by 7 as:
000000100000010000001000000100000010000001...
We can combine them via:
010101010101010101010101010101010101010101... even
001001001001001001001001001001001001001001... divisible by 3
000000100000010000001000000100000010000001... divisible by 7
....
....
We can make this list as long as we want. count in binary.
Now take diagonal elements, and flip them:
[1]10101010101010101010101010101010101010101...
0[1]1001001001001001001001001001001001001001...
00[1]000100000010000001000000100000010000001...
....
Now, no matter how how many sets we've listed, the set represented by this diagonal isn't on the list... (this is called diagonalization).
In our case, it shows that there are subsets (or mappings) that cannot be represented by predicates. So the mappings/sets exist (superset of all integers has that diagonal set), but we cannot have a predicate define that set.
What kind of sets cannot be defined by predicates (computer programs). One of them is the halting program. In fact, there's an infinite number of such sets that cannot be defined by a program.
 Alex; 20180716
 

Normally computer problems are seen as turning inputs into outputs, where output can be any number.
For example, square function would take a number and return its square.
Also, inputs and outputs are arbitrary numbers (if the input is a file, then the input number is the contents of the entire file). Same for output.
Godel came up with a way to map list of numbers into a single number: lets say we wanted to represent list 45, 2, 23, 0, 5 as a number, we could take 2^45 * 3^2 * 5^23 * 7^0 * 11^5. Note that we can represent any list of numbers as a single number (by doing powers of the prime factors).
So computer programs map numbers to other numbers. Number in, number out.
What if we're limited to programs that can only output a single bit as input. In logic these are called predicates. Number in, 0 or 1 out.
Are those programs just as powerful as those that can produce any number as output?
For example, can we build a function such as square using nothing but functions that return one bit? Obviously we're allowed to call our predicate program multiple times.
So let us setup a predicate that looks like: is_square_over(x,n){ if(x*x > n) return true; return false;
How can we use this predicate to calculate square of x?
The simplest approach is to do a for loop, for(i=0;;i++) if(is_square_over(x,i)) return i1;
But this for loop doesn't scale. If someone asks for square of 1000000 we would be looping for 1000000 times, while the square calculation is much simpler than that.
The trick is to find the uppwer bound on the output. Lower bound is already 0, the upper bound can start at 2, and then doubled until the predicate returns false.
Once we have the lower and upper bounds, we can do a binary search for the last instance where predicate returns false. It would look something like this:
sub square {
my ($x) = @_;
my $h = 1;
while(is_square_overN($x,$h)){
$h *=2;
}
my $l=int($h/2);
my $cnt = ($h+1)  $l;
while($cnt > 0){
my $step = int($cnt/2);
my $it = $l + $step;
if(is_square_overN($x,$it)){
$l = $it + 1;
$cnt = $step + 1;
}else{
$cnt = $step;
}
}
return $l;
}
Peformance wise, the square implementation using the predicate isn't that much slower; factor of log difference.
 Alex; 20180713
 

We sometimes hear of folks writing programs to generate first N digits of pi. Where N is a million, etc. It seems we can write programs to produce digits of pi: If left running forever, it will produce correct digits of pi for as long as it's running. If left running forever, it will produce pi.
Now, given this program (in a black box), can we write another program to verify that it truly does produce pi? The verification program would need to generate its own digits of pi, and compare them against the output of the black box program. If it finds a wrong one, it returns false. But will it ever return true?
This is just another variation on the halting problem.
But what if the verification program just returns "no". Will it be correct?
Consider that the box is finite (there's only so many bits it can hold). Eventually any finite machine must fall into a loopso any finite box that spits out digits of pi will eventually fail on some digit. Even a black box the size of the universe will not produce pi. So it's safe to return "no".
 Alex; 20180712
   

Kiddo getting MMR vaccinations.
 Alex; 20180703
 

that's one hot day...
 Alex; 20180629
 

Walking all over in Canada. Apparently they have fireworks every friday and saturday.
Played in the hotel poolkiddo loves the pool.
...and driving home :)
 Alex; 20180617
 

Driving out to Niagara Falls; going to the Canadian side.
 Alex; 20180616
         

Spent the morning watching the Berkshire Hathaway annual meeting.
Kiddo is 6months today :)
In other news, the reinforcement (or boosting) logic on top of the ntuple classifier: Apparently the performance curve flattens out... Running on EMNIST, after 100 training iterations, for n=10, using 32 tables gets us 92% accuracy, using 64 tables gets us 94% accuracy, using 128 tables gets us 94.6% accuracy, and using 256 tables gets us 95% accuracy. And then it just flat lines. I'm sure the code can be pushed to get 96% accuracy, perhaps by using bigger n, or using more tables, but it seems it's just not worth it.
So back to the idea land...
 Alex; 20180505
 

Finished reading The Structure of Scientific Revolutions by Thomas S. Kuhn. Pretty neat book. Essentially a bit of change in perspective on the clarity and correctness of science. Text books record only theories that survived the test of time, giving an impression that these theories and ideas behind them had a very structured progression. Reality is often much more disjointed and unpredictable than that.
 Alex; 20180504
 

Starting new job today :)
 Alex; 20180430
 

Implemented a reinforcement (or boosting) logic on top of the ntuple classifier that recognizes EMNIST digits. After about 100 iterations of that, the code went from ~80% accuracy to 95%!
 Alex; 20180429
  

Filler...
 Alex; 20180426
 

Why everyone is stressing about the 10year Treasury yield. Uh, oh.
In other news, finally heard back from $CORP_NAME. The background check is mostly done, except for a few things that they're willing to do after I join. So they've asked me for the start date.
Coincidentally, today is exactly two months since my last day at FINRA :)
As much as I like relaxing, I'll probably start next week.
 Alex; 20180423
 

So apparently my old ZenPad, that `broke' almost a year ago (the wifi wasn't coming upno matter how many resets I've done, etc.; I thought it was a hardware issuewifi chip burned out, or something) suddenly came back to live and started working again (wifi and everything).
In other news, discovered that my calendar was wrong. One of my classes actually starts 10 minutes earlier (6:05pm) than is marked in my calendar (6:15pm). So I've been showing up late most of the semester! This is pretty embarrassing :/
UPDATE: So to add to my embarrassment, this time, I showed up on time... but to the wrong classroom. 232NE vs 234NE. They're nearly identical, and I really didn't pay attention which one I walked into. Strangely enough, some students actually showed up in the wrong classroomso for about 20 or so minutes we were wondering what happened to the rest of the class :/
 Alex; 20180417
 

Yey, kiddo had his first meal today: baby oatmeal!
 Alex; 20180416
  

This doesn't happen every day: an actual snowy day in NYC... in April!
 Alex; 20180402
 

Visited two Toys R Us today.
In other news, accepted $CORP_NAME offer.
 Alex; 20180328
 

Ok, I think I might need to create some exams soon :/
Heard back from $SOCIAL_MEDIA_CORP. They have made a ``difficult decision not to move forward at this time.'' So... amm... The Paradox of Choice: I'm pretty relieved. I've given them a try, a very fair chance to hire me, and they chose not to, so... yey!
Guess it would've been weird of them to hire someone whose never had (and still doesn't have) a $SOCIEAL_MEDIA_CORP account :)
 Alex; 20180327
 

Or, you can directly go to a desired entry.

NOTICE: We DO NOT collect ANY personal information on this site. 

© 19962016 by End of the World Production, LLC. 
 