The thread: Measured, not assumed — page 7
Two bits a value, and what undoes them
The parentheses of a range minimum over 65,536 values are 131,072 bits. Everything that makes them answerable is 651,800 more — five times the payload — and one encoding choice nobody quotes accounts for a sixth of it on its own.
The data that is not a numberA million characters of the same thing
Every measurement this collection has published about real text was taken on twenty-four thousand characters, because the phrase count was quadratic. It is linear now, so here is the same corpus at forty times the size — and what forty times does to its own numbers.
Two parametersWhat the queries know that the map does not
A greedy rule that chooses landmark cells by rerunning a sample of past queries needs two hundred of them to draw level with a rule that reads only the map — and what it finally chooses, on map after map, is the four corners. Give the queries a destination instead of scattering them, and twenty are enough to beat the corners by 29% on eight maps out of eight. A query log is worth reading exactly to the extent that it is not uniform.
CountingThe price of remembering an answer
Give selection sort a table of the pairs it has already compared and it makes 16,805 comparisons on 256 elements instead of 32,640. The table pays for itself once a comparison costs more than 3.1 word operations, and that price stays near three at every size measured. Give it the full closure of everything its answers imply and it makes 13,079. The closure pays only past 425 operations a comparison at that size, and past 1,659 at twice the size. The cheap memory is worth having on long keys. The complete one is worth having almost nowhere.
The floorsA result the size of its own noise
Sorting rotations by their first four characters beat the full Burrows–Wheeler transform by 0.022 bits a symbol. On sixteen streams of the same source the mean difference is 0.004 with a standard deviation of 0.011, and the short context is ahead on ten of them. The 1.2% was one draw. What survives is sharper than the result it replaces: from four characters upwards, every run break either rule makes is at a word boundary, and nothing above four characters decides anything at all.
One pass, and no roomA detector that learns its own quiet
A burst detector built from a fast decayed counter needs a threshold, and the one it was measured with came from sixty independent runs of the background — an oracle no deployed system has. Two counters and a running mean square of their difference can supply it from the stream itself, and on a stream that does not move they hit a 5% false-alarm rate at every setting tried. What they lose is the burst: a scale that learns every reading learns the burst as noise and forgets it in 20 to 40 seconds whatever its size. Holding the scale while alarmed brings the oracle's horizon back for small bursts, turns a slow drift into an alarm that will not stop, and still cannot see past the lag at which the slow counter holds more of the burst than the fast one.
When the algorithm is a tableThe counts only short words carry
A spelling checker capped at two edits still gives a false suggestion to many words its vocabulary lacks, and the one evidence about a writer's intention it already holds is how often each vocabulary word is used. Scored by its nearest words' counts, weighed a tenth for each edit, an answer was predicted to separate a slip from an unfamiliar word on short queries. On the streams the checker was built against, whose typos are of words drawn evenly from the list, the false suggestions go to commoner words than the corrections do, and the test gives up sixteen corrections for each false suggestion it removes. Drawn as a writer mistypes, in proportion to use, the counts separate the two at last — and still trade worse than a plain cap of one edit, except on words of four letters or fewer, where a threshold removes a ninth of the false suggestions for one correction in ninety-six.
What is taught wronglyA parameter that waits for another
Four merge fold shapes over thirty-two evenly loaded shards leave errors of 665, 667, 665 and 667 — a fifth of a per cent apart. Give the same four shapes shards whose loads span seventeen-fold and they leave 148, 183, 323 and 403. The parameter did nothing until a second parameter moved, and every measurement that fixed the second one saw nothing.
The other axisThe structure that was supposed to halve
A bidirectional index holds the transform of the text and the transform of its reversal — 35,335 bits each, 70,670 together, exactly twice one index. The deferral that named it hoped it would stop the index doubling. It does not remove the doubling; it reuses it.
What is taught wronglyThe half of a fall that is the logarithm
Phrases per character on a real collection of essays fall by a factor of 2.34 as it grows. A shuffle of the same characters falls by 1.70. Nearly three quarters of the movement is arithmetic, and no definition of the measure says so.
Two parametersA stop that is correct and never sooner
A two-ended search can stop when the two frontiers' keys together reach the best route found, and it can also stop when either frontier's own estimate reaches it alone. Both rules are safe, so a search may use whichever fires first. On forty weighted grids the second never fires: at the moment the first one stops the search, the larger of the two own-keys stands at 64% of the route. The extra rule costs 60% more counted work and a second priority queue to find that out.
CountingThe comparisons that name the answer
Returning the 4,000 smallest of 65,536 keys in order needs 61,536 comparisons to eliminate the rest and 42,100 to order the ones returned. That was the floor, 103,636, and a tournament made 125,388. What the floor never charged is saying which 4,000 come back. Charge that, and the floor is 125,341. On the same input the tournament is 47 comparisons above it, and for every k up to a hundred it is exactly on it.
When it does not fitA key passed along the row
A full B+-tree leaf that offers a key to its immediate neighbours before splitting leaves an ascending stream 67.2% full. Let it look one sibling further and the same stream fills its leaves completely, 2,048 leaves where there were 3,048, for about the same key moves. On random keys each doubling of the reach adds a few points of fill and more writes than it saves, and at the whole parent a key travels sixteen leaves on average. On a stream that is mostly in order the same reach costs almost nothing.
CountingThe record that forgets on purpose
A table of every pair selection sort might ask is 131,072 bits at 256 elements and catches every repeated question. A direct-mapped cache of 1,024 pairs is 16,384 bits and catches 91% of them, at a break-even of 3.53 word operations a comparison against the table's 3.12. What decides the cache is not how long the sort waits before asking again — 98% of its repeats come within one pass — but how many other questions it asks meanwhile, and a model with nothing in it but collisions predicts every measured share to within 2.6%.
The floorsWhere a context stops naming the letter before
A sort of rotations by their first k characters matters only where those k characters fail to decide the character before them, and that share of positions can be counted in one pass with no sorting. It was proposed as a replacement for sweeping k. Counted as proposed, it misses by a factor of four: on a stream of independent words 37% of positions are still ambiguous at four characters, where the sweep has already gone flat. Counted only where the context sits inside a word that goes on past it — the one place more context can still help — it lands within a step of the sweep on four sources of five.
One pass, and no roomThe clock a burst cannot outlast
A burst detector that holds its scale while alarmed keeps a burst visible to a ceiling set by two half-lives, and cannot stop alarming on a slow drift. The repair proposed for it was a schedule: learn again once an excursion has outlasted anything a burst could produce. Timed by how long the alarm has lasted, the schedule changes almost nothing, because on a gentle rise the alarm breaks every few seconds and the clock never reaches the ceiling. Timed by how long the difference between the counters has stayed positive, it keeps every burst to within about a second of the ceiling and falls below the target false-alarm rate on steep drifts — and one quantity, the drift's offset measured in spreads of the difference, says where it still fails.
When the algorithm is a tableA line the corpus finds
Substitution costs built as a transversion penalty plus the distance between the two letters' values on a graded property beat every partition of the pairs, when the model is told the property, and fall below the plainest matrix when told the wrong one. Fitting each letter's position on a line instead was predicted to find the property's order from about four times the told model's data and then behave like it. It finds the order on every corpus of 100 sequence pairs or more but one, and every miss starts with the two letters whose values differ by a fifth of a bit. But knowing the order is not knowing the costs. Its alignments are no better than six free classes' until 800 pairs, sixteen times the data the told model needed for the same score, and on a generator with no graded property at all it finds a line anyway, a quarter as long as a real one at 200 pairs, and pays for it against the plain partition.
What is taught wronglyThe floor a merge does not settle at
Compute a fold's level floors from the shard histograms and the prediction over-shoots by 1.73. A merged summary's floor is not the floor a summary settles at on the same arrivals — it is 0.90 of it at four shards and 0.70 at sixty-four, straight in log₂ m at a 3% residual, because merging preserves the heavy counters and never runs their eviction cascade.
StructuresThe shape a range question is about
A range minimum is a question about a tree, and the tree is determined by the array. Twelve values, eleven parent links, and the answer to every one of the seventy-eight ranges is a lowest common ancestor — with the values themselves no longer needed.
The other axisA bit for every bit
A grid over 3,612 points is 43,344 bits of payload. The smallest any structure can be that distinguishes one permutation of 3,612 things from another is 37,485. There is 16% to play for, and the deferral that asked for a compressed grid assumed there was much more.
The index that replaces the textTwo thousand documents of two hundred characters
Every collection this field has measured has been a dozen documents of ten thousand characters. A real collection is usually the other shape, and the other shape moves every term in a document index — one of them by a factor of seventy.
Two parametersHow long a reweighting stays true
Johnson's one Bellman–Ford run costs under one per cent of an all-pairs computation because it is divided over every source. Asked one query at a time it is divided over nothing, and it still repays itself after 2.7 queries — because the preparation is one Bellman–Ford and every query saves a third of another. What decides the trade is not the query count but whether the graph holds still: at half a per cent of arcs redrawn between queries the stored potential is worth exactly nothing, and its life is geometric at a per-arc failure rate of 6.6%.
CountingThe spread a merge sort does not have
On 65,536 random keys quicksort's comparison count has a standard deviation of 42,470 and merge sort's has one of 141.9 — three hundred times smaller, on means only a quarter apart. Merge sort's count is a fixed bill minus 65,535 independent leftovers, one per merge, each worth about a coin toss, so its spread grows as the square root of n where quicksort's grows as n. From 256 keys up a single run of merge sort pins its mean to one per cent, and a third of quicksort's whole spread is decided by its first pivot.
One pass, and no roomThe keys worth counting before sampling
The sampling estimator for the third frequency moment spends most of its copies on keys that barely contribute to the answer. Finding the heavy keys first with a Count-Sketch, counting them, and sampling only the rest cuts the state for ten per cent from 5,260 words to about 1,500 on a moderately skewed stream, and from 23,598 to under a thousand on the stream built to be worst for the sampler. It costs five times more than it saves on a flat stream, where there is no heavy key to find. What decides it is how the frequencies are shaped, not how long the stream is, and the frequencies must be counted, not read off the sketch, because the sketch's error on a heavy key is cubed.
The floorsThe merge that can say equal
When two sorted lists draw their keys from a few values, equal keys can go on either side of each other, so far fewer answers are needed and the counting floor falls. Merging four keys into four over four values needs only 27 answers, not 70, and the floor falls from seven comparisons to five. A merge whose comparisons answer only 'at most' still needs seven on its worst input, exactly as many as distinct keys, because it can never learn that two keys are equal. A merge whose comparisons can answer 'equal' needs five and meets the floor. Solved exactly over every input, the third answer is the whole difference, and on average it is worth more than an optimal strategy.
What is taught wronglyThe warning that is silent for the right reason
The statistic shipped to warn that a merge prediction is about to fail reads 0.160 on a stationary stream, 0.172 on a bursty one and 0.909 on a drifting one. It was asked to be looked at again because it does not catch a burst. It does not, and the reason is that on a burst there is nothing to catch.
StructuresThe runs a permutation does not leave
A run of length L costs 2⌊log₂ L⌋ + 1 bits and replaces L, so coding runs pays above a mean run of six. This grid's mean run is 2.34, chance gives 1.90, and coding its runs makes it 17.8% larger.
What the libraries doThe crossing that never arrives
Output-sensitive document listing exists because a pattern can occur four thousand times in eight documents. On a real collection of two thousand short documents it occurs 1.04 times per document, and the whole apparatus buys nothing at all.
Two parametersA potential mended where it broke
A stored reweighting on a 256-vertex graph with negative arcs costs 10,045 relaxations to rebuild, and rebuilding it every time an update breaks it stops paying once half a per cent of arcs change between queries. Mending it from the arcs that broke costs 16 to 442 relaxations instead, and the stored potential stays at two thirds of the per-query cost at every rate of change. When the change is a vertex whose costs all move together, a repair reaches nearly every vertex. It still costs a third of a rebuild.
When it does not fitThe runs a range query can skip
A Bloom filter over key prefixes can tell a range query that a run holds nothing it wants, which a filter over keys cannot. On a four-level store of a million keys, a range expected to hold one key costs 4.76 transfers with no such filter and 2.41 with one built on 20-bit prefixes. The best length moves two bits for every tenfold longer range. By a hundred keys the most any range filter could save is a fifth of a transfer, because only the smallest level is ever empty. And the same memory spent on 20-bit prefixes makes a lookup for an absent key three times dearer.