THE TOPICS •Some Mathematics •Some Weird Programming Languages •Analyzing Keyboard Performances •Generating Ancient Greek Graded Readers •Emulating the Apple ][ •Writing an Operating System
EIGHT COMMANDS > increment the data pointer < decrement the data pointer + increment the byte at the data pointer - decrement the byte at the data pointer . output the byte at the data pointer as an ASCII-encoded character , accept one byte of input, storing its value in the byte at the data pointer [ if the byte at the data pointer is zero, jump forward after matching ] ] if the byte at the data pointer is nonzero, jump back after matching [
def run(self, stop=None): if stop is None: stop = len(self.program) while self.pc < stop: c = self.program[self.pc] if c == ">": self.p += 1 elif c == "elif c == "+": self.mem[self.p] += 1 elif c == "-": self.mem[self.p] -= 1 elif c == ".": sys.stdout.write(chr(self.mem[self.p])) elif c == ",": self.mem[self.p] = ord(sys.stdin.read(1)) elif c == "[": ...
ZERO = FALSE SUCC = lambda a: lambda b: lambda c: (b)((a)(b)(c)) ONE = (SUCC)(ZERO) TWO = (SUCC)(ONE) THREE = (SUCC)(TWO) FOUR = (SUCC)(THREE) def church_number(n): return SUCC(church_number(n - 1)) if n else FALSE
GIT FUNDAMENTALS • each file’s contents (called a blob) are hashed and that hash becomes a key into a dictionary of files • each directory is viewed as a sorted list of filenames paired with hash of file’s contents • this directory listing can then be hashed and so an entire directory tree can be hashed • a tree’s hash + commit message + committer + timestamp + parent(s) can then be hashed
class List(NodeBase): def shrink(self, content): return [ self.repo.shrink(item) for item in content ] def expand(self): return [ self.repo.expand(item) for item in self.content ]
THE BASIC IDEA •record a performance of the exercise / piece as MIDI (or similar) events •align the performed notes with the “score” notes •identify errors as well as fluctuations in timing, velocity, etc. •basically a performance “diff”
align with Needleman-Wunsch algorithm def note_similarity(score_note, performance_note): if score_note[0] == performance_note[1]: return 1 elif abs(score_note[0] - performance_note[1]) < 3: return 0.5 else: return 0 with a -1 for insertions and deletions plan to tweak to include velocity, duration, etc
THE BASIC IDEA •overcoming limitations of inductive, corpus-based learning of a language •generate graded readers •build up from much smaller pieces •optimize ordering of vocabulary •present in larger context even when context is not yet understood in target language
THE MYTH OF VOCABULARY COVERAGE if you learn top 100 words, you’ll know... one word in 99.9% of verses 50% of words in 91.3% of verses 75% of words in 24.4% of verses 90% of words in 2.1% of verses 95% of words in 0.6% of verses 100% of words in 0.4% of verses
THE MYTH OF VOCABULARY COVERAGE if you learn top 100 words, you’ll know... one word in 99.9% of verses 50% of words in 91.3% of verses 75% of words in 24.4% of verses 90% of words in 2.1% of verses 95% of words in 0.6% of verses 100% of words in 0.4% of verses
# a dictionary mapping targets to a set of items that are # needed (and initially missing) MISSING_IN_TARGET = defaultdict(set) ... # for each item, a score of how bad it is that it is missing MISSING_ITEMS = defaultdict(int) for missing in MISSING_IN_TARGET.values(): for item in missing: MISSING_ITEMS[item] += 1. / (2 ** len(missing)) if not MISSING_ITEMS: break next_item = sorted(MISSING_ITEMS, key=MISSING_ITEMS.get)[-1] for target in TARGETS_MISSING[next_item]: MISSING_IN_TARGET[target].remove(next_item)
OTHER WORK AND FUTURE PLANS •consideration of other prerequisite knowledge •intrinsic difficulty of a word (cognates, etc) •partial dynamic interlinears •inline replacement
THE BASIC IDEA •run the Python VM on bare metal •remove parts of CPython that assume an operating system •minimize libc •add built-ins to directly access memory and I/O ports •write drivers, etc in Python •even “must be written in assembly” parts use python-like syntax
OTHER THINGS •Forth Interpreter •Pascal Interpreter (as a prolegomenon to even more insanity) •Relational Algebra Implementation •Quantum Computing •Unicode Collation Algorithm Implementation •Global Illumination Renderer •Hacking Data Files from Lord of the Rings Online and Skyrim
jtauber
javiergs
a4lg
mot_techtalk
kinakomoti321
okuzawats
masahiro_nishimi
guikingone
maroon1st
dicoding
shgishzk
akase244
sainu
lara
tammielis
andyhume
shpigford
maggiecrowley
sugarenia
roundedbygravity
akmur
zenorocha
holman
paulrobertlloyd
hannesfritz