Gradient Descent and Classification: 梯度下降法和分类

Week 6 strings

Earthquakes Example

date_express <- "^[0-9]{4}/[0-9]{2}/[0-9]{2}"
head(grep(quakes, pattern = date_express))
# return the rows

head(grep(quakes, pattern = date_express, value = TRUE))
# value = T ==> return values
# similar to quakes[grep(quakes, pattern = date_express)]

grep( ) Family

• grep() returns a logical indicating a match. (which row of the parrten)
• regexpr() returns the location of the first match with attributes like the length of the match.
• gregexpr() works similarly to regexpr(), but returns all matching locations. ‘g’ for global. (all matching patterns' locations in each row)
• regmatches() takes strings and the output of regexpr() or gregexpr() and returns the actual matching strings.

# Is there a match?
# find if partten in the lines
grep("a[a-z]", "Alabama")

# Information about the first match.
# find in each string, the 1st time partten shows up
regexpr("a[a-z]", "Alabama")
[3]

# Information on all matches.
# find the all position that partten shows up
gregexpr("a[a-z]", "Alabama")
[[1]] 3 5
# first a in 3rd, second a in 5th

# What are the matches?
# use gregexpr/regexpr find the values in each string which contains the partten
regmatches("Alabama", gregexpr("a[a-z]", "Alabama"))
[[1]]
[1] "ab" "am"

Earthquakes Example

# with/without "-", 1 or more digit num, with ".",
# and follow 4 digit num in the end
coord_exp <- "-?[0-9]+\\\\.[0-9]{4}"
full_exp <- paste(coord_exp, "\\\\s+", coord_exp, sep = "")

# . ==> any character
# \\. ==> special character for "."
# \\s ==> space

• If we using regmatches() get the result with list format, and the 2 value in SAME sting, we can use sapply() to split the string

• also, according to the lab3, if rhe 2 values in one line of list but in DIFF string, we can unlist or trans-direction (row to column or inverse) with sapply().

  coords_split <- sapply(coords, strsplit, split="\\\\s+")
  
  p1char_num <- data.frame(t(sapply(regmatches(p1char_group, 
  													p1char_findnum), as.numeric)))
  

Web Scraping

• see lecture slides