18.06.05: ‘Systems of linear equations as matrices’ Lecturer: Barwick Friday 12 February 2016 18.06.05: ‘Systems of linear equations as matrices’ Suppose ๐11 ๐12 ๐ ๐22 ๐ด = ( 21 โฎ โฎ ๐๐1 ๐๐2 โฏ ๐1๐ โฏ ๐2๐ ) โฑ โฎ โฏ ๐๐๐ an ๐ × ๐ matrix, and ๐ฃ1 ๐ฃ ( 2 ) ∈ R๐ โฎ ๐ฃ๐ a vector. 18.06.05: ‘Systems of linear equations as matrices’ Last time, we learned that the following are logically equivalent: 1. ๐ฃ โ = ๐ด๐ฅโ for some vector ๐ฅโ ∈ R๐ ; 2. the system of linear equations ๐ฃ1 = ๐11 ๐ฅ1 + ๐12 ๐ฅ2 + โฏ + ๐1๐ ๐ฅ๐ ; ๐ฃ2 = ๐21 ๐ฅ1 + ๐22 ๐ฅ2 + โฏ + ๐2๐ ๐ฅ๐ ; โฎ ๐ฃ๐ = ๐๐1 ๐ฅ1 + ๐๐2 ๐ฅ2 + โฏ + ๐๐๐ ๐ฅ๐ ; has at least one solution; 3. ๐ฃ โ lies in the span of the column vectors ๐ด1โ , ๐ด2โ , … , ๐ด๐โ . 18.06.05: ‘Systems of linear equations as matrices’ Can this system of equations be solved? 1 = 2๐ฅ + 4๐ฆ; 1 = 6๐ฅ + 8๐ง; 1 = 10๐ฆ + 12๐ง. 18.06.05: ‘Systems of linear equations as matrices’ How about this one? 3 = 2๐ฅ + 5๐ฆ; −7 = 3๐ฅ + 3๐ฆ; 3 = 5๐ฅ + 2๐ฆ. 18.06.05: ‘Systems of linear equations as matrices’ This one? 3 = 2๐ฅ + 5๐ฆ; 0 = 3๐ฅ + 3๐ฆ; −3 = 5๐ฅ + 2๐ฆ. 18.06.05: ‘Systems of linear equations as matrices’ The following are logically equivalent: 1. there are infinitely many vectors ๐ฅโ such that ๐ด๐ฅโ = 0; 2. there is a nonzero vector ๐ฅโ ∈ R๐ that is orthogonal to all of the row vectors ๐ดโ 1 , ๐ดโ 2 , … , ๐ดโ ๐ ; 3. the row vectors ๐ดโ 1 , ๐ดโ 2 , … , ๐ดโ ๐ do not span all of R๐ . 18.06.05: ‘Systems of linear equations as matrices’ Why are those last two the same? On the one hand, if there’s a vector ๐ฅโ ∈ R๐ that is orthogonal to all of the row vectors ๐ดโ 1 , ๐ดโ 2 , … , ๐ดโ ๐ , then certainly the row vectors ๐ดโ 1 , ๐ดโ 2 , … , ๐ดโ ๐ do not span all of R๐ . But what about the other direction? If there’s a vector ๐ฅโ not in the span of ๐ดโ 1 , ๐ดโ 2 , … , ๐ดโ ๐ , then it doesn’t have to be orthogonal to each of ๐ดโ 1 , ๐ดโ 2 , … , ๐ดโ ๐ … 18.06.05: ‘Systems of linear equations as matrices’ However, there’s a procedure (called the Gram–Schmidt process) that produces from ๐ฅโ a new vector ๐ฆโ that is orthogonal to each of the vectors ๐ดโ 1 , ๐ดโ 2 , … , ๐ดโ ๐ . We will return to this later in the course! 18.06.05: ‘Systems of linear equations as matrices’ How many solutions have we here? 0 = 2๐ฅ + 4๐ฆ; 0 = 4๐ฆ + 6๐ง. 18.06.05: ‘Systems of linear equations as matrices’ How many here? 0 = ๐ฅ + ๐ฆ − ๐ง; 0 = ๐ฅ − ๐ฆ + ๐ง; 0 = −๐ฅ + ๐ฆ + ๐ง. 18.06.05: ‘Systems of linear equations as matrices’ Here’s a fun exercise: in R4 , can we find four vectors ๐ฃ1โ , ๐ฃ2โ , ๐ฃ3โ , ๐ฃ4โ such that the angle between any pair of them is ๐/4? 18.06.05: ‘Systems of linear equations as matrices’ Here’s one that might come as a surprise: suppose ๐ด an ๐ × ๐ matrix. Then the following are logically equivalent: โ are linearly independent; 1. the column vectors ๐ด1โ , ๐ด2โ , … , ๐ด๐ 2. the row vectors ๐ดโ 1 , ๐ดโ 2 , … , ๐ดโ ๐ are linearly independent; 18.06.05: ‘Systems of linear equations as matrices’ 3. the system of linear equations 0 = ๐11 ๐ฅ1 + ๐12 ๐ฅ2 + โฏ + ๐1๐ ๐ฅ๐ ; 0 = ๐21 ๐ฅ1 + ๐22 ๐ฅ2 + โฏ + ๐2๐ ๐ฅ๐ ; โฎ 0 = has exactly one solution. ๐๐1 ๐ฅ1 + ๐๐2 ๐ฅ2 + โฏ + ๐๐๐ ๐ฅ๐ ; 18.06.05: ‘Systems of linear equations as matrices’ Here’s a system of 512 linear equations in 512 variables ๐ฅ1 , ๐ฅ2 , … , ๐ฅ512 : 513 = ๐ฅ2 + ๐ฅ3 + โฏ + ๐ฅ511 + ๐ฅ512 ; 514 = ๐ฅ1 + ๐ฅ3 + โฏ + ๐ฅ511 + ๐ฅ512 ; โฎ 1023 = ๐ฅ1 + ๐ฅ2 + โฏ + ๐ฅ510 + ๐ฅ512 ; 1024 = ๐ฅ1 + ๐ฅ2 + โฏ + ๐ฅ510 + ๐ฅ511 . How many solutions does it have?