• Log InLog In
  • Register
Liquid`
Team Liquid Liquipedia
EDT 00:29
CEST 06:29
KST 13:29
  • Home
  • Forum
  • Calendar
  • Streams
  • Liquipedia
  • Features
  • Store
  • EPT
  • TL+
  • StarCraft 2
  • Brood War
  • Smash
  • Heroes
  • Counter-Strike
  • Overwatch
  • Liquibet
  • Fantasy StarCraft
  • TLPD
  • StarCraft 2
  • Brood War
  • Blogs
Forum Sidebar
Events/Features
News
Featured News
Team TLMC #5 - Finalists & Open Tournaments0[ASL20] Ro16 Preview Pt2: Turbulence10Classic Games #3: Rogue vs Serral at BlizzCon9[ASL20] Ro16 Preview Pt1: Ascent10Maestros of the Game: Week 1/Play-in Preview12
Community News
BSL 2025 Warsaw LAN + Legends Showmatch0Weekly Cups (Sept 8-14): herO & MaxPax split cups4WardiTV TL Team Map Contest #5 Tournaments1SC4ALL $6,000 Open LAN in Philadelphia8Weekly Cups (Sept 1-7): MaxPax rebounds & Clem saga continues29
StarCraft 2
General
#1: Maru - Greatest Players of All Time Weekly Cups (Sept 8-14): herO & MaxPax split cups Team Liquid Map Contest #21 - Presented by Monster Energy SpeCial on The Tasteless Podcast Team TLMC #5 - Finalists & Open Tournaments
Tourneys
Maestros of The Game—$20k event w/ live finals in Paris Sparkling Tuna Cup - Weekly Open Tournament SC4ALL $6,000 Open LAN in Philadelphia WardiTV TL Team Map Contest #5 Tournaments RSL: Revival, a new crowdfunded tournament series
Strategy
Custom Maps
External Content
Mutation # 491 Night Drive Mutation # 490 Masters of Midnight Mutation # 489 Bannable Offense Mutation # 488 What Goes Around
Brood War
General
BW General Discussion Soulkey on ASL S20 A cwal.gg Extension - Easily keep track of anyone BGH Auto Balance -> http://bghmmr.eu/ ASL20 General Discussion
Tourneys
[Megathread] Daily Proleagues BSL 2025 Warsaw LAN + Legends Showmatch [ASL20] Ro16 Group D [ASL20] Ro16 Group C
Strategy
Simple Questions, Simple Answers Muta micro map competition Fighting Spirit mining rates [G] Mineral Boosting
Other Games
General Games
Path of Exile Stormgate/Frost Giant Megathread Borderlands 3 General RTS Discussion Thread Nintendo Switch Thread
Dota 2
Official 'what is Dota anymore' discussion LiquidDota to reintegrate into TL.net
League of Legends
Heroes of the Storm
Simple Questions, Simple Answers Heroes of the Storm 2.0
Hearthstone
Heroes of StarCraft mini-set
TL Mafia
TL Mafia Community Thread
Community
General
US Politics Mega-thread Canadian Politics Mega-thread Things Aren’t Peaceful in Palestine Russo-Ukrainian War Thread UK Politics Mega-thread
Fan Clubs
The Happy Fan Club!
Media & Entertainment
Movie Discussion! [Manga] One Piece Anime Discussion Thread
Sports
2024 - 2026 Football Thread Formula 1 Discussion MLB/Baseball 2023
World Cup 2022
Tech Support
Linksys AE2500 USB WIFI keeps disconnecting Computer Build, Upgrade & Buying Resource Thread High temperatures on bridge(s)
TL Community
BarCraft in Tokyo Japan for ASL Season5 Final The Automated Ban List
Blogs
I <=> 9
KrillinFromwales
The Personality of a Spender…
TrAiDoS
A very expensive lesson on ma…
Garnet
hello world
radishsoup
Lemme tell you a thing o…
JoinTheRain
RTS Design in Hypercoven
a11
Evil Gacha Games and the…
ffswowsucks
Customize Sidebar...

Website Feedback

Closed Threads



Active: 1619 users

[Math Help] AMGM Inequality for Optimization

Blogs > OnceKing
Post a Reply
1 2 Next All
OnceKing
Profile Blog Joined January 2009
United States939 Posts
November 17 2011 05:41 GMT
#1
For my calculus homework today, I was asked to find the value of a such that the curvature of y = e^(ax) was greatest for x = 0.
So I solved this problem probably the 'normal' way -- plug into the curvature function, set x = 0, use the first derivative to maximize the value of the function and find a's value.

It turned out to be a very ugly function to be taking the derivative of (in my opinion, I suppose), and I was thinking of trying to use the Arithmetic Mean-Geometric Mean Inequality to find the maximum value instead. However, I couldn't do it and clearly I can conclude that I really don't understand it at all. So, the question I pose to you is: how would I use the AMGM Inequality to optimize this? Better yet, if someone can give me a complete explanation for how to use it for optimization in general that would be very awesome and balleriffic.

I understand the basic idea: AM >= GM, so for the maximum value we would set AM = GM and solve. But with this problem I was just completely confused with how to do it.

Shown below is what the curvature function ended up being with x = 0.
K(0) = a^2/(1+a^2)^(3/2)

What I tried to do
+ Show Spoiler +
I'm solving for a in a^2/(1+a^2)^(3/2), then I can use the AMGM inequality.
So, if I let a_1 = a^2 and a_2 = (1+a^2)^(-3/2), then I know (a_1+a_2)/2 >= sqrt(a_1*a_2).
And then solving for a becomes a huge mess.
Is it possible that in this case, AMGM is just an undesirable and inefficient way compared to the derivative method for optimization?


"Every man has his follies - and often they are the most interesting thing he has got."
xxpack09
Profile Blog Joined September 2010
United States2160 Posts
November 17 2011 05:49 GMT
#2
I don't even understand how AM/GM is supposed to apply to this at all, I would just use the standard optimization techniques.

btw expect this thread to be closed soon (homework)
Steel
Profile Blog Joined April 2010
Japan2283 Posts
November 17 2011 06:01 GMT
#3
I agree, just use the derivative and practice your algebra Derivatives are always relatively simple to do (integrals are the ugly unsolvable ones )...just need more practice I think.
Try another route paperboy.
Navillus
Profile Joined February 2011
United States1188 Posts
November 17 2011 06:03 GMT
#4
Just a heads up as I understand it HW help threads aren't allowed here.
"TL gives excellent advice 99% of the time. The problem is no one listens to it." -Plexa
infinitestory
Profile Blog Joined April 2010
United States4053 Posts
Last Edited: 2011-11-17 06:11:16
November 17 2011 06:06 GMT
#5
To bring some closure to this topic, I tried all the AM-GM I know and it doesn't work. For optimization, Jensen's Inequality and Cauchy-Schwarz are usually more well-suited, although I don't think either of those are particularly effective here either.

EDIT: Ok if mani is cool with it, I guess this is fine
Translator:3
OnceKing
Profile Blog Joined January 2009
United States939 Posts
November 17 2011 06:06 GMT
#6
Okay I should quantify this thread then.

This is NOT a homework thread. I am not asking you or anyone to do my homework for me. It involves a homework problem, yes, but I solved it as stated in the OP using the derivative method and got the right answer (sqrt 2, -sqrt 2). This is instead an exploration into solving a problem in an alternative method.
"Every man has his follies - and often they are the most interesting thing he has got."
Manifesto7
Profile Blog Joined November 2002
Osaka27151 Posts
November 17 2011 06:07 GMT
#7
Im cool with it.
ModeratorGodfather
micronesia
Profile Blog Joined July 2006
United States24701 Posts
November 17 2011 06:11 GMT
#8
Yea I left it alone also lol :p

The OP put isn't asking us to do anything... it's more of a math Q&A initially caused by a hw problem than a "do my hw" thread.
ModeratorThere are animal crackers for people and there are people crackers for animals.
igotmyown
Profile Blog Joined April 2009
United States4291 Posts
Last Edited: 2011-11-17 06:45:13
November 17 2011 06:17 GMT
#9
Local maximum and minimum occur when the derivative = 0.
Derivative of a^2/(1+a^2)^(3/2) is
2a/(1+a^2)^(5/2)

Edit: oops
(2a+2a^3-3a^3)/(1+a^2)^(5/2)
2a-a^3=0
a=0 or a=sqrt(1/2), a=-sqrt(1/2)

that's 0 when a = 0. Check whether that's minimum or maximum

I don't remember if that's really the curvature function, so I'm assuming your K(0) is correct.
Plexa
Profile Blog Joined October 2005
Aotearoa39261 Posts
Last Edited: 2011-11-17 06:40:55
November 17 2011 06:19 GMT
#10
(a_1+a_2)/2 >= sqrt(a_1*a_2), max when equal
a_1^2+2a_1*a_2+a_2^2 = 4*a_1*a_2
a_1^2-2a_1*a_2+a_2^2 = 0

Then :
(a_1-a_2)^2=0 => a_1 = a_2 (as expected)

Now: a^2 = (1+a^2)^(-3/2) which apparently doesn't give the right answer!
Administrator~ Spirit will set you free ~
Empyrean
Profile Blog Joined September 2004
16998 Posts
November 17 2011 06:21 GMT
#11
By the way, TL isn't a homework help forum :<
Moderator
Plexa
Profile Blog Joined October 2005
Aotearoa39261 Posts
November 17 2011 06:22 GMT
#12
On November 17 2011 15:11 micronesia wrote:
Yea I left it alone also lol :p

The OP put isn't asking us to do anything... it's more of a math Q&A initially caused by a hw problem than a "do my hw" thread.

It's cool empyrean
Administrator~ Spirit will set you free ~
infinitestory
Profile Blog Joined April 2010
United States4053 Posts
Last Edited: 2011-11-17 06:27:05
November 17 2011 06:24 GMT
#13
ok so I actually found a way to do this, but it's pretty convoluted

Considering the partial fraction decomposition of a^2/(1+a^2)^2 = 1/(1+a^2)-1/(1+a^2)^2 led me to a^2/(1+a^2)^(3/2) = 1/(1+a^2)^(1/2) - 1/(1+a^2)^(3/2). Let x=1/(1+a^2)^(1/2), and so we now want to maximize x-x^3, which is equal to x(1-x^2) or 2*((1/2-(1/2)*x^2)*(x^2)^(1/2). By the weighted form of AM-GM, (((1/2-(1/2)*x^2)*(x^2)^(1/2))^(2/3) <= (2/3)((1/2-(1/2)*x^2)+(1/2)*x^2) = 1/3. This maximum can only be achieved when (1/2-(1/2)*x^2)=(x^2), or x^2=1/3. Substituting gives 1/(1+a^2) = 1/3, or a^2 = 2.

On November 17 2011 15:19 Plexa wrote:
Now: a^2 = (1+a^2)^(-3/2) which you can solve as
0=a^2-(1+a^2)^(2/3) is a difference of two squares

according to Mathematica this gives the wrong answer
Translator:3
Plexa
Profile Blog Joined October 2005
Aotearoa39261 Posts
November 17 2011 06:30 GMT
#14
On November 17 2011 15:24 infinitestory wrote:
ok so I actually found a way to do this, but it's pretty convoluted

Considering the partial fraction decomposition of a^2/(1+a^2)^2 = 1/(1+a^2)-1/(1+a^2)^2 led me to a^2/(1+a^2)^(3/2) = 1/(1+a^2)^(1/2) - 1/(1+a^2)^(3/2). Let x=1/(1+a^2)^(1/2), and so we now want to maximize x-x^3, which is equal to x(1-x^2) or 2*((1/2-(1/2)*x^2)*(x^2)^(1/2). By the weighted form of AM-GM, (((1/2-(1/2)*x^2)*(x^2)^(1/2))^(2/3) <= (2/3)((1/2-(1/2)*x^2)+(1/2)*x^2) = 1/3. This maximum can only be achieved when (1/2-(1/2)*x^2)=(x^2), or x^2=1/3. Substituting gives 1/(1+a^2) = 1/3, or a^2 = 2.

Show nested quote +
On November 17 2011 15:19 Plexa wrote:
Now: a^2 = (1+a^2)^(-3/2) which you can solve as
0=a^2-(1+a^2)^(2/3) is a difference of two squares

according to Mathematica this gives the wrong answer

Something is going funny, lemme check over things
Administrator~ Spirit will set you free ~
igotmyown
Profile Blog Joined April 2009
United States4291 Posts
Last Edited: 2011-11-17 06:40:21
November 17 2011 06:39 GMT
#15
On November 17 2011 15:19 Plexa wrote:
(a_1+a_2)/2 >= sqrt(a_1*a_2), max when equal
a_1^2+2a_1*a_2+a_2^2 = 4*a_1*a_2
a_1^2-2a_1*a_2+a_2^2 = 0

Then :
(a_1-a_2)^2=0 => a_1 = a_2

Now: a^2 = (1+a^2)^(-3/2) which apparently doesn't give the right answer!

So you proved that the geometric mean is maximized relative to the arithmetic mean when it's equal.

Then you assumed the numerator and denominator of a problem are equal, to achieve said maximization.
Plexa
Profile Blog Joined October 2005
Aotearoa39261 Posts
November 17 2011 06:46 GMT
#16
On November 17 2011 15:39 igotmyown wrote:
Show nested quote +
On November 17 2011 15:19 Plexa wrote:
(a_1+a_2)/2 >= sqrt(a_1*a_2), max when equal
a_1^2+2a_1*a_2+a_2^2 = 4*a_1*a_2
a_1^2-2a_1*a_2+a_2^2 = 0

Then :
(a_1-a_2)^2=0 => a_1 = a_2

Now: a^2 = (1+a^2)^(-3/2) which apparently doesn't give the right answer!

So you proved that the geometric mean is maximized relative to the arithmetic mean when it's equal.

Then you assumed the numerator and denominator of a problem are equal, to achieve said maximization.

It doesn't take very long to realise that there is something amiss, but I think the bigger question is understanding why you wouldn't go for this approach as opposed to infinitestory's
Administrator~ Spirit will set you free ~
SonuvBob
Profile Blog Joined October 2006
Aiur21549 Posts
November 17 2011 06:47 GMT
#17
On behalf of the 99%, I'd just like to say: Thanks for making us feel stupid, guys.
Administrator
igotmyown
Profile Blog Joined April 2009
United States4291 Posts
Last Edited: 2011-11-17 06:49:09
November 17 2011 06:48 GMT
#18
On November 17 2011 15:24 infinitestory wrote:
ok so I actually found a way to do this, but it's pretty convoluted

Considering the partial fraction decomposition of a^2/(1+a^2)^2 = 1/(1+a^2)-1/(1+a^2)^2 led me to a^2/(1+a^2)^(3/2) = 1/(1+a^2)^(1/2) - 1/(1+a^2)^(3/2). Let x=1/(1+a^2)^(1/2), and so we now want to maximize x-x^3, which is equal to x(1-x^2) or 2*((1/2-(1/2)*x^2)*(x^2)^(1/2). By the weighted form of AM-GM, (((1/2-(1/2)*x^2)*(x^2)^(1/2))^(2/3) <= (2/3)((1/2-(1/2)*x^2)+(1/2)*x^2) = 1/3. This maximum can only be achieved when (1/2-(1/2)*x^2)=(x^2), or x^2=1/3. Substituting gives 1/(1+a^2) = 1/3, or a^2 = 2.

Show nested quote +
On November 17 2011 15:19 Plexa wrote:
Now: a^2 = (1+a^2)^(-3/2) which you can solve as
0=a^2-(1+a^2)^(2/3) is a difference of two squares

according to Mathematica this gives the wrong answer


A very convoluted way to prove you maximize x-x^3 at -1 and 1, and you also aren't allowed to let x<0 or 1-x^2 <0 for the inequality to hold. But clever.
igotmyown
Profile Blog Joined April 2009
United States4291 Posts
November 17 2011 06:54 GMT
#19
On November 17 2011 15:46 Plexa wrote:
Show nested quote +
On November 17 2011 15:39 igotmyown wrote:
On November 17 2011 15:19 Plexa wrote:
(a_1+a_2)/2 >= sqrt(a_1*a_2), max when equal
a_1^2+2a_1*a_2+a_2^2 = 4*a_1*a_2
a_1^2-2a_1*a_2+a_2^2 = 0

Then :
(a_1-a_2)^2=0 => a_1 = a_2

Now: a^2 = (1+a^2)^(-3/2) which apparently doesn't give the right answer!

So you proved that the geometric mean is maximized relative to the arithmetic mean when it's equal.

Then you assumed the numerator and denominator of a problem are equal, to achieve said maximization.

It doesn't take very long to realise that there is something amiss, but I think the bigger question is understanding why you wouldn't go for this approach as opposed to infinitestory's


But what you did was found a way to maximize a^2+(1+a)^2)^(--3/2) (or some variant of the sum), relative to the product, rather than maximize the product itself.
Plexa
Profile Blog Joined October 2005
Aotearoa39261 Posts
November 17 2011 06:57 GMT
#20
On November 17 2011 15:54 igotmyown wrote:
Show nested quote +
On November 17 2011 15:46 Plexa wrote:
On November 17 2011 15:39 igotmyown wrote:
On November 17 2011 15:19 Plexa wrote:
(a_1+a_2)/2 >= sqrt(a_1*a_2), max when equal
a_1^2+2a_1*a_2+a_2^2 = 4*a_1*a_2
a_1^2-2a_1*a_2+a_2^2 = 0

Then :
(a_1-a_2)^2=0 => a_1 = a_2

Now: a^2 = (1+a^2)^(-3/2) which apparently doesn't give the right answer!

So you proved that the geometric mean is maximized relative to the arithmetic mean when it's equal.

Then you assumed the numerator and denominator of a problem are equal, to achieve said maximization.

It doesn't take very long to realise that there is something amiss, but I think the bigger question is understanding why you wouldn't go for this approach as opposed to infinitestory's


But what you did was found a way to maximize a^2+(1+a)^2)^(--3/2) (or some variant of the sum), relative to the product, rather than maximize the product itself.

Ya, I see it.
Administrator~ Spirit will set you free ~
1 2 Next All
Please log in or register to reply.
Live Events Refresh
OSC
19:00
Mid Season Playoffs
Spirit vs PercivalLIVE!
Cham vs TBD
ByuN vs Jumy
Liquipedia
[ Submit Event ]
Live Streams
Refresh
StarCraft 2
Nina 147
ProTech74
StarCraft: Brood War
PianO 334
Nal_rA 66
Bale 27
Icarus 4
Dota 2
NeuroSwarm111
League of Legends
JimRising 622
Counter-Strike
Coldzera 332
Stewie2K267
Super Smash Bros
Westballz22
Other Games
summit1g7721
C9.Mang0414
XaKoH 99
RuFF_SC248
ViBE46
Trikslyr29
Organizations
Other Games
gamesdonequick973
StarCraft 2
Blizzard YouTube
StarCraft: Brood War
BSLTrovo
sctven
[ Show 16 non-featured ]
StarCraft 2
• Sammyuel 32
• practicex 24
• AfreecaTV YouTube
• intothetv
• Kozan
• IndyKCrew
• LaughNgamezSOOP
• Migwel
• sooper7s
StarCraft: Brood War
• RayReign 13
• Diggity4
• BSLYoutube
• STPLYoutube
• ZZZeroYoutube
League of Legends
• Lourlo1043
• Stunt365
Upcoming Events
RSL Revival
5h 31m
Maru vs Reynor
Cure vs TriGGeR
Map Test Tournament
6h 31m
The PondCast
8h 31m
RSL Revival
1d 5h
Zoun vs Classic
Korean StarCraft League
1d 22h
BSL Open LAN 2025 - War…
2 days
RSL Revival
2 days
BSL Open LAN 2025 - War…
3 days
RSL Revival
3 days
Online Event
3 days
[ Show More ]
Wardi Open
4 days
Monday Night Weeklies
4 days
Sparkling Tuna Cup
5 days
LiuLi Cup
6 days
Liquipedia Results

Completed

Proleague 2025-09-10
Chzzk MurlocKing SC1 vs SC2 Cup #2
HCC Europe

Ongoing

BSL 20 Team Wars
KCM Race Survival 2025 Season 3
BSL 21 Points
ASL Season 20
CSL 2025 AUTUMN (S18)
LASL Season 20
RSL Revival: Season 2
Maestros of the Game
StarSeries Fall 2025
FISSURE Playground #2
BLAST Open Fall 2025
BLAST Open Fall Qual
Esports World Cup 2025
BLAST Bounty Fall 2025
BLAST Bounty Fall Qual
IEM Cologne 2025
FISSURE Playground #1

Upcoming

2025 Chongqing Offline CUP
BSL World Championship of Poland 2025
IPSL Winter 2025-26
BSL Season 21
SC4ALL: Brood War
BSL 21 Team A
Stellar Fest
SC4ALL: StarCraft II
EC S1
ESL Impact League Season 8
SL Budapest Major 2025
BLAST Rivals Fall 2025
IEM Chengdu 2025
PGL Masters Bucharest 2025
MESA Nomadic Masters Fall
Thunderpick World Champ.
CS Asia Championships 2025
ESL Pro League S22
TLPD

1. ByuN
2. TY
3. Dark
4. Solar
5. Stats
6. Nerchio
7. sOs
8. soO
9. INnoVation
10. Elazer
1. Rain
2. Flash
3. EffOrt
4. Last
5. Bisu
6. Soulkey
7. Mini
8. Sharp
Sidebar Settings...

Advertising | Privacy Policy | Terms Of Use | Contact Us

Original banner artwork: Jim Warren
The contents of this webpage are copyright © 2025 TLnet. All Rights Reserved.