我就研究緊活點解人既decision唔係optimized
heuristics
我之前都做unconscious mental process
基本上人腦唔是為左optimization而運作
但係從evolution角度黎講 如果人腦唔係optimize緊咁點樣生存到而家
定係好似tirole/benabou嗰個套咁 一啲suboptimal既behavior反而達致optimal outcome
同意，但同一時間好多optimized 左既responses 會因為社會變左而變得係某d 情況下唔adaptive
巴打有計傾
有兩種可能
1.constrainted maximization
2. 個environment轉左 mechanism轉唔切
Me research大概圍繞住第一點做
optimisation
我就研究緊活點解人既decision唔係optimized
heuristics
我之前都做unconscious mental process
基本上人腦唔是為左optimization而運作
但係從evolution角度黎講 如果人腦唔係optimize緊咁點樣生存到而家
定係好似tirole/benabou嗰個套咁 一啲suboptimal既behavior反而達致optimal outcome
同意，但同一時間好多optimized 左既responses 會因為社會變左而變得係某d 情況下唔adaptive
巴打有計傾
有兩種可能
1.constrainted maximization
2. 個environment轉左 mechanism轉唔切
Me research大概圍繞住第一點做
個post推慢左
讀緊game theory,入面有好多simplex optimization,用手計計到頭暈
Agger 好煩nash bargaining solution
Dynamic programming 已經好好做。
Behavioral econ 有啲咩multi-selves agent, dynamically inconsistent agent, 連dynamic programming 都唔apply，煩到嘔泡。
Nash bargaining 亦都係最簡單，有幾多relationship中的surplus只係按bargaining power分。
本人表示深受其害
X2
好奇下econ係咪好straight要搵global optimal pt
因為ml好多都係搵local就算
yes, global and unique。
咁都幾係 而家好多ml都係minimize某個functional就算
因為ML好多係non-convex optimization
但係econ應該都唔多convex?
但上面話Econ搵global and unique optimal point喎
唔strictly convex唔會有global and unique架喎
個post推慢左
放左advisor飛機 要補鑊做野
個post推慢左
放左advisor飛機 要補鑊做野
前日有個ug放左我飛機, 我屌到佢柒左
個post推慢左
放左advisor飛機 要補鑊做野
前日有個ug放左我飛機, 我屌到佢柒左
我趕唔切火車
下個星期再見佢 唔知會唔會被屌
個post推慢左
放左advisor飛機 要補鑊做野
前日有個ug放左我飛機, 我屌到佢柒左
Optimization 其實學乜野
點optimize某objective subject to 唔同constraint (equality or inequality)
幾時有solution, 點計呀etc
深啲既有dynamic optimization, decision making over time etc.
like high school finding local maximum pt/minimum pt
Dynamic programming
讀緊game theory,入面有好多simplex optimization,用手計計到頭暈
Agger 好煩
Dynamic programming 已經好好做。
Behavioral econ 有啲咩multi-selves agent, dynamically inconsistent agent, 連dynamic programming 都唔apply，煩到嘔泡。
Nash bargaining 亦都係最簡單，有幾多relationship中的surplus只係按bargaining power分。
好在我唔係modeling個邊
optimization其實係好正既topic 我做緊machine learning research
Machine learning入面好多algorithm 都係solve optimization problems
當中會用到好多real analysis同functional analysis既techniques
最正個位係可以自己諗某啲algorithms去solve某一類optimization problems然後paper
讀緊game theory,入面有好多simplex optimization,用手計計到頭暈
Agger 好煩
Dynamic programming 已經好好做。
Behavioral econ 有啲咩multi-selves agent, dynamically inconsistent agent, 連dynamic programming 都唔apply，煩到嘔泡。
Nash bargaining 亦都係最簡單，有幾多relationship中的surplus只係按bargaining power分。
本人表示深受其害
X2
好奇下econ係咪好straight要搵global optimal pt
因為ml好多都係搵local就算
yes, global and unique。
咁都幾係 而家好多ml都係minimize某個functional就算
因為ML好多係non-convex optimization
但係econ應該都唔多convex?
但上面話Econ搵global and unique optimal point喎
唔strictly convex唔會有global and unique架喎
基本上係convex
Randomized choice係其中可以make sure constraint set係convex set一個手段
Optimization 其實學乜野
點optimize某objective subject to 唔同constraint (equality or inequality)
幾時有solution, 點計呀etc
深啲既有dynamic optimization, decision making over time etc.
like high school finding local maximum pt/minimum pt
Dynamic programming
讀緊game theory,入面有好多simplex optimization,用手計計到頭暈
Agger 好煩
Dynamic programming 已經好好做。
Behavioral econ 有啲咩multi-selves agent, dynamically inconsistent agent, 連dynamic programming 都唔apply，煩到嘔泡。
Nash bargaining 亦都係最簡單，有幾多relationship中的surplus只係按bargaining power分。
好在我唔係modeling個邊
optimization其實係好正既topic 我做緊machine learning research
Machine learning入面好多algorithm 都係solve optimization problems
當中會用到好多real analysis同functional analysis既techniques
最正個位係可以自己諗某啲algorithms去solve某一類optimization problems然後paper
functional 同 real analysis 都好正
個post推慢左
放左advisor飛機 要補鑊做野
前日有個ug放左我飛機, 我屌到佢柒左
我趕唔切火車
下個星期再見佢 唔知會唔會被屌
個post推慢左
放左advisor飛機 要補鑊做野
前日有個ug放左我飛機, 我屌到佢柒左
我趕唔切火車
下個星期再見佢 唔知會唔會被屌
俾我係advisor都屌你
個post推慢左
放左advisor飛機 要補鑊做野
前日有個ug放左我飛機, 我屌到佢柒左
我趕唔切火車
下個星期再見佢 唔知會唔會被屌
俾我係advisor都屌你
做人公平d好 我都俾唔少prof放過飛機
放左advisor飛機 要補鑊做野
前日有個ug放左我飛機, 我屌到佢柒左
我趕唔切火車
下個星期再見佢 唔知會唔會被屌
俾我係advisor都屌你
做人公平d好 我都俾唔少prof放過飛機
我個prof覆email好慢,成日約唔到時間
Optimization 其實學乜野
點optimize某objective subject to 唔同constraint (equality or inequality)
幾時有solution, 點計呀etc
深啲既有dynamic optimization, decision making over time etc.
like high school finding local maximum pt/minimum pt
Dynamic programming
讀緊game theory,入面有好多simplex optimization,用手計計到頭暈
Agger 好煩
Dynamic programming 已經好好做。
Behavioral econ 有啲咩multi-selves agent, dynamically inconsistent agent, 連dynamic programming 都唔apply，煩到嘔泡。
Nash bargaining 亦都係最簡單，有幾多relationship中的surplus只係按bargaining power分。
好在我唔係modeling個邊
optimization其實係好正既topic 我做緊machine learning research
Machine learning入面好多algorithm 都係solve optimization problems
當中會用到好多real analysis同functional analysis既techniques
最正個位係可以自己諗某啲algorithms去solve某一類optimization problems然後paper
functional 同 real analysis 都好正
巴打講多啲QM有好多banach space,想知多啲functional analysis做啲咩
, 我更加唔識物理 又要準備溫書考qualifying Optimization 其實學乜野
點optimize某objective subject to 唔同constraint (equality or inequality)
幾時有solution, 點計呀etc
深啲既有dynamic optimization, decision making over time etc.
like high school finding local maximum pt/minimum pt
Dynamic programming
讀緊game theory,入面有好多simplex optimization,用手計計到頭暈
Agger 好煩
Dynamic programming 已經好好做。
Behavioral econ 有啲咩multi-selves agent, dynamically inconsistent agent, 連dynamic programming 都唔apply，煩到嘔泡。
Nash bargaining 亦都係最簡單，有幾多relationship中的surplus只係按bargaining power分。
好在我唔係modeling個邊
optimization其實係好正既topic 我做緊machine learning research
Machine learning入面好多algorithm 都係solve optimization problems
當中會用到好多real analysis同functional analysis既techniques
最正個位係可以自己諗某啲algorithms去solve某一類optimization problems然後paper
functional 同 real analysis 都好正
所以其實app math都有正既地方 而且冇咁離地
Agger 好煩
Dynamic programming 已經好好做。
Behavioral econ 有啲咩multi-selves agent, dynamically inconsistent agent, 連dynamic programming 都唔apply，煩到嘔泡。
Nash bargaining 亦都係最簡單，有幾多relationship中的surplus只係按bargaining power分。
好在我唔係modeling個邊
optimization其實係好正既topic 我做緊machine learning research
Machine learning入面好多algorithm 都係solve optimization problems
當中會用到好多real analysis同functional analysis既techniques
最正個位係可以自己諗某啲algorithms去solve某一類optimization problems然後paper
functional 同 real analysis 都好正
巴打講多啲
我d FA已經唔記得七七八八
某D Linear PDE可以用functional analysis的方法去deduce d property
e.g. compact operators
又可以用黎construct measures, e.g. Haar measure
同埋functional analysis本身會deal with d Banach / Hilbert spaces, 但係數學本身都會遇到呢D spaces, e.g. space of continuous functions, L^p spaces等等
probability都會見到, e.g. weak convergence
所以functional analysis算係analysis的基礎野
