Skip to content

Commit c347768

Browse files
committed
Quartz sync: Mar 5, 2026, 5:31 PM
1 parent d3f5ba2 commit c347768

24 files changed

Lines changed: 585 additions & 39 deletions

content/Algorithms/Analysis of Algorithms.md

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -210,7 +210,6 @@ The condition `i=2` is not a comparison but an assignment. Thus the loop's condi
210210
^q4
211211
<h6 class="question">Q4) Find the TC of the below code -</h6>
212212

213-
214213
![[Pasted image 20251227142309.png | 300]]
215214

216215
<strong><u>Sol</u></strong>$^n$ -

content/Algorithms/Dynamic Programming.md

Lines changed: 14 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,17 @@
1+
>[!SUMMARY] Table of Contents
2+
>- [[Dynamic Programming#Single Source Shortest Path (SSSP)|Single Source Shortest Path (SSSP)]]
3+
> - [[Dynamic Programming#Djikstra's Single Source Shortest Path|Djikstra's Single Source Shortest Path]]
4+
> - [[Dynamic Programming#Matrix Based approach|Matrix Based approach]]
5+
> - [[Dynamic Programming#Spanning Tree Approach|Spanning Tree Approach]]
6+
> - [[Dynamic Programming#Bellman-Ford algorithm|Bellman-Ford algorithm]]
7+
>- [[Dynamic Programming#All Pairs Shortest Path (ASSP)|All Pairs Shortest Path (ASSP)]]
8+
> - [[Dynamic Programming#Floyd-Warshall's Algorithm|Floyd-Warshall's Algorithm]]
9+
>- [[Dynamic Programming#0/1 Knapsack (Binary Knapsack)|0/1 Knapsack (Binary Knapsack)]]
10+
> - [[Dynamic Programming#Tabulation Method|Tabulation Method]]
11+
>- [[Dynamic Programming#Sum of Subsets (SOS)|Sum of Subsets (SOS)]]
12+
>- [[Dynamic Programming#Longest Common Subsequence (LCS)|Longest Common Subsequence (LCS)]]
13+
>- [[Dynamic Programming#Matrix Chain Multiplication (MCM)|Matrix Chain Multiplication (MCM)]]
14+
115
Dynamic Programming is an algorithm design method used for solving problems whose solutions are viewed as a result of making a set/sequence of decisions.
216
- One way of making these decisions is to make them one at a time in a step-wise (sequential) manner and never make any erroneous decision.
317
- When applying Greedy Methods, for many problems it is not possible to make step-wise decisions **based on local information** available at every step in such a manner that the sequence of decisions in optimal.

content/Basic CS Definitions.md

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,8 @@
11
1. **Algorithm -** A finite, well-defined sequence of steps that detail on how to achieve any particular task.
22
2. **Data Structure -** A logical structure with a set of operations which explains how some data will be stored, retrieved, and modified.
33
3. **Program -** A formal set of instructions written in a programming language that a computer executes to implement an algorithm and perform a computation.
4-
4. **Computation -** The process of executing a sequence of well-defined operations on data according to a set of rules which may or may not terminate.
4+
4. **Programming -** Process of taking an algorithm and encoding it into a formal language for execution by a computational model to perform a computation.
5+
5. **Computation -** The process of executing a sequence of well-defined operations on data according to a set of rules which may or may not terminate.
56

67
The format difference between an algorithm and a program is that an algorithm must always terminate while a program can be non-terminating (example - OS).
78

content/Mathematical Foundations of Generative AI/VDM and GANs.md

Lines changed: 19 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -75,6 +75,8 @@ $$
7575
\end{aligned}
7676
$$
7777

78+
We write $P_X || P_\theta$ to denote that we are calculating the divergence of $P_X$ from $P_\theta$, or in other words "how well does $P_\theta$ approximate $P_X$".
79+
7880
- *Convex function -* A function which has one unique minimum value (can have multiple minima).
7981
- *Strictly Convex function -* A function which has only one global minima.
8082
- *Left Semi-Continuous function -* A function in which the value at a point is equal to the limit when approached from the left.
@@ -88,6 +90,7 @@ Properties of $f$-divergence -
8890
2. $D_f(P_X\,||\, P_\theta) = 0$ iff $P_X = P_\theta$.
8991

9092
<h4 class="special">Examples of f-divergence -</h4>
93+
9194
1. $f(u) = u\,log\,u$ leads to the **KL (Kullback-Leilber) Divergence** - ^20cce2
9295

9396
$$
@@ -125,11 +128,21 @@ $$
125128
So one way to solve an integral like the [[VDM and GANs#f-divergence|f-Divergence]] is by using the above two mentioned laws and equating it to the expected value of function $h(x)=f\left(\frac{P_X(x)}{P_\theta(x)}\right)$. It would be a mathematically valid representation but is not directly computable from the data since the true data distribution $P_X$ is unknown.
126129

127130
### Conjugate of a convex function
128-
The conjugate of a convex function $f(u)$ is written as
131+
The conjugate of a convex function $f(u)$ (Fenchel Conjugate) is written as -
129132

130133
$$
131-
f^*(t) = \sup_{u \in \text{dom(f)}} \left\{ut - f(u)\right\}
134+
\begin{aligned}
135+
f^*(t) &= \sup_{u \in \text{dom(f)}} \left\{ut - f(u)\right\} \\[8pt]
136+
&\qquad\text{OR in general,} \\[8pt]
137+
f^*(t) &= \sup_{u \in \text{dom(f)}} \left\{<u,t> - f(u)\right\} \\[8pt]
138+
\end{aligned}
132139
$$
140+
141+
- Here $u$ belongs to the domain of $f$, but $t$ isn't from the range of $f$.
142+
- $t$ is some arbitrary vector in $\mathbb R^n$ that is used to probe $f$.
143+
- $t$ is used to make a linear function $<t,x>$ where $t$ defines the slope of the linear function.
144+
- The conjugate checks by how much does this linear function outperforms $f$.
145+
133146
At every point $t$, one constructs multiple lower bounds on that particular $u$ and chooses the tightest of those lower bounds (supremum/max) as value of the conjugate.
134147

135148
Properties of a conjugate of a convex function -
@@ -163,9 +176,11 @@ $$
163176

164177
where $\mathbb{T}: X \rightarrow \operatorname{dom} f*$ is the space of functions containing solutions for the inner optimization problem.
165178

166-
The space of functions $\mathbb{T}$ we are optimizing over may or may not contain $T^*(x)$ that is the solution to the inner optimization problem. This can occur either because $\mathbb{T}$ is a restricted function class (e.g., neural networks), or because the supremum defining the conjugate is not attained within $\mathbb{T}$.
179+
%% The space of functions $\mathbb{T}$ we are optimizing over may or may not contain $T^*(x)$ that is the solution to the inner optimization problem. This can occur either because $\mathbb{T}$ is a restricted function class (e.g., neural networks), or because the supremum defining the conjugate is not attained within $\mathbb{T}$. %%
180+
181+
For the solution $t^*$ of the inner optimization problem $<t,u> - f(u)$ is maximal. Because we are allowed to pick any arbitrary $t$ as this point, this optimal can be achieved. But upon restricting $t=T(x)$ by considering a function $T$ that belongs to a class of function $\mathbb T$, if $t^*$ lies outside the range of $T$ then $<t,u> - f(u)$ need not be maximal anymore. Thus in such a case -
167182

168-
Because we are restricting $\mathbb{T}$ and $\mathbb{T} \subseteq \, \{\text{all measurable functions}\}$, by using the fact for any arbitrary function $F$ that $\sup_{t \in A} F(x) \le \sup_{t \in B} F(x)$ if $A \subseteq B$ we can say,
183+
%% Because we are restricting $\mathbb{T}$ and $\mathbb{T} \subseteq \, \{\text{all measurable functions}\}$, by using the fact for any arbitrary function $F$ that $\sup_{t \in A} F(x) \le \sup_{t \in B} F(x)$ if $A \subseteq B$ we can say, %%
169184

170185
$$
171186
\begin{aligned}
Lines changed: 91 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,91 @@
1+
# Approximations
2+
We can approximate the value at any point of the curve by using polynomials of increasing orders.
3+
## Linear Approximation
4+
Consider a point $x=a$ on the graph at which we wish to approximate the value of the function using a line centered around some $x=a$. We center at $a$ because the approximation is fundamentally about **local behavior**, and locality is measured relative to the point you care about.
5+
6+
The equation of this approximation line would be $L(x)=m(x-a)+c$. Around some really small $\epsilon-$neighbourhood around $x=a$ we can say that the curve of the graph and the line coincide, thus $f(x) = L(x)$.
7+
8+
![[Pasted image 20260223084831.png|550]]
9+
10+
By plugging in $x=a$ we can get $c = f(a)$. Due to this we can even say that,
11+
$$
12+
\begin{alignedat}{3}
13+
&&\,\,f(x) &= m(x-a) + c \\[8pt]
14+
&\Rightarrow &m &= \frac{f(x)-f(a)}{x-a} \\[8pt]
15+
&\Rightarrow &\lim_{x\rightarrow a}m &= \lim_{x\rightarrow a}\frac{f(x)-f(a)}{x-a} \\[8pt]
16+
&\Rightarrow & m &= f'(a) \\[8pt]
17+
\end{alignedat}
18+
$$
19+
20+
Using this information we can finalize it all by doing,
21+
22+
$$
23+
\begin{alignedat}{3}
24+
&&L(x) &= m(x-a) + c \\[8pt]
25+
&\Rightarrow &f(x) &= f(a) + f'(a)(x-a) \\[8pt]
26+
\end{alignedat}
27+
$$
28+
29+
If we consider a fixed point $x$ instead of $a$ and we wish to find the value of any point some $\triangle x$ distance away from $x$, we can rewrite the above equations as
30+
31+
$$
32+
f(x+\triangle x) = f(x) + f'(x)\triangle x
33+
$$
34+
35+
This is the First-Order Approximation of a function, also called as the **Linear Approximation**.
36+
## Quadratic Approximation
37+
Similar to how we had a linear centered around $x=a$ for linear approximation, we can also have a parabola centered around $x=a$ for a second order approximation of the function. This parabola can be represented as,
38+
39+
$$
40+
p(x) = A(x-a)^2 + B(x-a) + C
41+
$$
42+
43+
Again, by plugging in $x=a$ we get that $C = p(a)$. As in some $\epsilon-$neighbourhood around $a$ the parabola and the curve of the graph would coincide, we can say that $p(a) = f(a)$.
44+
45+
We can get $B$ by doing,
46+
$$
47+
\begin{alignat*}{3}
48+
&& f(x) &= A(x-a)^2 + B(x-a) + C \tag{1}\\[8pt]
49+
&\Rightarrow & \,\,f(x) &= A(x-a)^2 + B(x-a) + f(a) \\[8pt]
50+
&\Rightarrow & \,\,\frac{f(x) - f(a)}{x-a} &= A(x-a) + B(x-a) \\[8pt]
51+
&\Rightarrow & \lim_{x \rightarrow a}A(x-a) + B(x-a) &= \lim_{x \rightarrow a}\frac{f(x) - f(a)}{x-a}\\[8pt]
52+
&\Rightarrow & B &= f'(a)\\[8pt]
53+
\end{alignat*}
54+
$$
55+
56+
We can get $A$ by taking the derivative of $(1)$ and solving for $A$.
57+
58+
$$
59+
\begin{alignat*}{3}
60+
&& f(x) &= A(x-a)^2 + B(x-a) + C \tag{1}\\[8pt]
61+
&\Rightarrow & \,\,f'(x) &= 2A(x-a) + B \\[8pt]
62+
&\Rightarrow & \,\,f'(x) &= 2A(x-a) + f'(a) \\[8pt]
63+
&\Rightarrow & \,\,A &= \frac{f'(x) - f'(a)}{2(x-a)} \\[8pt]
64+
&\Rightarrow & \,\,\lim_{x \rightarrow a}A &= \frac{1}{2}\lim_{x \rightarrow a}\frac{f'(x) - f'(a)}{(x-a)} \\[8pt]
65+
&\Rightarrow & \,\,A &= \frac{1}{2}f''(a)\\[8pt]
66+
\end{alignat*}
67+
$$
68+
69+
By putting it all together we get,
70+
$$
71+
f(x) = f(a) + f'(a)(x-a) + \frac{1}{2}f''(a)(x-a)^2
72+
$$
73+
74+
or if we write this in the terms of $\triangle x$, we get
75+
$$
76+
f(x+\triangle x) = f(x) + f'(x)\triangle x + \frac{1}{2}f''(x)\triangle x^2
77+
$$
78+
79+
This same method can then be extended to higher degrees to get higher order approximations of a function. That's the **Taylor Series**.
80+
## Taylor Series
81+
The Taylor series provides a higher order approximation for a function's value around some neighbourhood of any point $x$.
82+
83+
$$
84+
\begin{aligned}
85+
f(x+\triangle x) &= f(x) + f'(x) \triangle x + \frac{1}{2!}f''(x) \triangle x^2 + \frac{1}{3!}f'''(x) \triangle x^3 + \dots \\[8pt]
86+
&= \sum_{i=0} \frac{1}{i!} f^{(i)}(x) \triangle x^i
87+
\end{aligned}
88+
$$
89+
# Cauchy Schwarz Inequality
90+
# Directional Derivative
91+
Direction of steepest descent is $-\nabla f(x^*)$. This can be shown by either using the [[Basics of Calculus#Cauchy Schwarz Inequality|Cauchy Schwarz Inequality]] or [[Basics of Calculus#Taylor Series|Taylor Series]].

content/Mathematical Foundations of Machine Learning I/Calculus/Convexity.md

Lines changed: 0 additions & 16 deletions
This file was deleted.
Lines changed: 172 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,172 @@
1+
>[!SUMMARY] Table of Contents
2+
>- [[Optimization#Gradient Descent|Gradient Descent]]
3+
>- [[Optimization#Convexity|Convexity]]
4+
> - [[Optimization#Convex Sets|Convex Sets]]
5+
> - [[Optimization#Properties -|Properties -]]
6+
> - [[Optimization#Convex Combinations|Convex Combinations]]
7+
> - [[Optimization#Convex Functions|Convex Functions]]
8+
> - [[Optimization#Properties -|Properties -]]
9+
>- [[Optimization#Questions|Questions]]
10+
# Gradient Descent
11+
\<Intentionally left blank for now\>
12+
# Convexity
13+
## Convex Sets
14+
A set is a convex set if $\forall x_1, x_2 \in S$, $\lambda x_1 + (1 - \lambda)x_2 \in S$ where $\lambda \in [0,1]$. Geometrically speaking, the below sets are a convex set -
15+
16+
![[Pasted image 20260217104309.png|550]]
17+
18+
### Properties -
19+
1. If $A$ and $B$ are two convex sets then $A \cap B$ is also a convex set.
20+
$$
21+
\begin{aligned}
22+
\text{Let } x_1 \text{ and } x_2 \in A \cap B. &\text{We can say,} \\[8pt]
23+
\lambda x_1 + (1 - \lambda)x_2 &\in A \text{ because } x_1, x_2 \in A \\[8pt]
24+
\lambda x_1 + (1 - \lambda)x_2 &\in B \text{ because } x_1, x_2 \in B \\[8pt]
25+
\text{Thus we can say that } x_1, &x_2 \in A \cap B.
26+
\end{aligned}
27+
$$
28+
29+
This property is helpful in showing if a set is convex, by showing that the set is formed by intersection of two other convex sets.
30+
## Convex Combinations
31+
Let $S = \{x_1, x_2, \dots, x_n\} \subseteq \mathbb R ^n$. Then we say that $z \in \mathbb R ^n$ is a convex combination of vectors in $S$ if $\exists \lambda_1, \lambda_2, \dots, \lambda_n$ such that $\lambda_i \ge 0$ and $\sum \lambda_i = 1$ and
32+
33+
$$
34+
z = \lambda_1 x_1 + \lambda_2x_2 + \dots + \lambda_nx_n
35+
$$
36+
37+
The **convex hull** of a set $S$, denoted by $\operatorname{conv}(S)$ or $\operatorname{CH}(S)$ is the set of all convex combinations of the elements of the set $S$.
38+
39+
$$
40+
CH(\{x_1, x_2, \dots, x_n\}) = \left\{z\,\,|\,\,z=\sum_{i=1}^n\lambda_ix_i, \lambda_i \ge 0 \text{ and } \sum_{i=1}^n \lambda_i = 1\right\}
41+
$$
42+
43+
An alternate definition of a convex hull can be that, the convex hull of a set is the intersection of all convex sets that contain $\{x_1, x_2, \dots, x_n\}$. The two definitions can be shown to be equivalent.
44+
45+
**Important Exercise**[Show that Euclidean Balls are Convex Sets](https://youtu.be/PfDvKJ2UTPg?list=PLZ2ps__7DhBammhVmBE9f5eezTj2kDfTN&t=1004)
46+
## Convex Functions
47+
$\underline{\text{Definition 1}}$ – A function $f: \mathbb R^d \rightarrow \mathbb R$ is a convex function iff $\operatorname{epi}(f) \in \mathbb R^{d+1}$ is a convex set.
48+
49+
$\operatorname{epi}$ stands for **epigraph** of a function. An epigraph is the set of all points above the graph's curve -
50+
51+
$$
52+
\begin{aligned}
53+
\operatorname{epi}(f) = \{(x,y) \in \mathbb R^n \times \mathbb R \,\,|\,\, t \ge f(x)\}
54+
\end{aligned}
55+
$$
56+
57+
The opposite of an epigraph is a **hypograph**.
58+
59+
$\underline{\text{Definition 2}}$ – A function $f: \mathbb R^d \rightarrow \mathbb R$ is a convex function iff $\forall x_1, x_2 \in \mathbb R^d$ and all $\lambda \in [0,1]$ ^bec74f
60+
61+
$$
62+
\begin{aligned}
63+
f(\lambda x_1 + (1-\lambda x_2)) &\le \lambda f(x_1) + (1-\lambda)f(x_2) \\[8pt]
64+
\text{OR more} & \text{ generalized} \\[8pt]
65+
f\left(\sum_{k=1} \lambda_k a_k\right) &\le \sum_{k=1} \lambda_k f(a_k)
66+
\end{aligned}
67+
$$
68+
69+
For concave functions this inequality becomes,
70+
71+
$$
72+
f\left(\sum_{k=1} \lambda_k a_k\right) \ge \sum_{k=1} \lambda_k f(a_k)
73+
$$
74+
75+
76+
![[Pasted image 20260222181747.png|450]]
77+
78+
$\underline{\text{Definition 3}}$ – A function $f: \mathbb R^d \rightarrow \mathbb R$ is a convex function iff $f$ is differentiable and, ^104fc1
79+
80+
$$
81+
f(y) \ge f(x) + \nabla f(x)^T(y-x) \qquad \forall x,y \in \mathbb R^d
82+
$$
83+
84+
![[Pasted image 20260222181819.png|450]]
85+
86+
This means that the tangent plane lower bounds the function for every point in its domain.
87+
88+
$\underline{\text{Definition 4}}$ – A function $f: \mathbb R^d \rightarrow \mathbb R$ is a convex function iff it is twice differentiable.
89+
### Properties -
90+
91+
<h4>
92+
1) In a convex function f, x* is a global minima iff its gradient is 0.
93+
</h4>
94+
95+
$\underline{\text{Proof}} -$
96+
$(1)$ We need to first show that if $x^*$ is a global minima, $\nabla f(x^*) = 0$. This can be shown easily by using the concept of steepest descent. At any point, the direction of steepest descent is $-\nabla f(x^*)$. So if $\nabla f(x) \ne 0$ for any point $x$, we can show that there exists another point in the direction of [[Basics of Calculus#Directional Derivative|steepest descent]] such that $f(x - \eta \nabla f(x^*)) \lt f(x)$. Thus if $\nabla f(x^*) \ne 0$ then $x^*$ can't be a minima, let alone a global minima.
97+
98+
Thus points where no such direction of steepest descent would exist will have $\nabla f(x^*) = 0$. Thus the gradient at the global minima must be $0$.
99+
100+
$(2)$ Now we need to show the reverse, that in convex functions if $\nabla f(x^*) = 0$ then $x^*$ is a global minima. We know by [[#^104fc1|Definition 3]] that the tangent plane at any point of the graph lower bounds the function. So even at the global minima $x^*$ we can say,
101+
102+
$$
103+
\begin{alignedat}{3}
104+
&& f(y) &\ge f(x) + \nabla f(x)^T(y-x) \qquad \forall x,y \in \mathbb R^d\\[8pt]
105+
&\Rightarrow & f(y) &\ge f(x^*) + \nabla f(x^*)^T(y-x^*) \\[8pt]
106+
&\Rightarrow & f(y) &\ge f(x^*) \\[8pt]
107+
\end{alignedat}
108+
$$
109+
110+
Hence we can say that $x^*$ is a global minima.
111+
112+
<h4>
113+
2) If f and g are two convex functions, then h(x) = f(x) + g(x) is also a convex function.
114+
</h4>
115+
116+
This can be proved easily using [[#^bec74f|Jensen's Inequality]] of convex functions. Try it out!
117+
<h4>
118+
3) If f is a convex and non-decreasing function and g is any convex function, then h(x) = fog(x) is also a convex function.
119+
</h4>
120+
121+
This can again be proved easily using [[#^bec74f|Jensen's Inequality]] of convex functions. Try it out!
122+
123+
<h4>
124+
4) If f is a convex function and g is a linear function, then h(x) = fog(x) is also a convex function.
125+
</h4>
126+
This can again be proved easily using [[#^bec74f|Jensen's Inequality]] of convex functions along with the property of linear functions. Try it out!
127+
128+
**Note -** In general if $f$ and $g$ are convex, the composition $h(x) = fog(x)$ may not be convex.
129+
130+
---
131+
# Questions
132+
<h6 class="question">Q1) Show that log is a concave function -</h6>
133+
$\underline{\text{Sol}^n} -$
134+
We can use the reverse of [[#^bec74f|Jensen's Inequality]] for this purpose. If we are able to show that,
135+
136+
$$
137+
\operatorname{log}(\lambda x + (1 - \lambda)y) \ge \lambda \operatorname{log}(x) + (1 - \lambda) \operatorname{log}(y) \qquad x,y > 0, \lambda \in [0,1]
138+
$$
139+
140+
For this we can try removing the $\operatorname{log}$ from both sides and put everything to the left side and show that the unified equation is greater than or equal to $0$. So we can rewrite the above requirement as,
141+
142+
$$
143+
\begin{alignat*}{3}
144+
&&\operatorname{log}(\lambda x + (1 - \lambda)y) &\ge \lambda \operatorname{log}(x) + (1 - \lambda) \operatorname{log}(y) \\[8pt]
145+
&\Rightarrow &\,\,\operatorname{log}(\lambda x + (1 - \lambda)y) &\ge \operatorname{log}(x^\lambda) + \operatorname{log}(y^{(1 - \lambda)}) \\[8pt]
146+
&\Rightarrow &\,\,\operatorname{log}(\lambda x + (1 - \lambda)y) &\ge \operatorname{log}(x^\lambda \cdot y^{(1 - \lambda)}) \\[8pt]
147+
&\Rightarrow &\lambda x + (1 - \lambda)y &\ge x^\lambda \cdot y^{(1 - \lambda)} \\[8pt]
148+
&\Rightarrow &\lambda \frac{x}{y} + (1 - \lambda) &\ge x^\lambda \cdot y^{- \lambda} \\[8pt]
149+
&\Rightarrow &\lambda t + (1 - \lambda) &\ge t^\lambda &\because t=\frac{x}{y} \\[8pt]
150+
&\Rightarrow &\lambda t + (1 - \lambda) - t^\lambda &\ge 0 \tag{1} \\[8pt]
151+
\end{alignat*}
152+
$$
153+
154+
We can differentiate this term to get its minima, and if the minima is $\ge 0$ then we can say that the entire equation is always non-negative.
155+
156+
$$
157+
\begin{alignedat}{3}
158+
&& q(t) &= \lambda t + (1 - \lambda) - t^\lambda \\[8pt]
159+
&\Rightarrow& \,\,q'(t) &= \lambda + - \lambda t^{\lambda - 1} \\[8pt]
160+
\end{alignedat}
161+
$$
162+
163+
For the some $t^*$ to be the minima of $q$, $q'(t^*) = 0$. So,
164+
165+
$$
166+
\begin{alignedat}{3}
167+
&& q'(t^*) &= \lambda + - \lambda t^{*^{\lambda - 1}} \\[8pt]
168+
&\Rightarrow& \,\, 0 &= \lambda(1 - t^{*^{\lambda - 1}}) \\[8pt]
169+
\end{alignedat}
170+
$$
171+
172+
If $\lambda \ne 0$ then then only $t^* = 1$ can satisfy the above equation. Because $q(1) = 0$, we can say that $\forall t > 0, q(t) \ge 0$. Hence we have proven that $(1)$ is true and consequently proven that $\operatorname{log}$ is a concave function.

0 commit comments

Comments
 (0)