##
**Online Learning** via Prediction with **Experts' Advice**

**Victor Sanches Portella**

UBC - CS

June 2022

##
**Experts**' Problem and **Online Learning**

### Prediction with Experts' Advice

Player

Adversary

**\(n\)** Experts

0.5

0.1

0.3

0.1

Probabilities

1

0

0.5

0.3

Costs

**Player's loss:**

Adversary **knows** the strategy of the player

Picking a **random expert**

vs

Picking a **probability vector**

### Measuring Player's Perfomance

Total player's loss

Can be = \(T\) always

Compare with offline optimum

Almost the same as Attempt #1

Restrict the offline optimum

**Attempt #1**

**Attempt #2**

**Attempt #3**

Loss of Best Expert

Player's Loss

**Goal:**

### Example

### Cummulative Loss

### Experts

### General Online Learning

Player

Adversary

**Player's loss:**

**Convex**

**The player sees \(f_t\)**

Simplex

Linear functions

Some usual settings:

Experts' problem

### Why Online Learning?

Traditional ML optimization makes **stochastic assumptions on the data**

OL Â **strips away**Â the stochastic layer

Traditional ML optimization makes **stochastic assumptions on the data**

Less assumptions \(\implies\) **Weaker guarantees**

Less assumptions \(\implies\) **More robust**

Adaptive algorithms

**AdaGrad**

**Adam**

Parameter Free algorithms

**Coin Betting**

Meta-optimization

**Algorithms**

### Follow the Leader

**Idea**: Pick the best expert at each round

**where \(i\) minimizes**

**Can fail badly**

Player loses \(T -1\)

Best expert loses \(T/2\)

Works **very well** for quadratic losses

* picking distributions instead of best expert

### Gradient Descent

\(\eta_t\): step-size at round \(t\)

\(\ell_t\): loss vector at round \(t\)

**Sublinear** Regret!

**Optimal** dependency on \(T\)

Can we improve the dependency on \(n\)?

**Yes, and by a lot**

### Multiplicative Weights Update Method

Normalization

**Exponential** improvement on \(n\)

**Optimal**

Other methods had clearer "optimization views"

Rediscovered many times in different fields

This one has an optimization view as well!

### Follow the Regularized Leader

Regularizer

"Stabilizes the algorithm"

"Lazy" Gradient Descent

Multiplicative Weights Update

Good choice of \(R\) depends of the functions and the feasible set

### Online Mirror Descent

projection

### Online Mirror Descent

### Dual

### Primal

**Bregman**

**Projection**

Regularizer

**GD:**

**MWU:**

### OMD and FTRL are quite general

**Applications**

### Approximately Solving Zero-Sum Games

Payoff matrix \(A\) of row player

**Row player**

**Column player**

Strategy \(pÂ = (0.1~~0.9)\)

Strategy \(qÂ = (0.3~~0.7)\)

**Von Neumman min-max Theorem:**

Row player

picks row \(i\) with probability \(p_i\)

Column player

picks column \(j\) with probability \(q_j\)

Row player

gets \(A_{ij}\)

Column player

gets \(-A_{ij}\)

### Approximately Solving Zero-Sum Games

**Main idea: **make each row of \(A\) be an expert

For \(t = 1, \dotsc, T\)

\(p_1 =\) uniform distribution

Loss vector \(\ell_t\) is the \(j\)-th col. of \(-A\)

Get \(p_{t+1}\) via Multiplicative Weights

**Thm:**

**Cor:**

where \(j\) maximizes \(p_t^T A e_j\)

\(q_t = e_j\)

\(\bar{p} = \tfrac{1}{T} \sum_{t} p_t\)

\(\bar{q} = \tfrac{1}{T} \sum_{t} q_t\)

and

### Boosting

**Training set**

**Hypothesis class**

of functions

**Weak learner:**

such that

**Question:** Can we get with high probability a hypothesis* \(h^*\) such that

**only on a \(\varepsilon\)-fraction of \(S\)?**

**Generalization** follows if \(\mathcal{H}\) is simple (and other conditions)

### Boosting

For \(t = 1, \dotsc, T\)

\(p_1 =\) uniform distribution

Â \(\ell_t(i) = 1 - 2|h_t(x_i) - y_i|\)

Get \(p_{t+1}\) via Multiplicative Weights (with right step-size)

\(h_t = \mathrm{WL}(p_t, S)\)

\(\bar{h} = \mathrm{Majority}(h_1, \dotsc, h_T)\)

**Theorem**

If \(T \geq (2/\gamma^2) \ln(1/\varepsilon)\), then \(\bar{h}\) makes at most \(\varepsilon\) mistakes in \(S\)

Main ideas:

Regret only against distrb. on examples that \(\bar{h}\) errs

Due to **WL** property, loss of the player is \(\geq 2 T \gamma\)

\(\ln n\) becomes \(\ln (n/\mathrm{\# mistakes}) \)

Cost of any distribution of this type is \(\leq 0\)

### From Electrical Flows to Maximum Flow

**Goal:**

Route **as much flow as possible** from \(s\) to \(t\) while respecting the **edges' capacities**

We can compute in time \(O(|V| \cdot |E|)\)

This year there was a paper with a \(O(|E|^{1 + o(1)})\) alg...

What if we want something faster even if approx.?

### From Electrical Flows to Maximum Flow

**Fast Laplacian system solvers **(Spielman & Teng' 13)

We can compute **electrical flows** by solving this system

Electrical flows may not respect edge capacities!

Solves

in \(\tilde{O}(|E|)\) time

Laplacian matrix of \(G\)

**Main idea:** Use electrical flows as a "weak learner", and boost it using MWU!

**Edges = Experts**

**Cost = flow/capacity**

### Other applications (beyond experts)

Solving Packing linear systems with oracle access

Approximating multicommodity flow

Approximately solve some semidefinite programs

Spectral graph sparsification

Approximating multicommodity flow

Computational complexity (QIP = PSPACE)

##
**Anytime** Regret and **Stochastic Calculus**

### Fixed-time vs Anytime

**MWU** regret

when \(T\) is known

when \(T\) is **not** known

**Â anytimeÂ **

**fixed-time**

Does knowing \(T\) gives the player an advantage?

Mirror descent in anytime can fail terribly

[Huang, Harvey, **VSP**, Friedlander]

Optimum for 2 experts anytime is worse

[Harvey, Liaw, Perkins, Randhawa]

Similar techniques also work for fixed-time

[Greenstreet, Harvey, **VSP**]

Continuous-time experts show interesting behaviour

[**VSP**, Liaw, Harvey]

### Random Experts' Costs

What happens when we randomize experts' losses?

Culmulative loss

of expert \(i\) is a **random walk**

**independently**

Worst-case adversary:

Regret bound hold with **probability 1**

Regret bound against **any adversary**

**Idea: **Move to continuous time to use powerfulÂ **stochastic calculus** tools

Worked very well with 2 experts

### Moving to Continuous Time

Moving to **continuous time**:

Random walk \(\longrightarrow\) Brownian Motion

are Independent Brownian motions

where

### Moving to Continuous Time

**Discrete time**

**Continuous time**

**Cummulative loss**

**Player's cummulative loss**

**Player's loss per roundÂ **

### MWU in Continuous Time

**Potential based players**

**Regret bounds**

when \(T\) is known

when \(T\) is **not** known

**Â anytimeÂ **

**fixed-time**

**MWU!**

**Same as discrete time!**

**Idea:** Use stochastic calculus to guide the algorithm design

**with prob. 1**

### New Algorithms in Continuous Time

**Potential based players**

**Matches fixed-time!**

Stochastic calculus suggests pontential that satisfy the **Backwards Heat Equation**

This new algorithm has good regret!

Does not translate easily to discrete time

Â need to add **correlation between experts**

New algorithms with good quantile regret bounds!

**Fixed-time vs Anytime is still open!**

##
**Online Learning** via Prediction with **Experts' Advice**

**Victor Sanches Portella**

UBC - CS

June 2022

## Kitchen Sink

### Simplifying Assumptions

We will look only at \(\{0,1\}\) costs

1

0

0

1

0

0

1

1

Equal costs do not affect the regret

Cover's algorithm **relies** on these assumptions **by construction**

Our alg. and analysis extends to **fractional costs**

### Gap between experts

**Thought experiment:** how much probability mass to put on each expert?

**Cumulative Loss **on round \(t\)

\(\frac{1}{2}\) is both cases seems **reasonable**!

**Takeaway:** player's decision may depend **only** on the **gap** between experts's losses Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â

**Gap =** |42 - 20| = 22

**Worst **Expert

**Best **Expert

42

20

2

2

42

42

(and maybe on \(t\))

### Cover's Dynamic Program

**Player strategy based on gaps:**

Choice doesn't depend on the specific past costs

on the **Worst** expert

on the **Best** expert

Â We can compute \(V^*\) backwards in time via **DP**!

**Max regret to be suffered** at time \(t\) with gap \(g\)

\(O(T^2)\) time to compute \(V^*\)

At **round** \(t\) with **gap** \(g\)

**Max. regret for a game with \(T\) rounds**

Computing the optimal strategy \(p^*\) from \(V^*\) is easy!

Cover's DP Table

(w/ player playing optimally)

### Cover's Dynamic Program

**Player strategy based on gaps:**

Choice doesn't depend on the specific past costs

on the **Lagging** expert

on the **Leading** expert

Â We can compute \(V^*\) backwards in time via **DP**!

Getting an **optimal player** \(p^*\) from \(V^*\) is easy!

Max **regret-to-be-suffered** at round \(t\) with gap \(g\)

\(O(T^2)\) time to compute the table â€” \(O(T)\) amortized time per round

At **round** \(t\) with **gap** \(g\)

Optimal regret for 2 experts

### Connection to Random Walks

Optimal player \(p^*\) is related to Random Walks

For \(g_t\) following a Random Walk

Central Limit Theorem

Not clear if the **approximation error** affects the regret

The DP is defined only for **integer costs**!

**Lagging **expert finishes **leading**

Let's design an algorithm that is efficient and works for all costs

**Bonus:** Connections of Cover's algorithm with stochastic calculus

### Connection to Random Walks

**Theorem**

Player \(p^*\) is also connected to RWs

For \(g_t\) following a Random Walk

Central Limit Theorem

Not clear if the approximation error affects the regret

The DP is defined only for integer costs!

**Lagging **expert finishes **leading**

[Cover '67]

# of 0s of a Random Walk of len \(T\)

Let's design an algorithm that is efficient and works for all costs

**Bonus:** Connections of Cover's algorithm with stochastic calculus

### Discrete ItÃ´'s Formula

How to analyze a **discrete** algorithm coming from **stochastic calculus**?

Discrete ItÃ´'s Formula!

Discrete Derivatives

Surprisingly, we can analyze **Cover's algorithm** with discrete **ItÃ´'s formula**

**ItÃ´'s Formula**

**Discrete ItÃ´'s Formula**

### Discrete Algorithms

\(V^*\)** **satisfies the **"discrete" Backwards Heat Equation!**

**Not Efficient**

**Efficient**

**Discrete ItÃ´ \(\implies\)**

Regret of \(p^* \leq V^*[0,0]\)

**BHE = Optimal?**

Hopefully, **\(R\)** satisfies the **discrete BHE**

**Discretized player:**

We show the **total** is \(\leq 1\)

Cover's strategy

### Bounding the Discretization Error

In the work of Harvey et al., they had

In this fixed-time solution, we are not as lucky.

*Negative discretization error!*

We show the **total** discretization error is always \(\leq 1\)

### A Probabilistic View of Regret Bounds

Formula for the regret based on the **gaps**

Random walk \(\longrightarrow\) Brownian Motion

Reflected Brownian motion** (gaps)**

Conditions on the *continuous player* **\(p\)**

Continuous on \([0,T) \times \mathbb{R}\)

for all \(t \geq 0\)

### Stochastic Integrals and ItÃ´'s Formula

How to work with stochastic integrals?

**ItÃ´'s Formula:**

\(\overset{*}{\Delta} R(t, g) = 0\) everywhere

Â ContRegret \( = R(T, |B_T|) - R(0,0)\)

**Goal:**

Find a "**potential function**" \(R\) such that

(1) **\(\partial_g R\)** is a valid continuous player

(2) \(R\) satisfies the **Backwards Heat Equation**

Different from classic FTC!

**B**ackwards** H**eat** E**quation

### Stochastic Integrals and ItÃ´'s Formula

**Goal:**

Find a "**potential function**" \(R\) such that

(1) **\(\partial_g R\)** is a valid continuous player

(2) \(R\) satisfies the **Backwards Heat Equation**

How to find a good \(R\)?

?

Suffices to find a player \(p\) satisfying the **BHE**

\(\approx\) Cover's solution!

**Also a solution to an ODE**

Then setting

preserves **BHE **and

### Stochastic Integrals and ItÃ´'s Formula

How to work with stochastic integrals?

**ItÃ´'s Formula:**

\(\overset{*}{\Delta} R(t, g) = 0\) everywhere

Â ContRegret is given by \(R(T, |B_T|)\)

**Goal:**

Find a "**potential function**" \(R\) such that

(1) **\(\partial_g R\)** is a valid continuous player

(2) \(R\) satisfies the **Backwards Heat Equation**

Different from classic FTC!

**Backwards Heat Equation**

[C-BL 06]

### Our Results

An **Efficient** and **Optimal** Algorithm in Fixed-Time with Two Experts

**Technique**:

Solve an analogous continuous-time problem, and discretize it

[HLPR '20]

How to exploit the knowledge of \(T\)?

Discretization error needs to be analyzed carefully.

BHE seems to play a role in other problems in OL as well!

Solution based on Cover's alg

**Or inverting time in an ODE!**

We show \(\leq 1\)

\(V^*\) and \(p^*\) satisfy the **discrete BHE**!

**Insight**:

Cover's algorithm has connections to **stochastic calculus**!

### Questions?

### Known Results

**M**ultiplicative **W**eights **U**pdate method:

**Optimal** for \(n,T \to \infty\) !

If \(n\) is fixed, we **can do better**

**\(n = 2\)**

**\(n = 3\)**

**\(n = 4\)**

Player **knows **\(T\) !

**Minmax** regret in some cases:

What if \(T\) is **not known?**

Minmax regret

**\(n = 2\)**

[Harvey, Liaw, Perkins, Randhawa FOCS 2020]

They give an **efficient** algorithm!

### A Dynamic Programming View

Optimal regret (\(V^* = V_{p^*}\))

For \(g > 0\)

For \(g = 0\)

### Regret and Player in terms of the Gap

**Path-independent player:**

If

round \(t\) and gap \(g_{t-1}\) on round \(t-1\)

on the **Lagging** expert

on the **Leading** expert

Choice doesn't depend on the specific past costs

for all \(t\), then

gap on round \(t\)

A discrete analogue of a Riemann-Stieltjes integral

**A formula for the regret**

### A Dynamic Programming View

Maximum **regret-to-be-suffered** on rounds \(t+1, \dotsc, T\) when **gap on round \(t\) is \(g\)**

Path-independent player \(\implies\) \(V_p[t,g]\) depends **only** on \(\ell_{t+1}, \dotsc, \ell_T\) and \(g_t, \dotsc, g_{T}\)

Regret suffered on round \(t+1\)

Regret suffered on round \(t + 1\)

### A Dynamic Programming View

Maximum **regret-to-be-suffered** on rounds \(t+1, \dotsc, T\) if **gap at round \(t\) is \(g\)**

Â We can compute \(V_p\) backwards in time!

Path-independent player \(\implies\)

Â \(V_p[t,g]\) depends **only** on \(\ell_{t+1}, \dotsc, \ell_T\) and \(g_t, \dotsc, g_{T}\)

We then choose \(p^*\) that minimizes \(V^*[0,0] = V_{p^*}[0,0]\)

Maximum regret of \(p\)

### A Dynamic Programming View

For \(g > 0\)

**Optimal player**

**Optimal regret** (\(V^* = V_{p^*}\))

For \(g = 0\)

For \(g > 0\)

For \(g = 0\)

### Discrete Derivatives

### Bounding the Discretization Error

**Main idea**

\(R\) satisfies the **continuous BHE**

Approximation error of the derivatives

**Lemma**

### Known and New Results

**M**ultiplicative **W**eights **U**pdate method:

**Optimal** for \(n,T \to \infty\) !

If \(n\) is fixed, we **can do better**

**Worst-case** regret for **2 experts**

Player **knows** \(T\) (**fixed-time**)

Player **doesn't** know \(T\) (**anytime**)

**Question: **

Is there an **efficient** algorithm for the **fixed-time** case?

Ideally an algorithm that works for **general costs**!

\(O(T)\) time per round

Dynamic Programming

\(\{0,1\}\) costs

\(O(1)\) time per round

Stochastic Calculus

\([0,1]\) costs

[Harvey, Liaw, Perkins, Randhawa FOCS 2020]

[Cover '67]

### Our Results

**Result**:

An Efficient and Optimal Algorithm in Fixed-Time with Two Experts

\(O(1)\) time per round

was \(O(T)\) before

Holds for general costs!

**Technique**:

Discretize a solution to a stochastic calculus problem

[HLPR '20]

How to exploit the knowledge of \(T\)?

Non-zero discretization error!

**Insight**:

Cover's algorithm has connections to **stochastic calculus**!

This connection seems to extend to more experts and other problems in online learning in general!

#### IAM Student seminar - OL via the Expert's problem

By Victor Sanches Portella

# IAM Student seminar - OL via the Expert's problem

- 137