Bayesian Networks - Questions
Question 1
A standard test for prostate cancer, the prostate specific antigen (PSA) test,
has a sensitivity of around 95% and a specificity of around 55%.(The sensitivity
is the conditional probability of a positive test resu

Sen$ment Analysis
?
I love this lecture,
but the movie last
night wasnt great.
Presented by:
Dor Cohen
Gil Elbaz
Overview
o Mo$va$on
o Problem Statement
o Sen$ment Analysis using ML classiers
o Sen$ment An

Markov Networks - Questions
Question 1
Consider the distribution P over the random variables: V, T, S, L, B, A, X, D.
Assume that the distribution factorizes as
P(V, T, S, L, B, A, X, D) = Z1 V,T (v, t)L,T (l, t)S,L (l, s)S,B (b, s)L,A (l, a)
T,A (t, a) A

Variable Elimination - Questions
Question 1
Consider the Bayesian Network given in the figure above (1). Assume that
each of the variables are boolean valued. For each of the following, state the
total number of operations(multiplication and addition) the

Belief Propagation Algorithm - Questions
Question 1
1. Suppose we wish to perform inference over the Markov network M as
shown below. Each of the variables Xi are binary, and the only potentials
in the network are the pairwise potentials (Xi , Xj ), with

Introduction to Probabilistic Graphical Models Questions
Question 1
Prove the following claims:
1. (X cfw_Y, W |Z) (X Y |Z)
2. (X Y |Z) and (X, Y W |Z) (X W |Z)
3. (X cfw_Y, W |Z) and (Y W |Z) (cfw_X, W Y |Z)
Question 2
Answer the following questions:
1.

Hillel Mendelson
Igal Shprincis
Liron Shalom
Extracting sentiment from text is a challenging problem with
applications throughout NLP and IR.
systems have tackled the problem at different levels of granularity,
from the document level, sentence level, p

Agenda
POS Tagging Problem
Named Entity Recognition
Generative Tagging Models
Markov Models
Hidden Markov Models (HMM)
Trigram HMM
POS Tagging problem
The most famous problem in NLP is Part-of-speech (POS) tagging.
In POS tagging our goal is to bu

Generative Models
Presented by Dima Kovaliov and Shay Rodes
Motivation
Until now - Discriminative modeling:
We did not impose any assumptions on the underlying
distribution over the data.
Our goal was to learn an accurate predictor.
Discriminative modelin

CRF
Conditional Random Fields
Benjamin Eshet
Tal Ariel
Overview
Motivation for CRF
Linear Chain CRF
Training
Inference
Recall
HMM
, =
+1
(generative)
MEMM
| =
+1
(discriminative)
Motivation
Sequencing problems - consider the context
Linear Chai

Machine Learning - 097209
Homework Assignment 2
Due Date: 23.6.2016
(1) All students must submit in threesomes. Please submit only once
in moodle.
(2) Consulting students from different threesome is not allowed.
(3) Searching the Internet for solutions is