1. draw an initial state \(q_1\) from the initial state distribution \(\pi\)
2. For each state \(q_{i}\)…
   1. Drew observe something \(o_{t}\) according to the action distribution of state \(q_{i}\)
   2. Use transition probability \(a_{i,j}\) to draw a next state \(q_{j}\)

Isolated recognition: train a family of HMMs, one for each word or something. Then, given new data, perform scoring of the HMM onto the features.

components of HMMs

scoring

Given an observation \(o_1, …, o_{T}\) and a model, we compute $P(O | λ)$—the probability of a sequence given a model \(\lambda\)

Misinformation can decline people’s intent to vaccinate.

Data

VaxConcerns Taxonomy: sorting misinformation into a taxonomy which has multiple, hierarchical labels

So, you can classify a label in a bunch of places.

Approaches

Ask an LLM to do one of the following…

One-Shot, Multi-Label

Ignore hierchy, just go do multi label system

multi-pass, hierarchical label

try to predict groups at a time, each pass has a label group as input and needs to produce whether the label is given + sublabels

a monomer is named active if it recruits its neighbors to make a polymer that can act as an enzyme to catalyze other monomers to become active as well

If we are tele-operating a robot, we ideally want to minimize cost. We want to estimate a user’s goal via user inputs. Predict the most likely goal + assist for it.

“find a cost function for which user input \(u\) is optimal”.

• system does not know the goal
• the user may not change their goal on a whim

Hindsight Optimization

To solve this, we use QMDP: “select the most optimal actions to estimating cost-to-go assuming full observability”.

history

economics

• utility theory —

psychology

• Pavlov’s salivating togs: biological reinforcement learning
• Alan Turing thought about this regarding

neuroscience

• bio-simulated things

computer science

• how “systems composed of matter can have properties of the mind

engineering

• control theory

maffs

• statistics

operations research

• WWII logistics optimization
• allocating decisions, etc.
• simplex algorithm

societal impact

• classes can help amplify intention (whether good or evil)
• data-driven methods