A markov game with State Uncertainty solved using POMDPs.

DOI: 10.3389/fnagi.2020.605317

One-Liner

An excercize scheme has had some measured effect on theta/alpha ratio and Brain wave frequency on AD patients; prognosis of AD not controlled for.

Novelty

• Leveraged physical training scheme and measured EEG effects by quantifying theta/alpha ratio

Notable Methods

• Used theta/alpha ratio as assay for improvement, and found the exercise scheme did so p<0.05
• Only tested patients with AD w/o a control for stage

Key Figs

Figure 1

This figure tells us th N number of participants through the study

Patient Scoring Systems

How do we score the status of a patient? Well, we can begin by having a chart—SpO2, can breath, etc. etc.

Drawbacks:

1. these systems are quite generic
2. not very representative of some information

Method

1. MIMIC-IV 6000 ICU patient stays, 48994 vital signs—measuring across patient stays
2. dynamic time warping to create a similar matrix
3. clustering post-hoc to correlate patients together

• No Demo Day
• TODO Email need statement template

Needfinding

• Not all patients want to be treated the same way
• Attitudes towards heathcare system
• Fostering strong interaction; facilitate interaction

Problem: patients have attitudes that physicians can’t effectively communicate.

Action item: interview doctors and patients

Need two need statement.

PEFT is parameter efficient fine-tuning.

LoRA

Consider some matrix:

\begin{equation} W_0 \in \mathbb{R}^{d \times k} \end{equation}

Key intuition: gradient matricies have low intrinsic rank. We consider the following update:

\begin{equation} W_0 + \Delta W = W_0 + \alpha BA \end{equation}

where \(B \in \mathbb{R}^{d \times r}, A \in \mathbb{R}^{r \times k}\), and \(r \ll \min(d,k)\).

where \(\alpha\) is the trade off between pre-trained knowledge and task specific knowledge.