===== Methods =====

==== Jason Sanders - Angle-Frequency Approach ====

Construct a model for a single stream by using information about the form of the stream in angle-frequency space in the true potential. The model favours two narrow clumps in frequency space aligned with the angle structure.

==== Robyn Sanderson - Information maximization ====

Maximizes the [[http://en.wikipedia.org/wiki/Mutual_information | mutual information]] of the action distribution of stream stars to determine the best-fit potential. 

Needs:
  * full phase-space coordinates for stars in more than one stream (preferably at least 10-15, depending on observational errors)
  * at least ~100 stars in the largest stream
  * Enough stars per stream that the largest stream does not solely determine the fit
  * the major component of the sample to be accreted streams (i.e. background must be <50%)
Does not need:
  * number of streams in sample or stream membership assignments
  * all stars to be in streams
  * complete sampling of any particular stream
Steps:
  - Compute actions from phase-space coordinates given a potential
  - Maximize mutual information by varying potential to obtain best-fit
  - Use the [[http://en.wikipedia.org/wiki/Kullback%E2%80%93Leibler_divergence|KLD]] again to compare action distribution in best-fit potential to distribution in other potentials to get uncertainty



==== Andreas Küpper/Ana Bonaca - Streakline Method ====

From present-day position & velocity of the progenitor of interest, integrate the orbit of a cluster particle of a given mass in a test potential backwards for several Gyr. Then integrate forward to present time. While doing so, generate test stream particles. Compare the present-day distribution of stream test particles to the observations. Define likelihood, calculate likelihood, put in MCMC, stirr, ready.

==== Adrian Price-Whelan - Orbit Rewinder ====

Need: 
  * Observations of progenitor (can have missing data): l,b,dist.,**μ**,RV
  * Constraint on the mass of the progenitor
  * Observations of a sample of stars stripped from the progenitor:  (again) l,b,dist.,**μ**,RV 
  * A flag to specify which tail (leading/trailing) each star belongs to

Integrate orbits of progenitor and stars //back// in time; the stars recombine into the progenitor in the correct potential, but their orbits diverge in an incorrect potential. See [[http://arxiv.org/abs/1308.2670|Price-Whelan & Johnston (2013)]] for the basic idea, or [[http://arxiv.org/abs/1405.6721|Price-Whelan et al. (2014)]] for the new idea for accounting for observational uncertainties.
==== Jorge Peñarrubia ====

Minimize scatter/entropy of tidal streams.


==== Nathan Deg ====

We model the stream as though it is generated by the orbit of a single particle in a fixed potential.  We find the full PDF of a particular model using a Bayesian analysis.  See [[http://adsabs.harvard.edu/abs/2014arXiv1401.4070D]] for a full description of the technique.

==== Hans Buist ====

Look at evolution of galaxy potential and streams. For now: look at "progenitor" orbit compared with stream.


==== Daniele Fantin ====

  * Stream finder: modified great circle cell method (Mateu et al 11)
  * Method for revocering local potential based on energy conservation
==== Hao Tian ====

Working on recovery of stellar streams using a phase-space halo finder ROCKSTAR (Behroozi et al. 2013), with some parameters modified.

==== Kohei Hattori -- Energy conservatation ====

We use the variation of kinetic energy of stream particles along the stream, which can be interpreted as the variation of gravitational potential.
We can determine the flattening of the potential if the kinetic energy does not monotonically increase or decrease along the stream.