Skip to content

Latest commit

 

History

History
45 lines (36 loc) · 2.31 KB

parameters.md

File metadata and controls

45 lines (36 loc) · 2.31 KB

Parameters

This contains a list of parameters used in the model, alongisde their default values and meaning.

DRL parameters

  • gamma = 0.9999 - (Discount rate for advantage estimation and reward discounting)
  • n_doseOptions = 2 - (Agent can choose 2 different intensities of chemotherapy)
  • base = 0.1 - (Base reward given per timestep survived)
  • hol = 0.05 - (Additional reward for a treatment holiday)
  • day_interval = 1 - (Interval between decision points (time step for DRL evaluation of ODE model))
  • max_episode_length = 3000 - (Maximum length to run for during training (note: this currently is not functional))
  • num_workers* = 4 - (Multiprocessing.cpu_count(), sets workers to the number of available CPU threads)
  • model_path = "./models"
  • results_path = "./results" - (Directory to save output .csv in)
  • logging_interval = 100 - (Save the state of the network every logging_interval patients)
  • model_loaded_name ='6260_patients_MonJul191846322021'
  • verbose = 0 - (Boolean to determine level of detail printed to terminal during runtime)

Tumour model parameters

  • n_replicates = 5 - (Number of times to repeat ODE model)
  • n0 = 0.75 - (Initial size of tumour)
  • rFrac = 0.001 - (Initial proportion of susceptible cells)
  • paramDic = {} - (See section below)

Parameter Dictionary

  • rS = 0.027 - (Birth rate of susceptible cells)
  • rR = 0.027 - (Birth rate of resistant cells)
  • dS = 0.0 - (Death rate of susceptible cells)
  • dR = 0.0 - (Death rate of resistant cells)
  • dD = 1.5 - (Drug induced death rate of susceptible cells)
  • k = 1.0 - (Carrying capacity of tumour environment)
  • D = 0 - (Drug concentration in environment)
  • theta = 1 - (Scale factor for mapping cell counts to observed fluorescent area - assumes a cell radius of 10uM)
  • DMax = 1.0 - (Max doseage given to patient)
  • S0 = n0 * (1 - rFrac) - (Initial size of susceptible portion of tumour)
  • R0 = (n0 * rFrac) - (Initial number of resistant portion of tumour)
  • punish = -0.1 - (Punishment for exceeding 20% limit on tumour growth)
  • learning_rate = 1e-4 - (Learning rate for Adam Optimiser)

Also includes n0, rFrac, base, hol, and day_interval as interval, which are described above.

* These parameters are used in training only.