Keith Trnka ktrnka

RFP Writing Guidelines

Purpose: Brief guidelines for effective RFPs based on successful patterns.

Prefer brevity - Avoid AI-generated verbose content and repetition
Problem-first - Document problems clearly before jumping to solutions
Dual-purpose - Planning documents that become reference materials post-implementation
Complement codebase - RFPs provide context, code provides implementation details

EVTF

FTL

percent of hour in FTL_EARTH, NADIR, flagcomms, etc. Not all of the fields but most

	def prepare_time_matrix(X, time_steps=5, fill_value=None):
	if time_steps < 1:
	raise ValueError("time_steps must be 1 or more")

	assert isinstance(X, numpy.ndarray)
	time_shifts = [X]
	time_shifts.extend(numpy.roll(X, t, axis=0) for t in range(1, time_steps))
	time_shifts = reversed(time_shifts)

	X_time = numpy.dstack(time_shifts)

	class StackedEnsembleClassifier(sklearn.base.BaseEstimator, sklearn.base.ClassifierMixin):
	"""
	Ensemble of classifiers. Each classifier must have predict_proba and the head classifier is trained
	to predict the output based on the individual classifier outputs. Stratified k-fold cross-validation
	is used to get probabilities on held-out data.
	"""
	def __init__(self, classifiers, head_classifier, num_folds=3):
	self.classifiers = classifiers
	self.head_classifier = head_classifier
	self.num_folds = num_folds

	from __future__ import unicode_literals
	import logging
	import numpy
	import keras.models
	import keras.layers.core
	import keras.regularizers
	import sklearn.metrics
	import sklearn.base
	import keras.constraints
	import keras.layers.noise