karpathy

""" Trains an agent with (stochastic) Policy Gradients on Pong. Uses OpenAI Gym. """
import numpy as np
import cPickle as pickle
import gym

# hyperparameters
H = 200 # number of hidden layer neurons
batch_size = 10 # every how many episodes to do a param update?
learning_rate = 1e-4
gamma = 0.99 # discount factor for reward

kylemcdonald

arturbekasov

• Deep Subspace Clustering Network, http://arxiv.org/abs/1709.02508v1
• Breaking the Nonsmooth Barrier: A Scalable Parallel Method for Composite Optimization, http://arxiv.org/abs/1707.06468v2
• Probabilistic Models for Integration Error in the Assessment of Functional Cardiac Models, http://arxiv.org/abs/1606.06841v4
• Dynamic Safe Interruptibility for Decentralized Multi-Agent Reinforcement Learning, http://arxiv.org/abs/1704.02882v2
• Parametric Simplex Method for Sparse Learning, http://arxiv.org/abs/1704.01079v1
• Group Sparse Additive Machine, http://arxiv.org/abs/1206.4673v1
• The Unreasonable Effectiveness of Structured Random Orthogonal Embeddings, http://arxiv.org/abs/1703.00864v2
• Inferring Generative Model Structure with Static Analysis, http://arxiv.org/abs/1709.02477v1
• On Structured Prediction Theory with Calibrated Convex Surrogate Losses, http://arxiv.org/abs/1703.02403v2
• Best of Both Worlds: Transferring Knowledge from Discriminative Learning to a Generative Visual Dialog Model

nakosung

# Borrowed from NAF. Might be plugged into DeepSets.

class SigmoidFlow(nn.Module):
    def __init__(self,K):
        super().__init__()
        self.W = nn.Parameter(torch.FloatTensor(K,1))
        self.b = nn.Parameter(torch.FloatTensor(1,K))
        self.alpha = nn.Linear(1,K)
        
    def forward(self,x):