This paper introduces a new AI task - Embodied Question Answering. The goal of this task for an agent is to be able to answer the question by observing the environment through a single egocentric RGB camera while being able to navigate inside the environment. The agent has 4 natural modules:
https://i.imgur.com/6Mjidsk.png
1. **Vision**. 224x224 RGB images are processed by CNN to produce a fixed-size representation. This CNN is pretrained on pixel-to-pixel tasks such as RGB reconstruction, semantic segmentation, and depth estimation.

2. **Language**. Questions are encoded with 2-layer LSTMs with 128-d hidden states. Separate question encoders are used for the navigation and answering modules to capture important words for each module.

3. **Navigation** is composed of a planner (forward, left, right, and stop actions) and a controller that executes planner selected action for a variable number of times. The planner is LSTM taking hidden state, image representation, question, and previous action. Contrary, a controller is an MLP with 1 hidden layer which takes planner's hidden state, action from the planner, and image representation to execute an action or pass the lead back to the planner.

4. **Answering** module computes an image-question similarity of the last 5 frames via a dot product between image features (passed through an fc-layer to align with question features) and question encoding. This similarity is converted to attention weights via a softmax, and the attention-weighted image features are combined with the question features and passed through an answer classifier. Visually this process is shown in the figure below. https://i.imgur.com/LeZlSZx.png

[Successful results](https://www.youtube.com/watch?v=gVj-TeIJfrk) as well as [failure cases](https://www.youtube.com/watch?v=4zH8cz2VlEg) are provided.

Generally, this is very promising work which literally just scratches the surface of what is possible. There are several constraints which can be mitigated to push this field to more general outcomes. For example, use more general environments with more realistic graphics and broader set of questions and answers.

This reinforcement learning paper starts with the constraints imposed an engineering problem - the need to scale up learning problems to operate across many GPUs - and ended up, as a result, needing to solve an algorithmic problem along with it. 

In order to massively scale up their training to be able to train multiple problem domains in a single model, the authors of this paper implemented a system whereby many “worker” nodes execute trajectories (series of actions, states, and reward) and then send those trajectories back to a “learner” node, that calculates gradients and updates a central policy model. However, because these updates are queued up to be incorporated into the central learner, it can frequently happen that the policy that was used to collect the trajectories is a few steps behind from the policy on the central learner to which its gradients will be applied (since other workers have updated the learner since this worker last got a policy download). This results in a need to modify the policy network model design accordingly. 

IMPALA (Importance Weighted Actor Learner Architectures) uses an “Actor Critic” model design, which means you learn both a policy function and a value function. The policy function’s job is to choose which actions to take at a given state, by making some higher probability than others. The value function’s job is to estimate the reward from a given state onward, if a certain policy p is followed. The value function is used to calculate the “advantage” of each action at a given state, by taking the reward you receive through action a (and reward you expect in the future), and subtracting out the value function for that state, which represents the average future reward you’d get if you just sampled randomly from the policy from that point onward. The policy network is then updated to prioritize actions which are higher-advantage. If you’re on-policy, you can calculate a value function without needing to explicitly calculate the probabilities of each action, because, by definition, if you take actions according to your policy probabilities, then you’re sampling each action with a weight proportional to its probability. However, if your actions are calculated off-policy, you need correct for this, typically by calculating an “importance sampling” ratio, that multiplies all actions by a probability under the desired policy divided by the probability under the policy used for sampling. This cancels out the implicit probability under the sampling policy, and leaves you with your actions scaled in proportion to their probability under the policy you’re actually updating. IMPALA shares the basic structure of this solution, but with a few additional parameters to dynamically trade off between the bias and variance of the model. 

The first parameter, rho, controls how much bias you allow into your model, where bias here comes from your model not being fully corrected to “pretend” that you were sampling from the policy to which gradients are being applied. The trade-off here is that if your policies are far apart, you might downweight its actions so aggressively that you don’t get a strong enough signal to learn quickly. However, the policy you learn might be statistically biased. Rho does this by weighting each value function update by: 

https://i.imgur.com/4jKVhCe.png

where rho-bar is a hyperparameter. If rho-bar is high, then we allow stronger weighting effects, whereas if it’s low, we put a cap on those weights. 

The other parameter is c, and instead of weighting each value function update based on policy drift at that state, it weights each timestep based on how likely or unlikely the action taken at that timestep was under the true policy. 

https://i.imgur.com/8wCcAoE.png

Timesteps that much likelier under the true policy are upweighted, and, once again, we use a hyperparameter, c-bar, to put a cap on the amount of allowed upweighting. Where the prior parameter controlled how much bias there was in the policy we learn, this parameter helps control the variance - the higher c-bar, the higher the amount of variance there will be in the updates used to train the model, and the longer it’ll take to converge. 

Lecuyer et al. propose a defense against adversarial examples based on differential privacy. Their main insight is that a differential private algorithm is also robust to slight perturbations. In practice, this amounts to injecting noise in some layer (or on the image directly) and using Monte Carlo estimation for computing the expected prediction. The approach is compared to adversarial training against the Carlini+Wagner attack.

Also find this summary at [davidstutz.de](https://davidstutz.de/category/reading/).

If one is a Bayesian he or she best expresses beliefs about next observation $x_{n+1}$ after observing $x_1, \dots, x_n$ using the **posterior predictive distribution**: $p(x_{n+1}\vert x_1, \dots, x_n)$. Typically one invokes the de Finetti theorem and assumes there exists an underlying model $p(x\vert\theta)$, hence $p(x_{n+1}\vert x_1, \dots, x_n) = \int p(x_{n+1} \vert \theta) p(\theta \vert x_1, \dots, x_n) d\theta$, however this integral is far from tractable in most cases. Nevertheless, having tractable posterior predictive is useful in cases like few-shot generative learning where we only observe a few instances of a given class and are asked to produce more of it.

In this paper authors take a slightly different approach and build a neural model with tractable posterior predictive distribution $p(x_{n+1} | x_1, \dots, x_n)$ suited for complex objects like images. In order to do so the authors take a simple model with tractable posterior predictive $p(z_{n+1} | z_1, \dots, z_n)$ (like a Gaussian Process, but not quite) and use it as a latent code, which is obtained from observations using an analytically inversible encoder $f$. This setup lets you take a complex $x$ like an image, run it through $f$ to obtain $z = f(x)$ -- a simplified latent representation for which it's easier to build joint density of all possible representations and hence easier to model the posterior predictive. By feeding latent representations of $x_1, \dots, x_n$ (namely, $z_1, \dots, z_n$) to the posterior predictive $p(z_{n+1} | f(x_1), \dots, f(x_n))$ we obtain obtain a distribution of latent representations that are coherent with those of already observed $x$s. By sampling $z$ from this distribution and running it through $f^{-1}$ we recover an object in the observation space, $x_\text{pred} = f^{-1}(z)$ -- a sample most coherent with previous observations.

Important choices are:

* Model for latent representations $z$: one could use Gaussian Process, however authors claim it lacks some helpful properties and go for a more general [Student-T Process](http://www.shortscience.org/paper?bibtexKey=journals/corr/1402.4306). Then authors assume that each component of $z$ is a univariate sample from this process (and hence is independent from other components)
* Encoder $f$. It has to be easily inversible and have an easy-to-evaluate Jacobian (the determinant of the Jacobi matrix). The former is needed to perform decoding of predictions in latent representations space and the later is used to efficiently compute a density of observations $p(x_1, \dots, x_n)$ using the standard change of variables formula $$p(x_1, \dots, x_n) = p(z_1, \dots, z_n) \left\vert\text{det} \frac{\partial f(x)}{\partial x} \right\vert$$The architecture of choice for this task is [RealNVP](http://www.shortscience.org/paper?bibtexKey=journals/corr/1605.08803)

Done this way, it's possible to write out the marginal density $p(x_1, \dots, x_n)$ on all the observed $x$s and maximize it (as in the Maximum Likelihood Estimation). Authors choose to factor the joint density in an auto-regressive fashion (via the chain rule) $$p(x_1, \dots, x_n) = p(x_1) p(x_2 \vert x_1) p(x_3 \vert x_1, x_2) \dots p(x_n \vert x_1, \dots, x_{n-1}) $$with all the conditional marginals $p(x_i \vert x_1, \dots, x_{i-1})$ having analytic (student t times the jacobian) density -- this allows one to form a fully differentiable recurrent computation graph whose parameters (parameters of Student Processes for each component of $z$ + parameters of the encoder $f$) to be learned using any stochastic gradient method.

https://i.imgur.com/yRrRaMs.png

Frankle and Carbin discover so-called winning tickets, subset of weights of a neural network that are sufficient to obtain state-of-the-art accuracy. The lottery hypothesis states that dense networks contain subnetworks – the winning tickets – that can reach the same accuracy when trained in isolation, from scratch. The key insight is that these subnetworks seem to have received optimal initialization. Then, given a complex trained network for, e.g., Cifar, weights are pruned based on their absolute value – i.e., weights with small absolute value are pruned first. The remaining network is trained from scratch using the original initialization and reaches competitive performance using less than 10% of the original weights. As soon as the subnetwork is re-initialized, these results cannot be reproduced though. This suggests that these subnetworks obtained some sort of “optimal” initialization for learning.

Also find this summary at [davidstutz.de](https://davidstutz.de/category/reading/).