28

Here's a snippet from an article by Gary Marcus In particular, they showed that standard deep learning nets often fall apart when confronted with common stimuli rotated in three dimensional space into unusual positions, like the top right corner of this figure, in which a schoolbus is mistaken for a snowplow: . . . Mistaking an ...


19

In theory, most neural networks can approximate any continuous function on compact subsets of $\mathbb{R}^n$, provided that the activation functions satisfy certain mild conditions. This is known as the universal approximation theorem (UAT), but that should not be called universal, given that there are a lot more discontinuous functions than continuous ones,...


16

Explainable AI and model interpretability are hyper-active and hyper-hot areas of current research (think of holy grail, or something), which have been brought forward lately not least due to the (often tremendous) success of deep learning models in various tasks, plus the necessity of algorithmic fairness & accountability. Here are some state of the ...


15

In our deep learning lecture, we discussed the following example (from Unmasking Clever Hans predictors and assessing what machines really learn (2019) by Lapuschkin et al.). Here the neural network learned a wrong way to identify a picture, i.E by identifying the wrong "relevant components". In the sensitivity maps next to the pictures, we can see that the ...


9

The two tech reports below both call RNNs explicitly "recurrent net(work)s". Rumelhart, David E; Hinton, Geoffrey E, and Williams, Ronald J (Sept. 1985). Learning internal representations by error propagation. Tech. rep. ICS 8504. San Diego, California: Institute for Cognitive Science, University of California. Jordan, Michael I. (May 1986). ...


7

There are at least three free courses. MIT 6.S099: Artificial General Intelligence, organized by Lex Fridman, is a series of lessons and talks primarily given by a diverse set of guest appearances, such as Josh Tenenbaum (researcher and professor in computational cognitive science), Nate Derbinsky (who gives a lesson on cognitive architectures, Soar, etc.),...


7

Although I have only partially read (or not read at all) some of the following resources and some of these resources may not cover more advanced topics than the ones presented in the book you are reading, I think they can still be useful for your purposes, so I will share them with you. I would also like to note that if you understand the contents of the ...


5

This is more in the direction of 'what kind of problems can be solved by neural networks'. In order to train a neural network you need a large set of training data which is labelled with correct/ incorrect for the question you are interested in. So for example 'identify all pictures that have a cat on them' is very suitable for neural networks. On the other ...


5

Yes, there are many, actually. A Google search turned this paper Artificial Neural Networks in Medical Diagnosis (2011) by Al-Shayea up. Not only are they used in disease diagnosis, but even with things like prescribing medicines. In fact, the top project for a hackathon at my school analysed thousands of research articles, and took a patient's medication ...


5

If you're interested in the theory behind Double Q-learning (not deep!), the reference paper would be Double Q-learning by Hado van Hasselt (2010). As for Double deep Q-learning (also called DDQN, short for Double Deep Q-networks), the reference paper would be Deep Reinforcement Learning with Double Q-learning by Van Hasselt et al. (2016), as pointed out ...


5

There is Google Research Football, which is an open-source platform to develop reinforcement learning algorithms to play a game similar to FIFA or PES, although the football simulation is not as realistic as the current versions of FIFA or PES. You can play this game against different RL agents (e.g. DQN or IMPALA) and, of course, you can even develop your ...


5

I will be starting my PhD in natural language processing in a few days and this is very similar to my proposed topic. It's an open problem that ties NLP and AI into philosophy of science and epistemology and is, I think, extremely interesting. I say all this to drive home the point that this is not a simple problem. Two major theoretical concerns come to my ...


5

In general, calculation of distance between camera and object is impossible if you don't have further scene dependent information. To my knowledge you have 3 options: Stereo Vision If you have 2 cameras looking at the same scene from a different point of view you can calculate the distance with classical Computer Vision algorithms. This is called stereo ...


5

Given that policies are probability distributions, in principle, you can use any metric or measure of distance that can be used to compare two probability distributions. (Note that notions of distance are not necessarily metrics in a mathematical sense). A common measure is the KullbackÔÇôLeibler divergence (which is not a metric, in a mathematical sense, ...


4

I don't know if it might be of use, but many areas of NLP are still hard to tackle, and even if deep models achieve the state of the art results, they usually beat baseline shallow models by very few percentage points. One example that I've had the opportunity to work on is stance classification 1. In many datasets, the best F score achievable is around 70%....


4

[Disclaimer: I work for a company that provides a platform for developing conversational AI systems] The platform used by the company I work for has a sentiment analysis component, so you can recognise if the user input expresses certain emotions. The dialogues are encoded in 'flows', which are graphs with an initial trigger consisting of output nodes and ...


4

First of all, efficiency and convergence are two different things. There's also the rate of convergence, so an algorithm may converge faster than another, so, in this sense, it may be more efficient. I will focus on the proof that policy evaluation (PE) converges. If you want to know about its efficiency, maybe ask another question, but the proof below also ...


4

There are actually quite a few. Personally I would say these courses have high quality and strong focus on practice: Standford computer vision cs231. Check the assignments materials on this page. This course has good explanation/exercises of how generally neural nets and backprop works. Fastai course notebooks. You can listen to the lectures as well, but ...


4

The book Grokking Deep Learning, by Andrew Trask (a PhD student at Oxford University and a research scientist at DeepMind), a wonderful, clean, and plain-English discussion of the basic mechanics that go on under the hood of neural networks - from data flow to updating of weights. It is written without a slant on normally-wonky math, the concepts are ...


4

A global race is underway to discover a vaccine, drug, or combination of treatments that can disrupt the SARS-CoV-2 virus. The problem is, there are more than a billion such molecules. A researcher would conceivably want to test each one against the two dozen or so proteins in SARS-CoV-2 to see their effects. Such a project could use every wet lab in the ...


4

These are known as adversarial attacks, and the specific examples that are misclassified are known as adversarial examples. There is a reasonably large body of work on finding adversarial examples, and on making CNNs more robust (i.e. less prone to these attacks). An example is the DeepFool algorithm, which can be used to find perturbations of data which ...


4

The visualisation can be found in The need for small learning rates on large problems. This paper by D. Randall Wilson and Tony R. Martinez from 2001 investigates the role of learning rates in gradient descent algorithms. In general, different algorithms assign different meaning to the same word 'learning rate'. For example, the learning rate in a gradient ...


4

The 2015 article Cyclical Learning Rates for Training Neural Networks by Leslie N. Smith gives some good suggestions for finding an ideal range for the learning rate. The paper's primary focus is the benefit of using a learning rate schedule that varies learning rate cyclically between some lower and upper bound, instead of trying to choose a single fixed ...


3

A checkerboard with missing squares is impossible for a neural network to learn the missing color. The more it learns on training data, the worse it does on test data. See e.g. this article The Unlearnable Checkerboard Pattern (which, unfortunately, is not freely accessible). In any case, it should be easy to try out yourself that this task is difficult.


3

From my experience in industry, a lot of data science (operating on customer information, stored in a database) is still dominated by decision trees and even SVMs. Although neural networks have seen incredible performance on "unstructured" data, like images and text, there still do not appear to be great results extending to structured, tabular data (yet). ...


3

Neural networks seem to have a great deal of difficulty handling adversarial input, i.e., inputs with certain changes (often imperceptible or nearly imperceptible by humans) designed by an attacker to fool them. This is not the same thing as just being highly sensitive to certain changes in inputs. Robustness against wrong answers in that case can be ...


3

"Modern" Guarantees for Feed-Forward Neural Networks My answer will complement nbro's above, which gave a very nice overview of universal approximation theorems for different types of commonly used architectures, by focusing on recent developments specifically for feed-forward networks. I'll try an emphasis depth over breadth (sometimes called ...


3

In principle, yes, you can also evolve the genetic algorithm (or, in general, evolutionary algorithm), i.e. you can evolve its operations (such as the mutation and cross-over) and hyper-parameters (such as the size of the population or mutation rate). For example, you could use genetic programming to evolve the cross-over operation of a genetic algorithm. ...


3

You can split each polygon into a collection of triangles and sum up the areas. Not really sure why you would bother with ML. Anyway if you approximates these polygons as images you could maybe train a CNN. Look at the image classification networks which provide bounding boxes.


3

There are several common deep reinforcement algorithms and models apart from deep Q networks (or deep Q learning). I will list some of them below (along with a link to the paper that introduced them), but note that some of these may not be state-of-the-art (at least, not anymore, and it's likely that all of these will be replaced in the future). Double DQN (...


Only top voted, non community-wiki answers of a minimum length are eligible