A Quick Overview of Artificial Intelligence and Machine Learning
- 3. 1. The Origin: Understanding “Intelligence” 2. Key Ingredient I: Statistics & Data Analytics 3. Key Ingredient II: Optimization 4. Machine Learning 5. Artificial Neural Networks 6. Deep Learning 7. Other Topics and Tools 8. Research Examples 9. Challenges 3
- 7. 7 The first formal model of computational mechanisms of (artificial) neurons
- 8. 8 Multilayer perceptron (Rosenblatt 1958) Backpropagation (Rumelhart, Hinton & Williams 1986) Deep learning https://commons.wikimedia.org/wiki/File: Example_of_a_deep_neural_network.png
- 9. 9
- 10. 10 Norbert Wiener (This is where the word “cyber-” came from!)
- 11. ▪ Herbert Simon et al.’s “Logic Theorist” (1956) ▪ Functional programming, list processing (e.g., LISP (1955-)) ▪ Logic-based chatbots (e.g., ELIZA (1966)) ▪ Expert systems ▪ Fuzzy logic (Zadeh, 1965) 11
- 12. 12
- 13. Key Ingredient I: Statistics & Data Analytics 13
- 14. ▪ Descriptive statistics ▪ Distribution, correlation, regression ▪ Inferential statistics ▪ Hypothesis testing, estimation, Bayesian inference ▪ Parametric / non-parametric approaches 14 https://en.wikipedia.org/wiki/Statistics
- 15. ▪ Legendre, Gauss (early 1800s) ▪ Representing the behavior of a dependent variable (DV) as a function of independent variable(s) (IV) ▪ Linear regression, polynomial regression, logistic regression, etc. ▪ Optimization (minimization) of errors between model and data 15 https://en.wikipedia.org/wiki/Regression_analysis https://en.wikipedia.org/wiki/Polynomial_regression
- 16. ▪ Original idea dates back to 1700s ▪ Pearson, Gosset, Fisher (early 1900s) ▪ Set up hypothesis(-ses) and see how (un)likely the observed data could be explained by them ▪ Type-I error (false positive), Type-II error (false negative) 16 https://en.wikibooks.org/wiki/Statistics/Testing _Statistical_Hypothesis
- 17. ▪ Bayes & Price (1763), Laplace (1774) ▪ Probability as a degree of belief that an event or a proposition is true ▪ Estimated likelihoods updated as additional data are obtained ▪ Empowered by Markov Chain Monte Carlo (MCMC) numerical integration methods (Metropolis 1953; Hastings 1970) 17 https://en.wikipedia.org/wiki/Bayes%27_theorem https://en.wikipedia.org/wiki/Markov_chain_Monte_Carlo
- 19. ▪ Legendre, Gauss (early 1800s) ▪ Find the formula that minimizes the sum of squared errors (residuals) analytically 19 https://en.wikipedia.org/wiki/Least_squares
- 20. ▪ Find local minimum of a function computationally ▪ Gradient descent (Cauchy 1847) and its variants ▪ More than 150 years later, this is still what modern AI/ML/DL systems are essentially doing!! ▪ Error minimization 20 https://commons.wikimedia.org/wiki/File: Gradient_descent.gif
- 21. ▪ Extensively studied and used in Operations Research ▪ Practical optimization algorithms under various constraints 21 https://en.wikipedia.org/wiki/Linear_programming https://en.wikipedia.org/wiki/Integer_programming https://en.wikipedia.org/wiki/Floyd%E2%80%93Wa rshall_algorithm
- 22. ▪ Original idea by Turing (1950) ▪ Genetic algorithm (Holland 1975) ▪ Genetic programming (Cramer 1985, Koza 1988) ▪ Differential evolution (Storn & Price 1997) ▪ Neuroevolution (Stanley & Miikkulainen 2002) 22 https://becominghuman.ai/my-new-genetic-algorithm-for-time-series-f7f0df31343d https://en.wikipedia.org/wiki/Genetic_programming
- 23. ▪ Ant colony optimization (Dorigo 1992) ▪ Particle swarm optimization (Kennedy & Eberhart 1995) ▪ And various other metaphor-based heuristic algorithms https://en.wikipedia.org/wiki/List_of_metaphor-based_metaheuristics 23 https://en.wikipedia.org/wiki /Ant_colony_optimization_al gorithms https://en.wikipedia.org/wiki /Particle_swarm_optimizati on
- 25. ▪ Unsupervised learning ▪ Find patterns in the data ▪ Supervised learning ▪ Find patterns in the input-output mapping ▪ Reinforcement learning ▪ Learn the world by taking actions and receiving rewards from the environment 25
- 26. ▪ Clustering ▪ k-means, agglomerative clustering, DBSCAN, Gaussian mixture, community detection, Jarvis Patrick, etc. ▪ Anomaly detection ▪ Feature extraction/selection ▪ Dimension reduction ▪ PCA, t-SNE, etc. 26 https://reference.wolfram.com/language/ref/FindClusters.html https://commons.wikimedia.org/wiki/File:T-SNE_and_PCA.png
- 27. ▪ Regression ▪ Linear regression, Lasso, polynomial regression, nearest neighbors, decision tree, random forest, Gaussian process, gradient boosted trees, neural networks, support vector machine, etc. ▪ Classification ▪ Logistic regression, decision tree, gradient boosted trees, naive Bayes, nearest neighbors, support vector machine, neural networks, etc. 27 https://scikit-learn.org/stable/auto_examples/classification/plot_classifier_comparison.html https://scikit-learn.org/stable/auto_examples/ model_selection/plot_underfitting_overfitting.html
- 28. ▪ Environment typically formulated as a Markov decision process (MDP) ▪ State of the world + agent’s action → next state of the world + reward ▪ Monte Carlo methods ▪ TD learning, Q-learning 28 https://en.wikipedia.org/wiki/Markov_decision_process
- 30. ▪ Hopfield (1982) ▪ A.k.a. “attractor networks” ▪ Fully connected networks with symmetric weights can recover imprinted patterns from imperfect initial conditions ▪ “Associative memory” Input Output 30 https://github.com/nosratullah/hopfieldNeuralNetwork
- 31. ▪ Hinton & Sejnowski (1983), Hinton & Salakhutdinov (2006) ▪ Stochastic, learnable variants of Hopfield networks ▪ Restricted (bipartite) Boltzmann machine was at the core of the HS 2006 Science paper that ignited the current boom of “Deep Learning” 31 https://en.wikipedia.org/wiki/Boltzmann_machine https://en.wikipedia.org/wiki/Restricted_Boltzmann_machine
- 32. ▪ Multilayer perceptron (Rosenblatt 1958) ▪ Backpropagation (Werbos 1974; Rumelhart, Hinton & Williams 1986) ▪ Minimization of errors by gradient descent method ▪ Note that this is NOT how our brain learns ▪ “Vanishing gradient” problem 32 Computation Error correction Input Output
- 33. ▪ Rumelhart, Hinton & Williams (1986) (again!) ▪ Feed-forward ANNs that try to reproduce the input ▪ Smaller intermediate layers → dimension reduction, feature learning ▪ HS 2006 Science paper also used restricted Boltzmann machines as stacked autoencoders 33 https://towardsdatascience.com/applied-deep-learning-part-3- autoencoders-1c083af4d798 https://doi.org/10.1126/science.1127647
- 34. ▪ Hopfield (1982); Rumelhart, Hinton & Williams (1986) (again!!) ▪ ANNs that contain feedback loops ▪ Have internal states and can learn temporal behaviors of any long- term dependencies ▪ With practical problems in vanishing or exploding long-term gradients 34 https://commons.wikimedia.org/wiki/File:Neuronal-Networks- Feedback.png https://en.wikipedia.org/wiki/Recurrent_neural_network h o V nfold t 1 ht 1 ot 1 t ht ot t+1 ht+1 ot+1 V V V V ... ...
- 35. ▪ Hochreiter & Schmidhuber (1997) ▪ An improved neural module for RNNs that can learn long- term dependencies effectively ▪ Vanishing gradient problem resolved by hidden states and error flow control ▪ “The most cited NN paper of the 20th century” 35
- 36. ▪ Self-organizing map (Kohonen 1982) ▪ Neural gas (Martinetz & Schulten 1991) ▪ Spiking neural networks (1990s-) ▪ Reservoir computing (random RNNs; 2000s-) etc… 36 https://en.wikipedia.org/wiki/ Self-organizing_map https://doi.org/10.1016/j.neucom. 2019.10.104 https://doi.org/10.1515/nanoph-2016-0132
- 37. Deep Learning 37
- 38. ▪ Ideas originally around since the beginning of ANNs ▪ Became feasible and popular in 2010s because of: ▪ Huge increase in computational power thank to GPUs ▪ Wide availability of training data over the Internet 38 https://commons.wikimedia.org/wiki/File:Example_of_a_deep_neural_network.png https://www.techradar.com/news/computing-components/graphics-cards/best-graphics-cards-1291458
- 39. ▪ Fukushima (1980), Homma et al. (1988), LeCun et al. (1989, 1998) ▪ DNNs with convolution operations between layers ▪ Layers represent spatial (and/or temporal) patterns ▪ Many great applications to image/video/time series analyses 39 https://towardsdatascience.com/a-comprehensive-guide-to- convolutional-neural-networks-the-eli5-way-3bd2b1164a53 https://cs231n.github.io/convolutional-networks/
- 40. 40 https://arxiv.org/abs/1412.6572 https://en.wikipedia.org/wiki/Generative_ adversarial_network ▪ Goodfellow et al. (2014a,b) ▪ DNNs are vulnerable against adversarial attacks ▪ Utilize it to create co- evolutionary systems of generator and discriminator https://commons.wikimedia.org/wiki/File:A-Standard-GAN-and-b-conditional-GAN-architecturpn.png
- 41. ▪ Scarselli et al. (2008), Kipf & Welling (2016) ▪ Non-regular graph structure used as network topology within each layer of DNN ▪ Applications to graph- based data modeling, e.g, social networks, molecular biology, etc. 41 https://tkipf.github.io/graph-convolutional-networks/ https://towardsdatascience.com/how-to-do-deep-learning-on- graphs-with-graph-convolutional-networks-7d2250723780
- 42. ▪ Vaswani et al. (2017) ▪ DNNs with self-attention mechanism for natural language processing (NLP) ▪ Enhanced parallelizability leading to shorter training time than LSTM ▪ BERT (2018) for Google search ▪ Massive language models: GPT-3 (2020), Google's Switch Transformer (2021), etc. 42 https://arxiv.org/abs/1706.03762
- 43. 43 OpenAI GPT-3 / DALL-E https://www.theguardian.com/commentisfree/2020/sep/08/robot-wrote-this- article-gpt-3
- 45. 45 Time series analysis • Autoregression, ARMA/ARIMA, time series embedding, phase space reconstruction, etc. Natural language processing (NLP) • Classic syntactic/semantic approaches Information theory • Entropy, mutual information Computation theory • Automata, computational complexity
- 46. 46 Brain/neuroscience, cognitive science Complex systems and networks Robotics and control Consciousness, sentience, self
- 49. 49 Zamani Esfahlani, F. et al. (2018). A network-based classification framework for predicting treatment response of schizophrenia patients. Expert Systems with Applications, 109, 152-161. https://doi.org/10.1016/j.eswa.2018.05.005 Graduate Award for Excellence in Research (2018)
- 50. 50 Cao, Y., et al. (2022). Visualizing collective idea generation and innovation processes in social networks. IEEE Transactions on Computational Social Systems. https://doi.org/10.1109/TCSS.2022.3184628
- 51. 51 Dong, Y. et al. (2021). Utterance clustering using stereo audio channels. Computational Intelligence and Neuroscience, 2021, 6151651. https://doi.org/10.1155/2021/ 6151651
- 52. 52 Sayama, H. (2022). Social fragmentation transitions in large-scale adaptive social network simulations, Proceedings of the 14th International Conference on Parallel Processing and Applied Mathematics (PPAM 2022) / 7th Workshop on Complex Collective Systems, Springer, in press. https://arxiv.org/abs/2205.10489
- 53. Challenges 53
- 54. 54
- 55. 55
- 57. 57 Fall 2020: “How to safely reopen the campus”
- 58. 58
- 60. Are We Getting Any Closer to the Understanding of True “Intelligence"? 60
- 61. ▪ Don’t get drowned in the vast ocean of methods and tools ▪ Hundreds of years of history ▪ Buzzwords and fads keep changing ▪ Keep the big picture in mind – focus on what the real problem is and how you will solve it ▪ Being able to develop unique, original, creative solutions is key to differentiate your intelligence from AI/machines 61
- 62. 62 @hirokisayama