Learning episode

The full cycle during which the model completes a large number of trials to reach criterion performance is called a learning episode. For each episode, the model begins with a new set of connection weights and a new ordering of the input vectors. Each of the main components of the model are described briefly below. 

One of the more striking results of the distribution analysis is the large difference between the highest and lowest numbers of trials needed to reach criterion. The most efficient learning episode concluded in 3,900 trials whereas the slowest learning episode required 6,000 trials. This is a large difference it represents an increase of more than 50%. 

In the second experiment, the conditions were reversed. The weights were held constant and item presentation was allowed to vary. One set of weights was generated randomly in the usual way and was then used to initialize 10 new learning episodes. The order in which the input items were presented to the model was randomly determined for each episode, as in the original model. Thus, for these 10 episodes, the model always began with the same pattern of weights. Only the order of input presentation varied. 

Figure 14.3 A comparison of learning episodes under varying initial orders and varying input weights...

In contrast, consider the lower part of Figure 14.3, which summarizes the 10 learning episodes for which the initial weights were held constant and for which the input order was allowed to vary. No matter the order of input, the resulting configurations were very similar. The rows are virtually identical. Look, for example, at the patterns of connections shown under the VARY output unit. Almost all of the 10 episodes resulted in the same configuration. 

HINT In order to remember the nine diagnostic symptoms for a major depressive episode, learn the following mnemonic Depression = SIG E CAPS (depression, sleep, interest, guilt, energy, concentration, appetite, psychomotor, suicide). 

Dere, E., Souza-Silva, M. A., Topic, B., Spieler, R. E., Haas, H. L. and Huston, J. P. Histidine-decarboxylase knockout mice show deficient nonreinforced episodic object memory, improved negatively reinforced water-maze performance, and increased neo- and ventro-striatal dopamine turnover. Learn. Mem. 10 510-519, 2003. 

After the acute psychotic episode has resolved, the patient typically has residual features (e.g., anxiety, suspiciousness, lack of volition, lack of motivation, poor insight, impaired judgment, social withdrawal, difficulty in learning from experience, and poor self-care skills). Patients often have comorbid substance abuse and are nonadherent with medications. 

Episodic memory and learning, involving both explicit verbal and visual memory performance in contrast, recognition and implicit memory tasks appear to be spared. 

Focus your reflections just on those episodes that are relevant to particular learning outcomes which you feel you need in order to round off your development at this time. 

---